A Compositional Semantics for Focusing Subjuncts Daniel C. Lyons ‘Technical Report CSRI-234 “Angast 1989 (Computer Systems Reser Instat ‘University of Toreato MSS 1A4 ‘Toe Computr Systems Research Insts (CSRD is an inerdiseptnay group formed to conduct research ‘404 development relevant wo compte systems and their application. Ikan tine within the cay of Applied Science and Engineering, and the Faculty of Art and Science, at the Univers of Teeees woke ‘supponed in prtby the Naural Sciences and Engineering Research Councllof Canada,A Compositional Semantics for Focusing Subjuncts Daulel C. Lyons Department of Computer Science University of Toronto ‘Torono, Ontario, Canada ‘August 1989 A Thesie submitted in conformity with the requirements {forthe degree of Master of Science in the University of Toronto Copyright ©1989 Daniel C. LyonsAbstract Focusing subjunets such as only, even, and also are a class of sentence adverbials that ‘poses a challenge to a compositional semantics. The relationship between the focusing ‘subjanct and ita focus, an item that it draws attention to in the sentence and that is ‘usually accented, is particularly problematic, as are the preeuppositional aspects of the ‘meaning conveyed by focusing subjuncts. The meaning of focusing subjuncts has in the past ‘been expresed in first-order or higher-order logics, which are computationally impractical repreventatiens of meaning. ‘This research overcomes these problems by introducing a semantics for focusing eub- junets that is compositional, computationally practical, able to dietinguith between asserted ‘and non-asserted meaning, sensitive to intonation, and cross-categotial. The eemantics are ‘made compositional and computationally practical by adapting the system of Mats Rooth, ‘which employs a higher-order loge. A frame-based meaning representation is used instead. ‘This in facilitated by the use of an relatively unconstrained syntactic treatment of focus. ing subjuncts, and by capturing their meaning with two operators on sentential semantic forms. A two-part semantic representation along the lies proposed by Lauri Karttenen and Stanley Peters allows non-asserted aapects of meaning to be captured. ‘The semantics are incorporated into the program Ing0, a semantic interpreter that uses 8 GPSG grammar modified to account for focusing subjuncts.Acknowledgments T would ke to thank my supervisor Graeme Hirst for the time, ofort, and suggestions he has given me over the past year and a half, and my sxpervisor Diane Horton for her ‘very significant help with this thesis I am indebted to Brendan Gillon for his thoughtful ‘comments, which he provided on quite short notice. I would alo like to thank Barb Brunson. for pointing me at some of the most useful literature on forusing subjuncts and for her advice on linguistic points. Iam grateful to Chuck Pilkington for his invaluable production ausitance, to Mark Ryan for beating a path for me to follow these past months, and to Carolyn Cho for her constant encouragement. 1 would ike to acknowledge my financial benefactor, the Natural Sciences and Engineer- ing Research Council, whose support made possible the completion of this research Finally, I thank my family for their moral support and encouragement during my past even years of university. Thanks, Mom, Dad, and Kathy.Contents 1 Introduction 1 1.1 Focusing subjuncte . Eee 1.2 Compositional semantics... . 5 121 Advantage of compositionality 6 1.22 Montague: PTQ .. . 7 123° Hirst: Absity 8 1.8. The semantics of focusing ubjncts 10 LA Goals of tly research... 2 14.1 Simplifying assumptions 1B 15 Some definitions. . 8 2 Previous research 7 2.1 Overviey ofthe literature . . ra 22 Laying the groundwork... .. 18 2.8 Association with focus . 20 24 Multi-part semantics * 25 Rooth .. , an 25:1 iii of eater theorce ey 25.2 Rooth's proposal . 9 26 NLU aywems that handle focusing sabjncto === sss a8 2.7 Differences in coverage . 3. Towards a semantics for focusing subjuncts 3.1 Resolution of important issues . ee 3.4.1 Meaning of evenand only... 2.2... aaa 313 2 214 Distsbution ofthe focusing subjuncts . - 3.1.5 The role of stress. . 3.1.6 Ambiguity 3.17 Poeus, the linguistic featurew CONTENTS 3.1.8 Multiple foc of focusing subjuncts 3.4.9 Maltiplefocasing subjancts . . Eat 41.1.0 Tatroduction of focusing ubjunci Into sentence structures... 81 3.2. Semantic framework 33 Syntactic framework - .. . 34 Summary of assumptions... . 341 Cholces of formal frameworks - 34.2. Simplifying assumptions . . 3.43 Linguistic points . . ‘A compositional semantics for focusing subjuncts 50 4.1 The syntax of focusing subjuncts . cee 59 441.1 Pocusin GPSG........ 6 ‘Tho other new features... esses a istrative grammar . oe Introduction of focusing subjonets 64 Poeudo-prepostions 4 42 The semantics of focusing eubjuncte . 65 42.1 Backgromd ..... 65 422, Dealls ofthe semantics. Sam 4.3. Frame representation semantics . 6 ‘The implementation SA The grammar... see eee ee ees Data structures Grammar miles 0... ‘Head Feature Convention . Control Agreement Principle 5.1.5 Foot Feature Principle . 52. Morphology and the lexicon 53. The semantic component 5.3.1 Data structures 5.32 Semantic rules . 54 Knowledge-base semantics... .. . 5.4.1 Implementing frame determiners... . . 5.42 Implementing the new semantic types . . 55 Limitations... ‘ 2 56 Examples .. i 98 Conclusions 108 6.1 Goals 105CONTENTS 62 Contributions to the fla 6.3 What was learned. 6.4 Further Revearch . . 65 Condusion vi 106 107 108 109 an 18List of Figures aa 12 2a aa 4a 42 43 44 45 46 5a 52 53. 54 Aaity type corenpondences ‘An Absity frame statement. ‘The EST organization of grammar Some GPSG category abbreviations. 1D rules for toy grammar... . Lexical entries for toy grammar . Revised ID rules... . ae ‘Compazison of two methods of translating only . . ‘This diagram commutes. . ab Semantic rule forthe attert feature. Category for Ross sleeps . . Atypical grammar rule. . Atypical lexical grammar rule. Parse tree for Ross washed a dog. B BeChapter 1 Introduction 1.1 Focusing subjuncts Focusing subjuncts such as only, even, and also are a subclass of the sentence-element class of adverblals, according to the grammar of Quirk etal. (2985) Focasing subjancts draw attention to a part of a sentence, called the focus of the focusing aubjunct, which ‘often represents ‘new’ information. They are usually realized by adverbe, but occasionally: by prepositional phrases. Simply and in particular are focusing subjuncte ofthe following, sentences (Cirk etal, 1985, §8.116): (2) Iwas simply (taking my dog fora walk)? (2) (The workers), in particular, are dissatisfied with the government, Focusing eubjuncts fill the semantic role of modality: they enphasize, approximate, or restrict their foci (see appendix A). They modify the force or treth value of a sentence, ‘especially with respect to its applicability to the focused item. Ia this thesis, only the ‘adverb realizations of focusing subjancts are considered, and theee lexical items themselves will henceforth be referred to as focusing subjuncts. ‘The pusist takes the term to refer ‘exclusively to a grammatical fonction or sentence role, rather than a part of speech, ‘Quirk et al. (1985, §8.116) divide focusing subjuncts into severe categories. Restric- tives indicate that the utterance is true ether exclusively or predominantly, of the item focused. Examples ofthe former kind, called exclusives, include jut and only. Examples of the later kind, called particularizers, are especially and main. Additives, euch at also, even, and too, indicate that the utterance is additionally trae ofthe part focused? sppendix A summaries Quis defen and tasonomy of adverbial ‘rote! Inte example weences inoughot! Ina Uen, worden small capitals ave inonatina tes, snd angle brats (ence the focus of «focusing subjanct. Unscrepae satenen are preceded by a ‘tei (*), aad entenen that ae scntraditory are preceded by & bush atk (#). in some ofthe trample sentencs in Chaper I, the focusing rubjunct (or candidate) alc, and ogee braces [] ‘eed to at ff the conttnnt to which the fcaingsubjuact adie A isting ofthe focsing subjunce i inclded in sppenix A 12 (CHAPTER 1. INTRODUCTION Focusing subjuncts are interesting and difficult to deal with, for several reasons, ‘+ The words that act as focusing subjancts are polysemous. In some of their senses they play a completely diferent role. ‘Focusing subjuncts can adjoin to many different types of constituent. They can occur at almost any syntactic position in a sentence, ‘Focusing subjancts can precede o follow the item that they focus (draw attention to), and need not be adjacent to this item. ‘© They may focus almost any constituent. ' Because of the above three points, itis dificult to determine the intended ayntactic argument (adjunct) and focus ofa focusing subunct. ' The location of ntonational stress has an important effect on the meaning of eentence containing a focusing subjunct. ‘Focusing subjuncts don’t ft into the simple slot-fllr semantics that seam adequate {for handling many other sentence elements (see Section 1.2.3). At best, their semantic effect is to transform the semantic representation of the contituent they modify in ‘some predictable compositional way (Hirst, 1987, p. 72). ‘They carry a lot of pragmatic “baggage”. It ia necessary to resort to talking about presuppositions and conversational and conventional implicatures (see Section 1.5 below) in order to fully explain the meaning of a focusing eubjunct. ‘The frst six ofthese problems are elaborated upon below. ‘The last two are addreted in Section 1.3. Polysemy The diferent senses of only in (8) are obviously semaatically very lose. Yet only (3.1) contains only in the sense in which it ina focusing subjunct. In (3.2), it is ‘ conjunction, and in (8.3), only acte as an intensifier, weakening the force of the verb joking. (9) 1 Shee charming only [to her wealthy clint. 2. He could have done it, only he didn’t make an efor. 43. Twas only joking. ‘Many words that can act as focusing subjuncte also have senses in which they do not. ‘These polysemous words include just, also, merely, too, and rather. Some of the different senses of foo and just are given in (4) and (5), respectively. Many of the sample sentences are from Quirk et al. (19885).11. FOCUSING suBJUNCTS (4) Uses of too: 1, Tbrought the wine and the bread foo. [a conjunet) 2, John has seen it too. [an additive focusing subjunct) 3, Too true, [an emphasizer eubjunct] 4. You did too, asin (4.3), a8 an anveverative) 5, Pm not too sure about that. (an intensifier subjunct] 6. A: He's clever. B: Too right, he i. fa in (4.5) in Australian Exglish] (5) oes of jut: 1, 1 just can’t understand it, fan emphasizer eubjunct] 2. You can get a B grade just for that anewer, [am exclusive restrictive focusing eubjunct) 3. Just exactly what do you expect? [as in (6.2), modifying another focusing subjunct, ezact] 4, Sho'l just be out for a few minutes. fan intensifier subjunct) 5. When I eaw her, she had just come back, [a time-relationship item eubjunct] Syntactic position Focusing subjuncts can occupy almost any position ins sentence, © 1. Boen Bob was there. (sentence inital] 2, 1 simply was taking my dog for a walk. [before aniliary) 3. Thad simply been taking my dog for a walk. [after fist auxiliary) 4, Tould have simply bean taking my dog fora walk. [Between second and third auxiliaries) 5. Ihave been simply taking my dog for a walk. [before main verb] 6. She has invited as wll some of her own family. (after verb] 7. We dida't buy any beer, either. [sentence final] 1t appears that focusing subjunets can adjoin to any maximal projection (an observation ‘due to Barbara Brunton, personal communication): wo 1. John introdnced only [np (Rt1)] to Sue. [NP] 2. Ross even [vp washed (the Doc). [VP] ‘8 John has seen it [pp (near his back DooR)] even. [PP] 4, Pm old but I'm also [yp (waPry)). [AP] ‘5. Ben [y if the agroement were si0Ng0], it wouldn't be binding. ($)4 CHAPTER 1. INTRODUCTION Position relative to their focus Focusing subjuncts are Sexible with respect to their ‘pnition, so mach so that maltiple positions are possible even fora given focus (Quirk et al., 1085, §8.110): (8) Pvenoticed the fox in my garden and... 4, John has seen it (near his back D0OR) even. Promiseuity of focus Focusing subjuncta can focus the subject, the object, a main or ausiliary verb, a complement, another adverbial, or an adjective phrase. (®) 1. At leat (ten worker) reported sick yesterday. [eubjet] 2. His father wouldnt give the money nor would he (umn) it. [main verb] 4 The old man simply (covtD) not have done the deed. [(modal) auxiary verb] 4, Twas simply (tang my dog fora wal). [predation] 8 I gave the dog a bone, and T gave the (at) one aswell. indirect objet} 6. He became happy, and he alzo became (niet). complement] 1. Lil be away (tomorrow) as well. [adverbial] 8, Shei charming only to her (WaaurHy) client. [adjective phrase] Ambiguity of focus Consider the following centence: (20) John could also see his wife from the doorway. Depending on where the stress fala when the sentence is spoken, also has the choice of focasing the main verb, come part of the predication, or the whole predication (Quirk etal. 1985, §8.127) ((11.) and (11.5) were noticed by Diane Horton, private communication). (21) 1. (Jou) could also see his wife from the doorway. [so could Tom] 2, John could lao (sux) bie wife from the docrway. [as well as being able to talk to her} 8, John could also see (a18) wife from the doorway. {at well at Tom's wife} 4, John could also see (hie wirE) from the doorway. [ae well as her brother] 55 John could also (eee hie Wire) from the doorway. [as well as do other things from the doorway]1.2. COMPOSITIONAL SEMANTICS 5 4. John cou alto eee his wife (fom the Doonway)- {os wll as from further inside the zoom) 1. John cou alao (ce his wife fom the noonway)- {he coal alo do other thing] Importance of intonation The sentence 10 is ambiguous among all ofthe interpreta- tions in 11 in written text, but not in spoken language. Intonation effects partly disam- Diguate the sentence. For example, only the readings (11.4) and (11.6) are possible ifthe ‘word wife is stressed intonationally. However, intonation alone does not indicate which of ‘these two sees ie Intended. 1.2 Compositional semantics Frege's Principle of Compositionality is often cited as a desirable feature of a system of semantics (Dowty, Wall and Peters, 1981, p. 8). As we shall ee in Section 1.3, focusing subjuncts present a challenge to a compositional semantics. Tn this section, I define what makes a semantics compositional, and explain the sense in which compositionality can be considered matter of degree. ‘Two compositional semantics are presented, Montague's (1073) PTQ system and Hirst’s (1987) Abeity. ‘The following two properties characterize compositional eemantice: (12) 1, Bach word and well-formed syntactic phrase mantic object. represented by a distinct se- 2, The semantic representation of a syntactic phrase is a eystematic function of, the representation of its constituent words and/or phrases. [Note that this definition assumes a syntax in which phrases are complex objects constructed from constituents according to some rules (phrase-structure rales). Ia a compositional semantic, the syntax drives the semantica. To each syntactic phrase construction rule there corresponds a semantic rule that specifies how the semantic objects of the constituents are (cystematically) comtined or composed to obtain a semantic object for the phrase. It is the word systematically in the definition that makes compostionality a scale rather ‘than an absolute property. ‘The eystematicity of a function is a nebulous, gray concept. ‘My intuition is that the more a function’s value depends on a set of arguments and not on ‘external variables or constants, the more it is a systematic function of these arguments. Ae ‘wel, the simpler the function, the more systematic it is ‘There ae several alternatives to compositional semantics (Hirst, 1987, p. 27). The first alternative is that the meaning of a phrase depends on the situation in which the sentence it occurs in is uttered. ‘Therefor, it is not just a systematic function of ite constituents? ‘meaning. ‘To the extent that the meaning of English sentences depends upon pragmmat-6 CHAPTER 1. INTRODUCTION jes (situation and context), any truly accurate semantics for English will not be purely compositional. ‘A second alternative is that the relationship between the meaning of a phrase and ite ‘parts is just not aystematic, For instance, in Woods's (1067) procedural semantics, whole ‘clauses are matched against a store of word patterns, each associated with a procedute call. ‘The meaning of a sentence is identified with the procedure call coreepondiag tothe pattern it matches. In thie somantie, a word or phrase may have meanings which vary arbitrarily’ from one pattern to another, and hence from sentence to sentence. Thus, per Hirst (1987, 27), “The meaning of an individual word varies iioeyncratically with the other words in ‘the same sentence.” 1.2.1 Advantages of compositionality ‘The motivation behind Frege's articulation ofthe idea of compositional semantics was that. ‘natura language itself is for the most past compositional. ‘+ People compose phrases and words whose meaning they understand in order to predict ‘the meaning of sentences they've never seen before. They do not have to memorize sdloeyneratic senses for every sentence they can comprehend. * Natural language alo appears to have the characteristic that each well-formed con- stituent has semantic significance. People understand the “meaning” of phrases ruch 1 with the new shampoo and the dog independently of a sentental context. Further- ‘nore, words and phrases retain, for the mott part, their meaning from sentence to sentence. They do not vary idioeyncraticaly depending on the context. + People don't construct a complete syntactic parse of a sentence before beginning to ‘ecipher ite meaning, (Certain computational benefits accrue to semantic interpretation with the use ofa com- ‘positional semantics. «Ifthe semantics is formulated correctly, the semantic representation attributed to a. constituent can guide the parsing and disambiguation of what fllowait ina sentence For instance, syntactically possible prepositional phrase attackmente.can be ruled out on semantic grounds if the preceding noun phrase has been interpreted. ‘Rather than storing and looking up idiceyneratic meanings for each sentence, the semantic interpreter builds sentence meanings recursively from a relatively emall store of word and idiom meanings. ‘The interpreter can thus achieve a fa: wider coverage ofthe language from a given sizeof the dictionary of meanings. + By virtue of being systematic, the algorithm (fonction) to construct the meaning of constituents is simple and uniform, having fewer special cases, in a compositional semantics,1.2. COMPOSITIONAL SEMANTICS 7 1.2.2 Montague: PTQ ‘Montague semantics may be the canonical compositional semantics. It was introduced by ‘Richard Montague in his “PTQ" (“Proper Treatment of Quantification”) paper (Montague, 1973}in the form ofa eyntax and semantics fora fragment of English. It has been extended widely in many different and probably incompatible ways by other researchers, including Mats Rooth, whote extensions of PTQ to treat even and only wil be outlined in section 25. ‘The foundation of Montague’s semantics isa formal mathematical model ofthe world. ‘Tho semantics are model-theoretic: the meaning of a sentence ita function from possi- ble worlds, or indexes, to trath values. Equivalently, the mesning is the set of possible worlés in which the sentence is true. Semantic objects in Montague semantics are model- theoretic objects such as individuals in the world model truth value, individual concepts (or intensions, which are functions from indexes to individuale). Other types of object such 1s properties of individual concepts (which are sets of individual cencepts) are recursively defined on top of these base types. In Montague semantics, exprtsions of intensional logic (IL) are used as an intermediate representation between Englith and the semantic ‘objects. English expressions are translated into I, expressions, which ean be interpreted to ‘determine the model-theoretic objecte that they denote, ‘The PTQ grammar consists of pais of rales. The frst ia gyntatic rule and the second is a translation (semantic) rule that specifies an IL trandlation for the English phrase constructed by the fret. Hirst calls this property tandem operation of the semantic and syntactic rules. The semantic rales make use of Montague's theory of types of semantic ‘objects, Each type of syntactic constituent corresponds to one of the types of IL. Each semantic rule combines objects of an expected type. ‘A common kind of translation rule isthe fanetional application rule, co named be- ‘cauteit forms the semantic object of a phrase by applying the translation of one ofits parts ‘asa fanction of that of another. A typical rule ofthis kind has thie form: (28) If-X is of type a and ¥ is of the type of functions from intentions ofa to f, then Zin ¥(*X) and has type 8. (Cis the intensional operator), The semantic and syntactic rules and types are paired such ‘that there is never a mismatch in the type of semantic objects the semantic rule expect. For example, the translation rale T4 (Dowty, Wall and Peters, 1981, p. 192), shown below {in (14), dictates that Joln/(‘watks’) be the translation for the phrase John walk, which ‘the cerexponding syntactic ale $4 (also shown in (14)) gives asthe value of Fy(a,6). Jofn’ and alts! aze the translations of John and walks, respectively. (14) St: Hae Pr and 6 € Pry, then Fy(a,6) € P,, where F(a, 6) = af’, and 6 is ‘the result of replacing the first verb in 6 by it third person singular present form,8 CHAPTER 1. INTRODUCTION ‘Té: Wa € Prand 6 € Pry,and a, 6 translate into a, 6 respectively, then F(a, 6) translates into o("!). ‘A second kind of rule introduces words ayncategorematically. These words are not as- ‘signed a semantic (or syntactic) type. They are specified explicitly by name rather than by type in the syntactic and sumantic rales. The method by which syncategorematic fn- troduction rules construct a semantic object is usually more complicated than functional application. One of the simpler rule pair (Si1a, Ta) introduces the word and (Dowty, Wall and Peters, 1981, p. 198): (48) Sila: If 4,9 € A, then Fa(d,¥) € P, where Fu(¢,¥) = ¢ and y. ‘Tita: If 4, YEP and 9, translate into, y respectively, then Fa, p) translates into [7 AW. AA third and final kind of semantic rule translates most of the words in the grammar. ‘Montague semantics is nothing if not compositional. Strong typing allows the single ‘device of functional application tobe used to construct the semantic objects of mont phrases, ‘Syncategorematic rules make the mapping to semantic objects of phrases from those of their ‘arts just a itl les systematic. Montague semantic isnot frequently implemented in AI systems because of two draw- backs, ‘The first is that itis computationally impractical. Ite semantic objects are‘nfnite objects and hugo sets of things such as functions, possible worlds, and functions among. ‘thove things. The second is that the meaning it attributes to a sentence isthe aet of post ble worlds in which the sentence is trae. In AI, “we are interested not oo much in whether te of affairs is or could be true in aome possible world, but rather in the state of af: {are itec™ (Hizst, 1987, p. 32). An addtional drawback of truth-conditional semantic in general is that they are hard to extend to sentences that are not declarative. 1.2.3 Hirst: Absity [Absity is a semantic interpreter, » program that maps text in a natural language to a representation of ite meaning. Absity ess a highly compositional semantics for English that th author says is “Montagueingpired” (Hirst, 1967.38). Like PT, it makes use of strong typing of semantic objects and of the operation of eyntax and semantics in tandem (Birt, 1987, p. 44). ‘Absty's underlying representation ofthe wold i a knowledgebase consisting of frames, rather than a model. A frame isa collection of stereotypical knowledge about tome top or concept (Hirst, 1987, p. 12). A frame is rually stored as « named structure having sociated with Ita act of slots or roles that may be assigned values or flrs. Abaity's semantic objects belong tothe types in a frame representation language called Frail (Char- ak, 1981; Wong, 198ta, 19810), rather than the type ofa model theory. Abeity wes the following types of semantic object:12. COMPOSITIONAL SEMANTICS 9 Syntactic type Semantic type Sentence Frame statement Proper noun Frame statement Pronoun Frame etatement Common noun Frame Determiner ‘Frame determiner Noun phrase Frame statement Preposition Slot name Propositional phrase | Slot-fller pair Subordinate clause | Slot-fller pair Verb Frame ‘Adverd Slot-fller pair Ansilary Slot-fller pair ‘Verb phrase Frame descriptor ‘Clause end punctuation | Frame determiner Figure 1.1: Absity type correspondences «frame name + aot name frame determiner slot-Aler pair frame description (fame with zero of more slot-ller pre) * either an instance or frame statement (atom or frame determiner with frame descrip- ton) ‘A frame determiner is a function that retrieves frames or adds them to the knovledge base. ‘A frame description describes a frame in the knowledge base. The filler of a sot is either ‘an atom, or it isan instance, specified by a frame statement, of a frame in the knowledge base. The relationship between syntactic categories and semantic types is shown in Figure 11 (Birst, 1987, p. 48). ‘The semantics of Absity are quite compositional. There is a very amall ect of methods ‘of composing objects. The method used depends entirely on the type of the objects © A alot and a frame statement produce a slot-filler pair with the frame statement a the filler. A dlot-filler pair and a frame name produce a frame description in whick the slot of ‘the slot-fller pair belongs to the named frame. ‘+ Adlot-filer pair and a frame description produce another frame description with the slot-fller pair farther modifying the frame description.10 CHAPTER 1. INTRODUCTION (2 20 (wath 20 (agent=(the tx (person 2 (roperame=Ros))) y (deg 7y))) (27s (thane 22 (ege=0))) Figure 1.2: An Absity frame statement ‘© A frame determiner and a frame description produce a frame statement in which the escription is the argument ofthe frame determiner. Other combinations of semantic objects are not allowed. Tau, the computation of the semantic object of a phrase from those of its parts is very syriematic, Absity’s semantic strong typing and compositional systematicity have significant con- soquencee for parsing. The major one relates to the handling of subject and object noun phrases. Note that a verb and a noun phrase have types that cannot be combined, since & ‘verb is 2 frame and a noun phrase i a frame statement. Absty overcomes this problem by ‘treating object and subject noun phrases a prepositional phrsaee, which have a type (slot~ filer pair) than can be combined with a verb's type (frame aame). This is accomplished by inserting cave flags called pseudo-prepsitions into the parse tre adjacent to theee noun phrases. These case fags represont the important semantic ole of syntactic postion. They fare processed by the semantics just as if they were prepositions: they belong to the type: slot, ‘The notation Hirst uses for frames is lustrated in Figure 12, whichis a frame statement translation of the sentence (28) Ross washed the dog with a new shampoo. 1.3. ‘The semantics of focusing subjuncts Semantic non-compositionality It is a non-trivial task to devise « compositions ee- ‘antics of focusing sabjuncts for a number of reasons. First, consider the sentence (27) Roos even washed (the 000). Semanticaly, it soums most natural that even take ite focus, the dog, as ite functional argument. However, the argument available to even is ite syntactic sister (tho verb phrase Ivp washed the 00a] inthe case of (17)). To make the focus of even available as an argument ‘roqures that even or its focus be moved during semantic interpretation. Adherence to the compositionality principle is complicated by the non-adjacency of the focusing subjunct to its focus, The semantics of focusing subjuncts that thie thea proposes overcomes thie ‘problem by allowing a focusing subjunct to takeits syntactic sater as ite syntactic argument.1.3. THE SEMANTICS OF FOCUSING SUBJUNCTS u Secondly, even assuming the availability of the focus as an argument, it is not clear what a focusing subjunct actually meant. Consider sentence 18, which means, roughly “Ross washed the dog and he also washed something elee and he was not expected to wash the dog.” (18). Rose washed even (the Doa). ‘The sentence without even might be assigned the semantic representation (218) (wath tense=past agent=Ross patient=dog) 1 would be inadequate to represent sentence (18) a8 just (20) (wash tense=past agent=Ross_patient=(even argument=dog)) since this does not capture the idea that Ross cleaned something else. More appropriate ‘seems the representation (21) (and (wash tense=past agent=Ross patient=dog) (wash tense=past agent=Ross patient=2X) d ‘which isto be read, roughly, as “Ross washed the dog and he also washed something elee.” ‘An approach of this type will be adopted in this thes. ‘Thindly, the semantic function that is assigned to a focusing subjunct must be capable of accepting arguments of different typee, since both the focus of a focusing eubjunct and ita ayntactic argument ({ts sister node) vary in type. But a usual feature of compositional semantics is that logical and linguistic arguments of a phrase or lexical item must be of specified type. Pragmatic considerations According to Fraser (1971), the sentence (22) would have the three-part interpretation given in (28). (22) Even (Ross) washed the dog. (23) 1, Rows washed the dog. 2, Other people washed the dog. 8. The speaker would not expect Row to wash the dog, However, these three parts of the meaning of (22) are not equal in the following respect. For a listener to acept (22) aa (truth-conditionally) trae, only (23.1) need hold. If (28.1) holds but (28.2) or (29.3) do not, the listener is likely to accept the sentence as true, but with a reservation such ax “Well, yes, Ross washed the dog, but no one ee did” or “Well, yet, he did wash the dog, but that is jast as one should expect” (Karttunen and Peters,2 CHAPTER 1. INTRODUCTION 1970, p. 12). In such cases the sentence is often taken to be strictly speaking true, though pethape inappropriate in most circumstances, on contextual grounds. Therefore, it can be argued that even does not alfect the truth conditions of the ventence. That is, (22) is trae exactly when (23.1) it. For this reason, the sentence ie sometimes aid to asvert (23.1) and ‘to presuppose (23.2) and (23.3) (ove Section 1.5). ‘Ropreoenting the content of (23.8) requires that the speaker's beliefs be represented. Distinguishing between the contributions of (28.1) and those of (23.2) and (23.3) to the ‘meaning ofthe sentence (22) requires the ability to distinguish defeaible, or “non-asserted”, cr “presuppositional” meanings from a santence's “assertion”. Those concerns fall within ‘the scope of the field called pragmatics (Levinson, 1983, pp. 5, 12) 1.4 Goals of this research ‘The goal ofthis thesis is a semantics for focusing subjancts that is © comyositional, © comyutationally practical, 4 able to different between asserted and non-asserted meaning, «sensitive to intonation, and 1 crose-categoral. ‘The semaxtics should be compositional for reasons given in Section 1.2.1. The semantics [develop is based on Rooth’s (1085) compositional semantice for even and only, described in Section 2.5 below. Being a compositional semantics, it is set within a formal theary of grammar. In order to make my semantice computationally practical, I eet it within ‘an Absity ike frame based semantic framework: the higher-order intensional logic used by Rooth is intractable. To provide evidence that the semantics is computational practical, it is incorporated into a computational semantic interpreter. ‘The semantics should differentiate between presupposed (or implicated) and asuerted meaning of focusing eubjuncts. The eemantica should be tenstive to intonation because the meaniag of foctsing eubjuncts depend #0 much on intonational stress. It should be croes-cateorial because focusing subjuncts can occur next to such a variety of syntactic arguments: the semantics must be eapable of handling focusing subjuncta in many different ingle definition or lexical entry for each focusing subjunct. ‘A secondary purpose ofthis thesis is to ilustrate the usefulness of focus (see Section 1.5) and two-part semantic representations in a semantic interpreter. A further secondary urpote is to show that it is poable to avoid the use of higher-order intensional logic, ‘which is computationally impractical and intractable, in capturing the meaning of focusing1.5. SOME DEFINITIONS 13 ssubjunctt, That is itis possible to replace the Intensionl loge with a frame-based repre- sentation of meaning in which intensions are not expressed without losing any aspect ofthe ‘meaning of focusing subjuncts themeslves, ‘This work can be seen as a development of Rooth’s and McCord’ semantics for focusing subjuncts (see Sections 2.5 and 2.6) within aframe-based semantic framework (Absity-atyle frames, see Sections 3.2 and 1.2.3) and within a formal theory of grammar (namely GPSG, 00 Seeticn 3.3). Le 1 Simplifying assumptions A completely thorough computational treatment of focusing subjuncts appears imponsile siven what linglate and AT researchers know now. Lingulsts have not provided a definitive syntactic explanation of focusing subjuncts and of thelr association with focus (see Section 23). The phenomenon of presupposition isnot well-enough understood, though it seme to involve te belief of speakers (sve Presupposition in Section 1.5, below). ‘The representation of beliefs isa problem in AI that is only partially solved. Unsolved AI problems include the ‘use of negative Information to make inferences and the maintenance of the consistency of knowledge bases, both of which enter into the semantics of focusing tubjuncts (see Section 5.4.2). The solution ofthese problems is beyond the scope of this thesis. Instead, the present work is a contribution towards the continuing effort to provide a {ull account of focusing subjuncts, It ties together some of the previous reeearch towards this effort and extends it in several directions. Like previous research in the field, this thesis makes certain simplifying assumptions with respect to focusing subjuncts, and then under- takes to develop a semantics that is adequate modulo these assumptions. Later researc ‘expected to extend the account of focusing subjancts to an even more general case in which some of the simplifying assumptions are dlscharged. 1.5 Some definitions 1t will be useful to define some of the linguistic and semantic concepts used in studying {focusing rubjuncts. The following sentence wil be used to illustrate these exampl (24) Only Muriel voted for Hubert. Presupposition The groundwork for the analysis of focusing subjuncts was laid by Horn (1968) (ste Section 2.2). He decomposes the meaning of (24) into & presupposition (26.1) and an assertion (25.2). (25) 1. Muriel voted for Hubert. 2. No one other than Muriel voted for Hubert. Horn employs Austin’s (1968) definition ofa presupposition as an implication of a sentence ‘thats prserved under negation of the sentence. For example, Austin considers the sentence“ (CHAPTER 1. INTRODUCTION (26) to be anomalous because its second clause violates the presupposition of the frst— ‘namely, that John has children—and therefore the fret clause has no truth value. (26) * All John’s children are bald, but John has no children. Hoorn also provides a definition of presupponition that he says is equivalent, due to Kata ‘and Postal (1964): “a condition that the asker of a question assumes will be accepted by anyone who tries to answer it.” ‘According to Levinson (1983, Chapter 4), presuppositions are characterized by thelr defeasiblity in certain contexts, * their links to partcalar aspects of surface structure, and ‘their behavior in complex sentences: a complex sentence doce not inherit intact all ‘the presuppositions of sentences embeded in It. ‘A more recent account of presupposition is presented by Horton (1988), who shows that constancy under negation isnot a necesary property of presuppositions. Presuppositions ‘avo also bean described ax preconditions fr the use of certalnexpresion. No single dafinition of presupposition is generally acepted, and none is embraced in ‘his paper. For our purposes, it will be suficent to note that inferences like (25.1) will ‘be represented separately from the more “obvious” truth-conditions of sentences. The efuness of making this distinction is broached in Section 8.18, I will abel as non ‘asserted meaning the propositions that Hora calls presuppositions euch as (25.1), while remaining neutral on the fasue of whether they are in fact presuppositions, The label ‘assorted meaning wil refer tothe main assertion or truth conition ofa sentence modified by a focasing subjanct Conventional implicature At least one author (Karttunen and Peters, 1979) (vee Sec- ton 24) hae deacibed (25.1) a8 a conventional implicature of (24). According to Levin- on (1983, pp. 127-128), thie le a category invented by Grice (1975) whose existence i ‘the subject of some debate. Grice used the term “implicature” to include all pragmatic Inferences discernible from an utterance, in contrast to “what is said or expreseed by the ‘truth conditions of expressions”. Grice intended “conventional implicature” to mean any ‘non-trath-conditional inference not derived from pragmatic principles such as the Gricean ‘maxims, but rather attached by convention to lexical items or expressions. ‘Therefore, they tare attached to lexical iteme rather than the truth-conditions of @ sentence, and may be calculated just from the sentence without using pragmatic principles or the context. Conven- ‘onal implicatures are not cancslable. An example ofa conventional implicature provided by Grice is that introduced by the word but, This word has the same truth-conditional ‘mesning as the conjunction and, but additionally conventionally implicates that there is come contrast between the conjuncts.1.5. SOME DEFINITIONS 18 Levinson noces that the existence of the category is challenged on the grounds that te ‘only use is to deny that truth-conditional semantics captures the meaning of words and ‘expresions. According to the challengers, among whom Levinson singles out Kempeon 1(1975), alleged cance are either truth-conditional entallments, conversational implicatures, ‘which are derived from pragmatic principles, or presuppositions. As noted above, I remain sevtral as to the status of inferences such aa (25.1). Focus For thepurposos ofthis thesis, the concept of focus as alnguiticfetureis relevant. ‘Focus is binary feature, similar to (say) gender and number, whichis either absent or present on every consiituent at surface structure. The presence of focus can be signalled ‘overtly using vatious syntactic and phonetic means, but the most common realization of ‘focus is by phonological stress in the focused constituent. ‘The precise meaning of the focus ‘feature is subject of discussion in this thesis (se Section 3.1.7). In thin thesis, the term focus by iaelf wil refer tothe focus feature. Foews of focusing subjunct isa separate concept defined below. ‘Focus and scope of focusing subjuncts In his work, Horn describes only (when mod- ‘fying an NP) axa predicate taking two arguments, “the term [o] within its sope” and some proposition [Fo] containing that term” (Hora, 169, . 99). The meaning ofthe predicate is thea to presuppose thatthe proposition Fis tre of , and to assert that zis the unique term of which Fis true. For example, in (24), taking = to be Mariel and F to ‘be the property of voting for Hubert produces the presupposition (25.1) and the assertion. (52). 1 a similar win, n Karttonen and Peter’ analysis of even (eee Section 2), ihe Simplicatare” (ay non-asered meaning of even i esi to depend on two argument, & Joous and a sme (Karten and Peters, 1979, p. 25). For example, they attribute the implicature (272) to sentence (27.1), taking Mary as the focus and Bill likes = as the scope of even (21) 1, Bul thes even Many. 2, Bil likes other people besides Mary. In general they describe the implicature contributed by even as (28) There are other 2 under consideration besides « [the focut] euch that scope]. ‘The scope is the open sentence ...2... ‘The two arguments of focusing subjuncts described by Horn and by Karttunen and Pe- ‘ters are concepts that are referred to in all of the literature on focusing subjuncts. Hance- forth, the terminology of Karttunen and Peters willbe used to describe these arguments. ‘The term focus of a focusing subjunet will refer to the item that the focusing subjanct:16 CHAPTER 1. INTRODUCTION ‘draws attention to, or focuses, This isthe same as Horn's argument 2 in his formulation of tho meaning of only. The term scope of the focusing subjunct will zefr to the open sentence in Karttunen and Peters's description of even, or, equivalently, the argument F {in Horn’s description of only. This open sentence can always be obtained by removing the focusing subjunet from a sentence, and then replacing the item serving asthe focus of the focusing eubjunet with a viable (soe Section 3.1.1).Chapter 2 Previous research 2.1 Overview of the literature ‘The scant attention that has been pald to the focusing eubjuncts in the linguistics and computational linguistics literature is devoted mainly to the adverbs only and even. In this chapter, we review five different accounts of the semantics of these words: ‘+ Horn (1969) establishes foundations that are on the whole accepted by the later ‘writers—namely, hs presupposition assertion analysis of only and his use ofthe focus and scope arguments described in Section 1.5. ‘+ Jackendoff’s (1972) theory of association with focus explains the association of even: (Gn particnlar) with its focus as a phenomenon of surface, rather than underlying, structure, It is challenged by Fraser (1971), and later by Rooth (1985). ‘© Rooth’s dissertation (1985) provides a revised account of asociation with focus, and semantic theory for the focusing subjuncts even and only. + One of the wall-accepted claims Hor makes about even and only is that they introduce non-asserted meaning into a sentence. Karttunen and Peters (1970) propoce a multi- part model-theoretic semantics to deal with this type of meaning, particularly the conventional implicatures that they maintain are introduced by even and only McCord’s article (1986) i the only one treating focusing subjancte that the present ‘author was able to discover in the computational linguistics literature. McCord de- scribes a semantic interpreter in which a number of the oeasing eubjuncts are given semantics that he anys fllow Horn (1969) and Dergmana (1961). In addition, several researchers have contributed work that isnot reviewed inthis chap- ‘er, but is referred to in the discussion of research issues in Section 3.1. Anderson (1972), ‘and Fraser (1971) present two competing (mainly syntactic) accounts of even that bridge ‘the gap between the earlier theories and Rooth’s theory. Bergmann (1081) describes a, Ww18 CHAPTER 2. PREVIOUS RESEARCH ‘multi-part semantics for (inter alia) even and only. Gussenhoven (1983) and Lada (1963) deal with the phenomenon of focus, especially with the relationship between focus and stress ((ntonation). 2.2. Laying the groundwork orn (1969) provides what was atthe time of its publiation a ground-breaking analysis of, ‘the meaning of only and even in terms of presupposition and assertion. Horn employs the following definition of presupposition, which he credits to Austin (1963): (20) Let $ and 5” be sentences, 1. W(S-+ 9%) and (18+ 8"), then § presupposes 5", 2, If (S— 8°) and (75 + 8), then $ entail '. (SS 1s to be read ‘From $ we can conclude 5°...) ‘That is, presuppositions are distinguished from entailments by their invariance under negae tion! ‘Horn considers the semantic contribution of only and even in a limited number of con- structions. Specifically, he looks at only modifying an NP or a VP, and even when it modifies ‘an NP. Hom describes only, when focusing an NP, as a two-place predicate (though itis not trath-valued) taking as arguments “the term withia ite scope” (the focus of only), and “some proposition containing that term” (Horn, 1969, p. 99): (80) Only (2, FY Prerappose|: Fe Alerts|: Guo # # & Fy) ‘That is, a sentence with only presupposes some proposition, or open sentence, Fs, con- structed from the wntence, and asserts that thie proposition i nt true of any Item other ‘than the focus, x (z is a parameter, not a variable). For example, the sentences (81) 1. Only Lucifer pities imal, 2, Only Lucifer pities Lucifer. 8, Lucifer pities only himedl share the presupposition "ilo doce not aves th quanton of exactly how — in constructed yet wentnce gation isnt 2 {usa problem, Parthermore, recent work ragga invasance ener nega aa delat rather thas «rle {ex prorppostions. Fors recat scooat of prevappontion that tals thee problems, ee Horton (1958) Foke aca expremion given by Hor isthe somewbas confesing Only (2= 4, Fs) om doe ct explain the perpowe or mening ofthe yma22, LAYING THE GROUNDWORK 19 (82) Lucifer pities Lucier. ‘but assert, respectively, (83) 1, No one other than Lucifer pties himeel 2. No one other than Lucifer pites Lucier. 3, Lucifer pls no one other than Lucifer. ‘with the appropriate substitutions for the focus» and scope F being, respectively, (84) 1. 2+ Lucifer and Fy ~ pites(y,») 2. 2m Lacifer and Fy + ptia(y, Lucifer) 3, 2 Lacifer and Fy + pthn(Lacter, ») Note that more recent defiaitions of presupposition supersede the one used by Hora, and it ‘is controrersal whether (52) shold be considered a prerupposition of the sentence in (31). orn recognises that intonational stress plays some role in determining the focus and ‘open sentence arguments of only, However, he does not specify exactly what ths role might be. ‘orn’s scheme is intended to handle nested occurrences of only, though he clams that bis analysis of only does not rely on the assumption that such occurrences are grammatical. He gives the following example (Horn, 1969, p. 100): (85). Only John eats only rice. ‘This sentence is sald to presuppose (88). John ests only rice. ‘hich can be in turn be decomposed into a presupposition, John ents rice, and an assertion, ‘John eats nothing but rice. The assertion of (35) is given as nly ree, nothing but rice, ‘Horn provides no explanation for the apparent non-grammaticality of sentences with two. ‘occurrences of even or also, such as this example: (88) * Even John even eats rice. Horn uses a different formulation for only when it focuses a verb phrase, as in the following sentence (80) Marie only (voted for Huet), she didn’t do the laundry. (31) Noy bt Janet { In sentences like this, only takes a property as one of its arguments. Therefore, it is a ‘second-order predicate:20 CHAPTER 2. PREVIOUS RESEARCH (40) Only (F, Fe) P: Fe A: 9(GG\(G4F & Ge) orn's formulation for VP-modifying only is awed, at least notatonally. It is not rigorous; for example, itis not clear whether 2 is intended to be fre or bound in the aerton (A) expression of (40). Another aw shows up when atempting to cast a sentence such as his example (41) Musil only (voreD) for Hubert (¢he didn't campaign for him). into the form of (40). This requires the argument substitutions of (42) to be made in order to produce the correct presuppositions and assertion, (42) 1. Fo vote (for someone) 2, 2 Mariel 3. Pe + Muriel voted for Hubert ut itis notetionlly inconsistent to have FF be the property of voting for someone, = bbe Muriel, and Fe not be the proposition that Musil campaigned for someone (ie. not necessarily Hubert). Therefore, Horns formulation stated as (40) cannot adequately handle entance suchas (4H). It can only handle sentences in which onde interpretd as having ‘he entire verb phrase as its focus. orn posits the following anal of een: (43) Bvea (2, F) P: (yy #28 Fy) A: Fe ‘That is, “even (lke also) aterts what only presupposes and presupposes the nogation of, ‘what only enserts” (Horn, 1969, p. 106). Horn credits Bruce Fraser with the observation that the data of (44) (with focus on Muriel) are explained by the fact that clefting specifies ‘uniqueness, as does only, while even and als, incompatibly, preuppose nos-uniqueness > nly (44) W's} *even } (Mariel) who voted for Hubert. also 2.3 Association with focus According to Jackendoff (1972, pp. 8, 16, the semantic representation ofa sentence coniate of four parte: "Wace the posibiliy ofthe wae I Mariel wh did X10 alio Mari! whe did (2). t's even Mori who di (2) a thee ease, Masi ot the fen of leven23, ASSOCIATION WITH FOCUS a ‘Functional structure Function-argument relationships ‘Modal structure The conditions under which the sentence purports [sc] to correspond to real-world situations. “Table of coreference The store of all coreferences between NP& ‘Focus and presupposition Distinction between new and old in‘ormation. New informa- ‘tion is that whichis assumed by the speaker not to be shered by him/her and the hearer. Jackendoffs usage ofthe terms focus and presupposition differs from the usage of the ‘terms inthis theis. His “focus” is semantic rather than eyatacti:. However, Jackendof's marker F (sce below) is the same as the linguistic focas feature defined in Setion 1.5. “Jackenof's “focus” Is just the semantic material asocated with the constituent marked F (rue (46) below). Jackandoff formally incorporates focus into his theory; he devotes ‘chapter of his book to it. JTuckendoff presents two different account of focus assignment. The cleft theory main- ‘tains that focus is ausigned at deep structure. Sentences containeg an emphatic stress are soem os beng cleft at dep structure, withthe focus being the predicate of the higher clause, ‘and the presupponiton being the lower clause (Jackendof, 1972, yp. 280-231). Several ar- ‘guests agnns thie sccount are advance, including the posit of foc that cannot be ‘lofted, such as verbs and quantifiers (Jackendof, 1972, pp. 234): (48) 1. Did Fred arr Bint 2, * Was it urr that Fred (did to) Bil? ‘Jechendoff adopts an alternative theory, which he attributes to Chomsky. In this theory, ‘focus is assumed to be assigned, at surface structure, to some phrase containing the main ‘intonational stres ofthe sentence. A semantic marker F Is assumed to be preset in surface ‘structure and associated with some node containing stress (Jackendoff, 1972, p. 242). Thie 4s the same as the focus feature defined in Section 1.5 of this thesis. Jackendoff identifies ‘focus with information status. That is, the F marker marks the portion of the sentence ‘whose semantic material is new information. Jackendoff's view on this matter is in accord with shat of Guasenhoven (1983), who identifies focus with information status, but at odds with Ladd (1983), who argues that itis possible for contextually “given” items to have focus Jackendoff postulates a rule, focus assignment, to divide a reading into focus and presupposition on the basis of syntax, including the F mazker. (48) Focus assignment rule: In a sentence S, with otherwise determined semantic representation SR, the s2- ‘mantic material associated with surface structure nodes dominated by F is the2 CHAPTER 2. PREVIOUS RESEARCH Focus ofS. Substitute an appropriate semantic variable = for Focus in SR to form ‘he function Presuppe(2)- The presupposition of § is then formed as 2.1 and the ‘sertion is 2.2 (Jackendoff, 1972, p. 247), nests Jia promt daoue dePreespa(e)} Yedetned JO ay is amenable to dicanion in aderdouson ‘Pocus € Az Presupps(2) (2.2) JackendoiT emphasises that the “appropriate semantic variable” must be determined se- ‘matically, limited by the notion of “possible contrast” with the focus. For example, the focus assignment rale derives (47.2) as the function Preeupps() for ‘the sentence (47.1), This yields the presupposition (47.8) and assertion (47-4) (Jackendof 1972, pp. 245-247). (41) 1, John uiees Bil. 2, the attitude of John toward Bill is 2 3, Az [the attitude of John toward Bill is 2] ie under discussion 4. lke € da [the attitude of John toward Billi 2] ‘Jeckeadoff proposes a rule of association with focus that applies to focusing sub- juncts. The rule states that even and “related worde” are associated with focus by having the focus in their range (see below) Focus is astigned at surface structure, notin deep structure, an the asorintion of even with ts focur occurs at surface structure. In the ease of even, the juntifcation for association with focus is that even aasociated with a constituent ‘plies urusualness or unexpectedness, and “if there is something unexpected abou: the constituent it must be new to the hearer, and hence by definition part of the focus.” Jack ‘endoff does not provide a similar explanation forthe association of other focusing subjuncts (euch as just and only) with focus. In Jackendof’s scheme, diferences between the ranges for the vatious focusing adverbs ‘account for their different distributions (Jackendofy, 1972, pp. 249-260). Range of even If evenis directly dominated by a node X, then X and all nodes dominated by X are in the range of even. Range of only If only is directly dominated by a node X, then X and all nodee dominated by X and to the right of only re in the range of only "Baas JaclendoiPy “acu is just the semantic material aocated withthe constituent macked ?, ‘when be tla about having the foci range, we may anette for “Tous” the sense defined in Section Ts of hie thei28. ASSOCIATION WITH FOCUS 2 ‘That is, only (and just, which has the same range) cannot precede ite focus, while even (48) 1, * (Joan) only gave his deughter anew bicycle. 2, (Joun) even gave his daughter a new bicycle, ‘Jackendoff provides no independent linguistic motivation for the difference in ranges. Fur- ‘thermore, there is some evidence that his formulation does not captize the distribution of ‘only. For example inthe sentence (49) I saw his (EupEx) brother only. ‘the focus of only is to its left. However, the idea of “range” does capture some important differences in the distributions of focusing subjoncts. For example, alone is notable for having a range exclusively to its left (Quirk et al., 1985, §8.118): (60) * Alone (ten workers) reported sick yesterday. ‘According to Jackendoff, the rule of association with focus correctly predicts the restric- tion on occurrences of even to one per sentence, assuming that “thers is only one focus per ‘seatence”. However, later in his chapter on focus, Jackendoff accepts that sentences such ss (61) have more than one focus (Jackendof, 1972, p. 260): (51) A: Well, what about Faro? What did az eat? B: Frup ate the BEANS. 1t appears that both Fred and beans are focused. Jackendoff accounts for multiple foci and sultiple emphatic sree by using two kinds of pitch accents, A and B. An A pitch accent is realled phonologically by high stress on some focus syllable, followed by a phrase that concludes with a fll in pitch, B pitch aceat is similar but concludes with a rise in pitch (GJackendoff, 1972, p. 258). A and B pitch accents are present as features on the focus marker F. B accent occurs on the “variable” whose value is chosen ist, for example the one & questioner is asking about. accent occurs on the vasiable choven 90 as to make ‘the sentence true forthe value of the other variable. That is, “the B accent defines an independent variable and the A accent « dependent variable” (Jackendof, 1972, p. 262). In ‘the example above, Fred receives B accent and beans receives A accent. Jackendof points cnt that the diference between independent and dependent variable is similar to “the ‘traditional notions of topic and comment” (Sackendoff, 1972, p. 262). Because Jackendof's analysis of multiple foci requires one focus to depend on the other, It docea't cover sentences in which there is no bass for selecting oxe ofthe foc} aa being ‘dependent on the other, such as this standard though somewhat unfortunate example: (52) John can't even sell WHISKEY to the INDIANS.4 CHAPTER 2. PREVIOUS RESEARCH 2.4 Multi-part semantics ‘Karttunen and Peters (1979) propose that much of the pretuppositional behavior associated ‘ith particles like too, also, enen, and only is really due to ecnventional implicatures (@efined in Section 1.5, and below), Karttunen and Peters claim that even in a sentence such as (63). Even Bill ikee Mary. Plays no role in determining ite truth conditions citing Gtalnaber (1974) for support. This claim ia azo consistent with Horn's (1969) assertion-preeupposition ansiysis of even. The speaker's “principal commitment” when uttering (53) i said tobe the content of the propo- sition (64) Bil kes Mary. ‘hich they call the sentence’s expressed meaning or denofation (my atserted meaning). The speaker also commits himself to (85) 1. Other people besides Bill like Mary. 2. Of the people under consideration, Bill isthe lear likely to like Mary. Sentence (63) isin some way wrong if (54) is false. However, Karttunen and Peters argue ‘thatthe speaker is usually credited with saying “something that is partially correct” if (54) fs true but (65.1) or (85.2) is fale. As noted in Section 1.3, x response to (63) in such circumstances might be qualified agreement, such at "Well ye, he does like her; but that just as one should expect” (Karttunen and Petere, 1979, p. 12) Karttunen and Peters take the preceding to be evidence thatthe propositions contained ‘in (85) are conventional implicatures of ventence (88), distinct from ite denotation, They are implicated, not asserted, because “the truth of what (53) actually soys depends solely ‘on whether Bill kee Mary.” ‘The implicatures are called conventional because they are ‘always present in even sentences hy virtue of common practice. That is, they do not arise ‘rom general conversational principles and the context of the utterance, rather “they simply ‘azise from the presence of the word even", and they cannot be canceled, eay by modifying (53) asin (60) # Even Bill likes Mary but no one elee does, Which is seltcontradictory (Karttunen and Poters, 1979, p. 12). Karttunen and Peters ‘comment that “conventional implicatures are not set apart so they can be challenged in a direct way"; therefore Grice's co-operative principle (Grice, 1975) entails that “a sentence ‘ought to be uttered only if t does not conventionally implicate anything that is subject to controversy” (Karttunen and Peters, 1979, p. 14). But this comnent could also be applied to all indirectly conveyed aspects of meaning in general.24. MULTLPART SEMANTICS 2% Karttunen and Peters propose to represent sentence meaning with a two-dimensional Jogie* in which “truth-conditional” and conventionally implicated aspects of meaning are separated. ‘This is necessary, they say, because expressed and implicated meanings of a phrase are inherited diferently by complex sentences in which the phrase is embedded, (67) Ijast noticed that even Bill likes Mary. ‘This sentence says that the speaker hat notiend that Rill ikes Mary, but not that he has noticed that other people like Mary or that Bll is the least likely person to do eo. That in constructing the meaning of (67), notice applies only to the proposition “expressed bby® the embedded sentence, not to ite conventional implicatures, On the other hand, the conventional implicatures of the embedded sentence are Inherited intact by sentence (57) (Karttunen and Peters, 1979, p. 18). Karttunen and Peters describe an extension of Montague’s PTQ (1973) grammar, in which every (English) phrase is associated with two expressions of intensional logie (IL). ‘The first is the extension expression, which ie dentical tothe (IL) expression produced by the PTQ grammar. It stands for “what logicians would call the denotation of the phrase.” ‘The second js the implicature expression, signifying “what the phrase conventionally ‘implicates (itt isa sentence), what it contributes (if it is smaller than a sentence) to the conventional implicatures of sentences having it as a part™ (Karttunen and Peters, 1979, p. 16). Bach lexical entry in their grammar has its associated extension and implicature ex- pressions. Asa phrase a is constructed using a syntactic rue, a paired translation rule (emantic re) assigns an extension expression (as in PT'Q), a*, and an implicsture ex- pression a' to the derived phrase. These expressions are computed as a function of the ‘extension ani implicature expressions of the derived phrase's constituents, Usually, the ‘extension expression is purely a function ofthe extension expressions of the constituents. ‘An example of pair of syntactic and translation rule ie (58). It corresponds to syntactic rule 5 in PTQ (Karttunen and Peters, 1979, p18). (68) 1. arte a transitive verb and f isa noun phrase ‘then af is a verb phrase, where J is the accusative form of 6. 2. Translation: (0A(*B*); Asfai(e,"B) 0 aM(e,"9) ‘ere, is the intensional operator of TL. ais the zeslt of applying a heritage function, 4, to the extension and implicature expressions of a. It is a transformation that must be applied to the inherited conventional implicatures of phrases that a takes as arguments. "ine not what Kettanen and eles clit. Bot they acknowledge tha thee eetm a very similar In concept to H Hersbergers (1973) *wo-dimeniona lope”6 CHAPTER 2. PREVIOUS RESEARCH ‘According to Karttnnen and Peters, the use of the heritage function is necowary to sceount forthe implicatures of fatensional verbe auch ae hope, believe, and tell. Consider for example the sentence (50) It wasn Bill who tapped Mary's phone. ‘which presupposes (60) Mary's phone was tapped. Now in sentence (61), (58) is embedded a a complement ofthe verb hope. (61) John vp hoped that it wasn't Bil who tapped Marys phone). Ordinarily, a sentence or phrase inherits all of the conventional implicatures of ite con- stituents, But in this cas, according to Karttunen and Peter, (61) does not inherit (59)s premupponition. Rather, (61) conventionally implicates thet John Believed Mary's phone ‘was tapped. Karttunen and Peter's solution ie thatthe verb phrase in (61) acquires the conventional implicatues of the embedded sentence (50) after these are modified by the ffects of the transformation hope. That is, hoped transforms a proposition such as (60) Into an acription of belief in this proposition (Karttunen and Petes, 1679, pp. 20-21, Karttunen and Peters Iustrate the ability of thelr grammar to handle problem cases such as even sentences. They imit their analysis to cases in which even procedes an exten- sionally interpreted noun phrase (Interpreted de re) on which it focuses, The denotation and ezitentalimplzature (see below) that they attribute to even are similar Horns (1960) ‘usertion nd presupporiton, respectively. According tothe author, ther analysis of even 1s easly generalized, for example, to account for even with an adjective focxs, or to explain ‘the semantics of to. ‘The implicature expreision for even has two pats. Firt, the existential implicature (ay presuppositional assertion) contsbuted by even (Kasttanen and Peters, 197, p. 25) is ven a (62) There are other = under consideration besides a such that Ter, « stands for the focus of even. The open sentence... 2... bound by the existential ‘quantifier there ore in (62) i what Karttunen and Peters call the scope of even® Note ‘thatthe focus and scope arguments of even correspond closely to the arguments Hora uses to describe the meaning of even For example, in (63), Even (Bil) kes Mary. the focus of evens Billand the scope is = Uke Mary. Karttunen and Peter illustrate that focus and scope are both relevant and independent components of the meaning of an even sentence by producing the sentence VTiclrdefions of focus and scope ae the one adopted in thi thas,25. ROOTH or (G4) tie hard for me to believe that Bill can understand even (Syntectic Structures). Which is ambiguous between two readings. In both readings, the book title Syntactic Struc- tures is the focus of even. On the fret reading, the sentence implicates that Bill's under- standing i limited, and even has the scope (85.1). On the second reading, the implicature is ‘that Bill can understand other books besides Syntactic Structures, and the scope is (65.2). (68) 1. It ishard for me to belive that Bill can understand =, 2. Bill can understand 2. ‘A second, scalar implicature of even is a8 follows: (@0) For all z under consideration besides a, the likelihood that ... z ... is greater ‘than the Hkelihood that Here, ... @... iste sentence obtained from the scope ... = .. by subtitling the focus, 4 fora, Karttunen and Peter define a special even rule that generates even in ther eyntax syncategorematicaly, meaning that the word iielf ie not assigned to any syntactic eate- sory. This is how PTQ treats particles such at every, the, and not. They define a meaaing postulate for even (actually, an IL definition of "even that introduces (62) and (66) as conventional implicatures of even Since the even rule combine a focus noun phrase (Kart. tunen and Peters se the PTQ notation “T-phraso) and a spe sentence (“t-phrase”) to produce a new sentence (tphrase) with even, a phrase like even Mary in Bil likes even ‘Mary “snot considered to be a constituent phrase” (Karttunen and Peters, 1979, p. 28), Note that even isnot treated in a manner simila to the way in which adverbs are treated in PIQ. It were, it would have to be assigned to some ayatactc category, say the category ‘of functions from noun phrases to noun phrases. ‘The semantic efect ofthe even rule is to sabetitute the focus of even forthe pronoun (variable in its expe sentence, producing the extension expression for th even sentence. ‘The raleconjoins the existential and scalar Implicatures described above vith the conven ‘ional implicatures contributed ina natural way by the focus and the scope to produce the Impliatureexpreson forthe new sentence. 2.5 Rooth In his dissertation, Rooth (1985) examines the interaction between focus, which is equated ‘with intonational prominence, and the adverbs even and only. ‘This is done within the context of a general examination of the semantic effect of focus. Rooth calls this effect ‘sociation with focus, a term introduced by Jackendoff (1972) (see section 2.2). Rooth's work is motivated by euch evidence as the minimal sentence pair (@7) 1, Tonly introduced Brut to Sue,8 CHAPTER 2. PREVIOUS RESEARCH | DS: “dap tractre 5: ruc” or Serctre PR: photolagal presentation Va LF: logical form Pa TF Figure 2.1: The EST organization of grammar 2. Lonly introduced Bill to Sus. Ifthe speaker introduces Bill and Tom to Sue, and performs no other introductions, then (67.1) is fale and (67-2) is trus, Though the two sentences difler only in the location of intonational prominence, they have diffrent trath conditions. ‘This is a case of only “antociating” with focus (which is on Bill in (67.1) and on Svein (67.2)) (Rooth, 1986, pp. 2-8). ‘Rooth bates his work upon the Extended Standard Theory (EST) organization of gram- ‘mar (Chomaly and Lasnik, 1977) shown in Figure 2.1. He adopta what he calls the “stan- dard” aavamytion about focus, that itis “a feature marked on syntactic phrases.” It follows ‘rom these ezrumptions that focus must be marked in S-structue, since it has correlated phonological/pho- the locs of semantic interpretation, and $S isthe link between the two (Rooth, 1985, p. 10). ‘That is, Rooth assumes the existence ofa focus feature (realized by intonation) on surface syntactic phrase, similar to Jackendoffs F marker. Rooth also aseumes even and only are ‘wo cases ofthe same phenomenon, #0 that that work on even predating his own applies equally well o only. Rooth refers to Selkirk (1984) to justify his linkage of focus with intonational prominence: ‘The core of Selkirk’ proposal is that the phonological reflex of focus Is pitch accent” (Roth, 1986, p. 18). Another aspect of Selkrk's proposal that Rooth accepts is that ‘complex phrase can optionally acquire focus by inheriting the feature from one ofits daughters: “Focuses can percolate (} to X-bas phrases from their heads and (i) toa phrase interpreted aoa function plus arguments from the argument phrases” (Rooth, 1985, p. 21). 2.5.1 Criticism of earlier theories Rooth examines two theories of association with focus. He calls one of them the scope ‘theory; it isdue to Anderson (1972). The otheria the standard theory predating Anderson's ‘work and credited to Fischer (1968). Both ofthese theories predate the Excended Standard "Theory, to Rooth takes come liberties in recasting them within an EST framework.25. ROOTH on ‘According to the earlier theory, which is essentially the one espoused by Fraser (1971), evenia adjacent to the focused phrase in deep structure, and is optionally moved to other (curfuce) positions. Anderson criticized this theory for being unable to accommodate mul- tiple fori, such as occur in (88): even cannot be adjacent in deep structure to more than ‘one focused phrase, (68) John only introduced (B111) to (Sve). ‘Anderton proposed (essentially) that even ‘ie generated in some single position in un- erlying structure” (Anderson, 1972, p. 898), which Rooth takes to mean DS, and is later ‘moved into its surface position (in $8). Then, an “interpretive principle”, which Rooth takes toman movement in LF, is applied. This determines which constituent, actually a focused phrase, serves as the focus of even, For instance, the surface structure (68) [s John [vp only [yp introduced [yp Bill> [rp to Sue] is interpreted by adjoining the focused phrase [yp Billlp to only, where it is in the appro- priate position to be an argument of only: (70) [3 John [ve only [yp Billa [ve introduced [x els [xp to Suel]). (the constituents subscripted “2* are co-indexed) Rooth pokes everal holes in Anderson's theory. First, it succumbs to the same erticiam as Ficher’s regarding multiple foc: only one constituent can be the argument of only in logical form. Second, the movement in LF that Anderson requires sometimes violates ‘constraints on movement that have, since Anderson published his paper, been recognized as being applicable to movement in LF as well as to syntactic movement, according to Rooth. ‘Rooth’s third objection to Anderson's theory is that there is no independent principle ‘that forces the appropriate constituent, namely the phrase with focus, to move into an argument relationship with oniy/even: ‘That foces influences the assertions and conventional implicatures of sentences lavolving only, even, and a small group of other adverbs seems to be a marginal fact about English, and we should not have to state a eeparate principle which ‘covers it (Rooth, 1985, pp. 40-41). ‘Anderson must posit a restriction that stipulates that only interact with focus, such as (71) Im LP, only must be the sister of a phrase bearing the feature F. 2.5.2 Rooth’s proposal Rooth rejects both the standard theory and the scope theory, and puts forth his own domain selection theory, which proposes that only take the VP adjacent to it in S- structure aa its argument. His proposal is not subject to the same critics as Anderson's30 CHAPTER 2, PREVIOUS RESEARCH ‘by vstue of not roqulsng the focused phrase to be an argument of ony) (Rooth, 1985, 4S). An extension of the theory allows only to take arguments other than VPs. Rooth provides a detailed PTQ-style(Montaguo, 1873) grammar that translates a focused phrase into his Intensional Logic with Focus (ILF). ‘The semantics attributed to ony is essentially that of Hor (1969) (se (30) of Section 22). However, for simplicity, Rooth drops the non-asertel messing of only. ‘The phrase Including the modifier only aserte that any property ofthe form F(a), where P is come relevant property derived from the argument VP using focus information, is exactly the property defined by the VP. Tb illustrate, the meaning of (72) John only (ve introduced Brut, to Sue. is Sif John has a property ofthe form ‘introduce y to Sue tien it is the property introduce Bil to Sue!™ (Rooth, 1985, p. 44, p.89)- Rooth employs essentially the following logic translation for only (hi preliminary ver- sion): (78) only” = APAzlVQII{z} & C(Q)] — Q=P]) Here, C is free viable representing the “domain of quantification". Specifically, C is “the characteristic function of a vet of properties, which wo think of as the set of relevant properties.” That i, only” takes a property and an object as arguments, and asserts that the argument property is the unique one that is both true ofthe object and relevant. For example, Rooth's theory would produce the translation (14) YPI[PUohn) & C(P)] —+ P= introduce (billue)] for both of the following sentenca: (78) 1, John only introduced Br to Sue. 2. John only introduced Bil to Sus. Different truth conditions for the two sentences are obtainnd by providing diferent values for C. The quantification (VP) is restricted to properties of the form ‘introduce y to Sue? in (75.1), and to relevant properties ‘introduce Bill to y" in (76.2). The desired values for are (16.1) and (76.2). (70) 1. AP3y(P = 2. AP3y/P. ‘Thote rmult in final translations (771) and (772) respectively for sentences (76.1) and (152: (77) 1. Velintroduee'ohn,2, sue) —+ ==ill 2. Valintroduce(john,bil,=) —+ ==aue]25, ROOTH Pa ‘The key feature of Rocth’s proposal, agua, is thatthe focused phrase isnot an argument ‘of only. For this reason, (75.1) and (75:2) need not be structurally distinguished in LP. a reclt of a contribution of focus to the selection of domains of quantification (Rooth, 1985, p. 45). 1 appears that Rooth’s approach involves the use of pragmatic information, The “domains of quantification” can be thought of as pragmatic background against which a sentence is ‘evaluated. Certainly, the faction C is related to the preeuppostions of a sentence: that the property (76.1) (the property of being introduced to Sue) is “relevant” ie a presupposition of (75:1). Rooth derives & modifed version of PTQ in which domains of quantification are incor- orated. He lays down scme groundwork for his derivation by showing how to construct 8 presupposition akeleton of a sentence with focus. This is just Jackendof’s (1972) fanction Presupps(2) described in Section 2.3. It i the renult of substituting a vaziable for the focused phrase(s) in asentence. Rooth's version of the presuppositional skeleton is the Pract. ‘The pset of a sentence « with normal denotation a! is defined as follows (Roth, 1985, p. 14): (78) poset ofa 1. the set of objects in the model matching a! in type, if« bears the feature F; 2, the unit set, fais anon focused non-complex phrase; 3 the sot of objects that can be obtained by picking one element from each of ‘the p-ssts corresponding to the component phrases of a, and applying the semantic rue for ato this sequence of elements, if ie a non-focused complex phrase According to Rooth, a critical diference betwoen his definition and Jackendofts is that theater's “involves a variablein the position ofa focused phrase”, whereas Rooth's requires “that a semantic object with variables in the postions of focused phrases [be] available” (Rooth, 1985, p. 16). Rooth notes that a proposal of Chomky's (1076) i like Jackendo’s ‘thls expect (the variablo in Chomsky’s propotal being a trace). Rooth's psets are actually used to construct the domains of quantification he requires ‘hia translation rule for only (Rooth, 1985, p. 59): (79) Translation rule for oly: [vp only VP) has the ILP translation A(C, VP, A2¥PIIP(2) & Pec] — P="VvP)) Vike mode ered tin jefinton i formal model Ale Montage; Fe the face fete2 CHAPTER 2. PREVIOUS RESEARCH ‘The restriction operator 2 produces an expression similar to its third argument, but in this caso with the variable C replaced with (essentially) the characteristic function of the pet ofthe intension of VP. Rooth explains his preference for this complex formala, in which a variable (C) always becomes bound: retained C because I find the possibilty that association with focus has some- thing to do with a general phenomenon of selecting domains of quantification Interesting (Rooth, 1985, p. 59). Continuing with our example, the poets for the verb phrases in (75.1) and (16:2) are ‘sventially the sets of propositions that satisfy (76.1) and (76.2), respectively. The actual ‘model-theoretic objects Rooth constructs are more complicated than this. The use of these values for C in the translation rule (79) produces the following translations forthe two verb phrases: (80) 1. 2PI[P(2) & By(P = "introduee(y, sue)]] —+ P ="introduce'(bill sue)} 2. AaVPI[P(z) & Iy[P = “introducebil, yl] —+ P= “introduce (bill, sue). Rooth’s theory accommodates multiple foci by generating an appropriate domain of ‘quantification for phrases with multiple foci (Reoth, 198, p. 61). For example, the pset of the intension of the verb phrase (81) [vp introduced Billy to Suey] is the set of properties of the form ‘introduce 2 to y'. This revults in the correct meaning. ‘being assigned to (61): if John has a property ofthe form ‘introduce = to y’, then it i the property ‘introduce Bll to Sue.” Rooth worries that his grammar doesn’: handle correctly all cases of “embedded focus”, ‘that is, a focused phrase dominated by a node also having focus. For example, Rooth produces the same translation for both ofthe following sentences: (62) 1. Jane (sent a 200K to MAR). 2, Jane (sent a/the book to MARY). ‘The VP may optionally inherit focus from its NPp and PP constituents in (821) and from its PPp constituent in (822). In either of theo cae, the meaning ofthe sentence make it an appropriate outoftheblue answer toa question about Janes activities. In either case, the pret ofthe VP is asigned the same fixed value: Rooth’s definition of p-sets “does not pay attention tothe p-sets of constituent phrases” offocued phrases (Rooth, 1965, p. 23). Becante ofthis, the VPs in (82.1) and (822) receive the same translation; hence so do the two sentences, In fact, Rooth cts Sik (1984) in explaining that (82.2) fers fom (82.1) in that it in appropriate only in context in which o/the look “old information” (fr instance, Jane's workinvolves books). Rooth appears to suggest that this shortcoming of his system's rellcts26. NLU SYSTEMS THAT HANDLE FOCUSING SUBJUNCTS 33 8 difference between the meaning of embedded and non-embedded foces. When focus is ‘non-embedied, its meaaing is explained in terms of the question the sentence answers and in terms of information fom the non-focased part ofthe sentence, When focus is embedded, ite meaning is explained in terms of novelty or givenness of discourse referents ofthe focused phrase: Bmbedded NP focus is linked to the novelty of the referent of the NP in the discourse ... ‘The interpretation for an embedded focus has nothing to do with the role of the focated NP in the sentence as a whole; the opposite is true of ‘non-embedded fod ..."The role of the non-embedded focus in| John introduced BILL to Suc] isto suggest that alternatives of the form ‘John introduced y to Sue! are under consideration. The semantic object which implements ths idea [e peet] incorporates semantic information from the non-focused part of the sentence (Rooth, 1985, p. 24). Rooth acknowledges that his proposal does not attempt to deal with this aspect of focus involving new /ld discourse referent Rooth summarizes the advantages of his theory over the scope theory as the following (Rooth, 1985, p. 81 «No syntactic bound variable is postulated; thus we do not expect the relation between ‘only/even and the focused phrase to be restricted by scope islands or local conditions fon variables, * Associaton with multiple fot is accommodated without special stipulation. ‘* Association with focus is derived as a “theorem” from independently motivated prin- ples, A notable disadvantage of Rooth’s theory is that, the way he exposes it, it accommodates, the translation of only caly in the construction [vp only VP]. This disadvantage is partly ‘emodied by a modification he proposes in Chapter III of hie dissertation that introduces a ‘rost-categorial family of operators to handle only in several other constructions. Rooth’'s

You might also like

• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Rating: 4 out of 5 stars

  4/5 (5810)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brené Brown

  Rating: 4 out of 5 stars

  4/5 (1092)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Rating: 4.5 out of 5 stars

  4.5/5 (844)
• • Magazines
  • Podcasts
  • Sheet music
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  Rating: 4 out of 5 stars

  4/5 (608)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Rating: 4.5 out of 5 stars

  4.5/5 (1716)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Rating: 4 out of 5 stars

  4/5 (590)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Rating: 4 out of 5 stars

  4/5 (1104)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Rating: 4 out of 5 stars

  4/5 (897)
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Rating: 4.5 out of 5 stars

  4.5/5 (540)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Rating: 4.5 out of 5 stars

  4.5/5 (2104)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Rating: 4.5 out of 5 stars

  4.5/5 (348)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  Rating: 4 out of 5 stars

  4/5 (1018)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Rating: 4.5 out of 5 stars

  4.5/5 (474)
• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Rating: 4 out of 5 stars

  4/5 (822)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  Rating: 4 out of 5 stars

  4/5 (1850)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Rating: 4.5 out of 5 stars

  4.5/5 (271)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Rating: 4.5 out of 5 stars

  4.5/5 (122)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Rating: 4.5 out of 5 stars

  4.5/5 (441)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Tóibín

  Rating: 3.5 out of 5 stars

  3.5/5 (1947)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Rating: 3.5 out of 5 stars

  3.5/5 (401)
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Rating: 4.5 out of 5 stars

  4.5/5 (4610)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Rating: 3.5 out of 5 stars

  3.5/5 (2259)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Rating: 4 out of 5 stars

  4/5 (4203)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Rating: 4.5 out of 5 stars

  4.5/5 (1929)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  Rating: 4.5 out of 5 stars

  4.5/5 (807)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Rating: 4 out of 5 stars

  4/5 (98)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Rating: 4.5 out of 5 stars

  4.5/5 (266)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  Rating: 4 out of 5 stars

  4/5 (1898)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Rating: 3.5 out of 5 stars

  3.5/5 (231)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Rating: 4.5 out of 5 stars

  4.5/5 (234)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  Rating: 3.5 out of 5 stars

  3.5/5 (2521)
• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  Rating: 3.5 out of 5 stars

  3.5/5 (738)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Rating: 4 out of 5 stars

  4/5 (3811)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  Rating: 4.5 out of 5 stars

  4.5/5 (2409)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Rating: 4 out of 5 stars

  4/5 (74)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Rating: 4.5 out of 5 stars

  4.5/5 (789)
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Rating: 4 out of 5 stars

  4/5 (45)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Rating: 3.5 out of 5 stars

  3.5/5 (792)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  Rating: 3.5 out of 5 stars

  3.5/5 (104)
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Rating: 3.5 out of 5 stars

  3.5/5 (137)
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  Rating: 4 out of 5 stars

  4/5 (104)
• 1778 5274 1 PB

  

  Document19 pages

  1778 5274 1 PB

  cheng long liu

  No ratings yet
• 1543 4381 1 PB

  

  Document13 pages

  1543 4381 1 PB

  cheng long liu

  No ratings yet
• Copular Constructions in English: A Multi-Dimensional Merge Approach

  

  Document22 pages

  Copular Constructions in English: A Multi-Dimensional Merge Approach

  cheng long liu

  No ratings yet
• A Fact-Aligned Corpus of Numerical Expressions: Sandra Williams, Richard Power

  

  Document5 pages

  A Fact-Aligned Corpus of Numerical Expressions: Sandra Williams, Richard Power

  cheng long liu

  No ratings yet
• Modality

  

  Document35 pages

  Modality

  cheng long liu

  No ratings yet
• Equality Versus Similarity Constructions in Englis

  

  Document27 pages

  Equality Versus Similarity Constructions in Englis

  cheng long liu

  No ratings yet