Professional Documents
Culture Documents
Glove PDF
Glove PDF
1. f (0) = 0. If f is viewed as a continuous Most of the details of these models are irrelevant
function, it should vanish as x → 0 fast for our purposes, aside from the the fact that they
enough that the lim x→0 f (x) log2 x is finite. attempt to maximize the log probability as a con-
text window scans over the corpus. Training pro-
2. f (x) should be non-decreasing so that rare
ceeds in an on-line, stochastic fashion, but the im-
co-occurrences are not overweighted.
plied global objective function can be written as,
3. f (x) should be relatively small for large val- X
ues of x, so that frequent co-occurrences are J=− log Q i j . (11)
not overweighted. i ∈corpus
j ∈context(i)
where we have used the fact that the number of Finally, we observe that while the weighting factor
like terms is given by the co-occurrence matrix X. X i is preordained by the on-line training method
Recalling our notation for X i = k X ik and inherent to the skip-gram and ivLBL models, it is
P
Pi j = X i j /X i , we can rewrite J as, by no means guaranteed to be optimal. In fact,
Mikolov et al. (2013a) observe that performance
V
X V
X V
X can be increased by filtering the data so as to re-
J=− Xi Pi j log Q i j = X i H (Pi ,Q i ) , duce the effective value of the weighting factor for
i=1 j=1 i=1 frequent words. With this in mind, we introduce
(13) a more general weighting function, which we are
where H (Pi ,Q i ) is the cross entropy of the dis- free to take to depend on the context word as well.
tributions Pi and Q i , which we define in analogy The result is,
to X i . As a weighted sum of cross-entropy error, X
Jˆ = f (X i j ) wTi w̃ j − log X i j 2 ,
this objective bears some formal resemblance to (16)
the weighted least squares objective of Eqn. (8). i, j
In fact, it is possible to optimize Eqn. (13) directly
which is equivalent1 to the cost function of
as opposed to the on-line training methods used in
Eqn. (8), which we derived previously.
the skip-gram and ivLBL models. One could inter-
pret this objective as a “global skip-gram” model, 3.2 Complexity of the model
and it might be interesting to investigate further.
As can be seen from Eqn. (8) and the explicit form
On the other hand, Eqn. (13) exhibits a number of
of the weighting function f (X ), the computational
undesirable properties that ought to be addressed
complexity of the model depends on the number of
before adopting it as a model for learning word
nonzero elements in the matrix X. As this num-
vectors.
ber is always less than the total number of en-
To begin, cross entropy error is just one among
tries of the matrix, the model scales no worse than
many possible distance measures between prob-
O(|V | 2 ). At first glance this might seem like a sub-
ability distributions, and it has the unfortunate
stantial improvement over the shallow window-
property that distributions with long tails are of-
based approaches, which scale with the corpus
ten modeled poorly with too much weight given
size, |C|. However, typical vocabularies have hun-
to the unlikely events. Furthermore, for the mea-
dreds of thousands of words, so that |V | 2 can be in
sure to be bounded it requires that the model dis-
the hundreds of billions, which is actually much
tribution Q be properly normalized. This presents
larger than most corpora. For this reason it is im-
a computational bottleneck owing to the sum over
portant to determine whether a tighter bound can
the whole vocabulary in Eqn. (10), and it would be
be placed on the number of nonzero elements of
desirable to consider a different distance measure
X.
that did not require this property of Q. A natural
In order to make any concrete statements about
choice would be a least squares objective in which
the number of nonzero elements in X, it is neces-
normalization factors in Q and P are discarded,
sary to make some assumptions about the distribu-
X tion of word co-occurrences. In particular, we will
Jˆ = X i P̂i j − Q̂ i j 2
(14)
assume that the number of co-occurrences of word
i, j
i with word j, X i j , can be modeled as a power-law
function of the frequency rank of that word pair,
where P̂i j = X i j and Q̂ i j = exp(wTi w̃ j ) are the
ri j :
unnormalized distributions. At this stage another
k
problem emerges, namely that X i j often takes very Xi j = . (17)
(r i j ) α
large values, which can complicate the optimiza-
tion. An effective remedy is to minimize the 1 We could also include bias terms in Eqn. (16).
The total number of words in the corpus is pro-
Table 2: Results on the word analogy task, given
portional to the sum over all elements of the co-
as percent accuracy. Underlined scores are best
occurrence matrix X,
within groups of similarly-sized models; bold
|X| scores are best overall. HPCA vectors are publicly
X X k
|C| ∼ Xi j = = k H | X |,α , (18) available2 ; (i)vLBL results are from (Mnih et al.,
r =1
rα
ij 2013); skip-gram (SG) and CBOW results are
from (Mikolov et al., 2013a,b); we trained SG†
where we have rewritten the last sum in terms of
and CBOW† using the word2vec tool3 . See text
the generalized harmonic number Hn, m . The up-
for details and a description of the SVD models.
per limit of the sum, |X |, is the maximum fre-
quency rank, which coincides with the number of Model Dim. Size Sem. Syn. Tot.
nonzero elements in the matrix X. This number is ivLBL 100 1.5B 55.9 50.1 53.2
also equal to the maximum value of r in Eqn. (17) HPCA 100 1.6B 4.2 16.4 10.8
such that X i j ≥ 1, i.e., |X | = k 1/α . Therefore we GloVe 100 1.6B 67.5 54.3 60.3
can write Eqn. (18) as, SG 300 1B 61 61 61
CBOW 300 1.6B 16.1 52.6 36.1
|C| ∼ |X | α H | X |,α . (19) vLBL 300 1.5B 54.2 64.8 60.0
ivLBL 300 1.5B 65.2 63.0 64.0
We are interested in how |X | is related to |C| when
GloVe 300 1.6B 80.8 61.5 70.3
both numbers are large; therefore we are free to
SVD 300 6B 6.3 8.1 7.3
expand the right hand side of the equation for large
SVD-S 300 6B 36.7 46.6 42.1
|X |. For this purpose we use the expansion of gen-
SVD-L 300 6B 56.6 63.0 60.1
eralized harmonic numbers (Apostol, 1976),
CBOW† 300 6B 63.6 67.4 65.7
x 1−s SG† 300 6B 73.0 66.0 69.1
H x, s = + ζ (s) + O(x −s ) if s > 0, s , 1 , GloVe 300 6B 77.4 67.0 71.7
1−s
(20) CBOW 1000 6B 57.3 68.9 63.7
giving, SG 1000 6B 66.1 65.1 65.6
SVD-L 300 42B 38.4 58.2 49.2
|X |
|C| ∼ + ζ (α) |X | α + O(1) , (21) GloVe 300 42B 81.9 69.3 75.0
1−α
where ζ (s) is the Riemann zeta function. In the dataset for NER (Tjong Kim Sang and De Meul-
limit that X is large, only one of the two terms on der, 2003).
the right hand side of Eqn. (21) will be relevant, Word analogies. The word analogy task con-
and which term that is depends on whether α > 1, sists of questions like, “a is to b as c is to ?”
The dataset contains 19,544 such questions, di-
if α < 1,
(
O(|C|)
|X | = (22) vided into a semantic subset and a syntactic sub-
O(|C| ) if α > 1.
1/α
set. The semantic questions are typically analogies
For the corpora studied in this article, we observe about people or places, like “Athens is to Greece
that X i j is well-modeled by Eqn. (17) with α = as Berlin is to ?”. The syntactic questions are
1.25. In this case we have that |X | = O(|C| 0.8 ). typically analogies about verb tenses or forms of
Therefore we conclude that the complexity of the adjectives, for example “dance is to dancing as fly
model is much better than the worst case O(V 2 ), is to ?”. To correctly answer the question, the
and in fact it does somewhat better than the on-line model should uniquely identify the missing term,
window-based methods which scale like O(|C|). with only an exact correspondence counted as a
correct match. We answer the question “a is to b
4 Experiments as c is to ?” by finding the word d whose repre-
sentation wd is closest to wb − wa + wc according
4.1 Evaluation methods
to the cosine similarity.4
We conduct experiments on the word analogy 2 http://lebret.ch/words/
task of Mikolov et al. (2013a), a variety of word 3 http://code.google.com/p/word2vec/
similarity tasks, as described in (Luong et al., 4 Levy et al. (2014) introduce a multiplicative analogy
2013), and on the CoNLL-2003 shared benchmark evaluation, 3C OS M UL, and report an accuracy of 68.24% on
80 70 70
70 65 65
60 60 60
Accuracy [%]
Accuracy [%]
Accuracy [%]
50 55 55
40 50 50
Semantic Semantic Semantic
Syntactic Syntactic Syntactic
30 45 45
Overall Overall Overall
20 40 40
0 100 200 300 400 500 600 2 4 6 8 10 2 4 6 8 10
Vector Dimension Window Size Window Size
Figure 2: Accuracy on the analogy task as function of vector size and window size/type. All models are
trained on the 6 billion token corpus. In (a), the window size is 10. In (b) and (c), the vector size is 100.
Word similarity. While the analogy task is our has 6 billion tokens; and on 42 billion tokens of
primary focus since it tests for interesting vector web data, from Common Crawl5 . We tokenize
space substructures, we also evaluate our model on and lowercase each corpus with the Stanford to-
a variety of word similarity tasks in Table 3. These kenizer, build a vocabulary of the 400,000 most
include WordSim-353 (Finkelstein et al., 2001), frequent words6 , and then construct a matrix of co-
MC (Miller and Charles, 1991), RG (Rubenstein occurrence counts X. In constructing X, we must
and Goodenough, 1965), SCWS (Huang et al., choose how large the context window should be
2012), and RW (Luong et al., 2013). and whether to distinguish left context from right
Named entity recognition. The CoNLL-2003 context. We explore the effect of these choices be-
English benchmark dataset for NER is a collec- low. In all cases we use a decreasing weighting
tion of documents from Reuters newswire articles, function, so that word pairs that are d words apart
annotated with four entity types: person, location, contribute 1/d to the total count. This is one way
organization, and miscellaneous. We train mod- to account for the fact that very distant word pairs
els on CoNLL-03 training data on test on three are expected to contain less relevant information
datasets: 1) ConLL-03 testing data, 2) ACE Phase about the words’ relationship to one another.
2 (2001-02) and ACE-2003 data, and 3) MUC7 For all our experiments, we set x max = 100,
Formal Run test set. We adopt the BIO2 annota- α = 3/4, and train the model using AdaGrad
tion standard, as well as all the preprocessing steps (Duchi et al., 2011), stochastically sampling non-
described in (Wang and Manning, 2013). We use a zero elements from X, with initial learning rate of
comprehensive set of discrete features that comes 0.05. We run 50 iterations for vectors smaller than
with the standard distribution of the Stanford NER 300 dimensions, and 100 iterations otherwise (see
model (Finkel et al., 2005). A total of 437,905 Section 4.6 for more details about the convergence
discrete features were generated for the CoNLL- rate). Unless otherwise noted, we use a context of
2003 training dataset. In addition, 50-dimensional ten words to the left and ten words to the right.
vectors for each word of a five-word context are The model generates two sets of word vectors,
added and used as continuous features. With these W and W̃ . When X is symmetric, W and W̃ are
features as input, we trained a conditional random equivalent and differ only as a result of their ran-
field (CRF) with exactly the same setup as the dom initializations; the two sets of vectors should
CRFjoin model of (Wang and Manning, 2013). perform equivalently. On the other hand, there is
evidence that for certain types of neural networks,
4.2 Corpora and training details training multiple instances of the network and then
We trained our model on five corpora of varying combining the results can help reduce overfitting
sizes: a 2010 Wikipedia dump with 1 billion to- and noise and generally improve results (Ciresan
kens; a 2014 Wikipedia dump with 1.6 billion to- et al., 2012). With this in mind, we choose to use
kens; Gigaword 5 which has 4.3 billion tokens; the 5 To demonstrate the scalability of the model, we also
combination Gigaword5 + Wikipedia2014, which trained it on a much larger sixth corpus, containing 840 bil-
lion tokens of web data, but in this case we did not lowercase
the analogy task. This number is evaluated on a subset of the the vocabulary, so the results are not directly comparable.
dataset so it is not included in Table 2. 3C OS M UL performed 6 For the model trained on Common Crawl data, we use a
worse than cosine similarity in almost all of our experiments. larger vocabulary of about 2 million words.
the sum W + W̃ as our word vectors. Doing so typ-
Table 3: Spearman rank correlation on word simi-
ically gives a small boost in performance, with the
larity tasks. All vectors are 300-dimensional. The
biggest increase in the semantic analogy task.
CBOW∗ vectors are from the word2vec website
We compare with the published results of a va-
and differ in that they contain phrase vectors.
riety of state-of-the-art models, as well as with
our own results produced using the word2vec Model Size WS353 MC RG SCWS RW
tool and with several baselines using SVDs. With SVD 6B 35.3 35.1 42.5 38.3 25.6
word2vec, we train the skip-gram (SG† ) and SVD-S 6B 56.5 71.5 71.0 53.6 34.7
continuous bag-of-words (CBOW† ) models on the SVD-L 6B 65.7 72.7 75.1 56.5 37.0
6 billion token corpus (Wikipedia 2014 + Giga- CBOW† 6B 57.2 65.6 68.2 57.0 32.5
SG† 6B 62.8 65.2 69.7 58.1 37.2
word 5) with a vocabulary of the top 400,000 most
GloVe 6B 65.8 72.7 77.8 53.9 38.1
frequent words and a context window size of 10.
SVD-L 42B 74.0 76.4 74.1 58.3 39.9
We used 10 negative samples, which we show in
GloVe 42B 75.9 83.6 82.9 59.6 47.8
Section 4.6 to be a good choice for this corpus. CBOW∗ 100B 68.4 79.6 75.4 59.4 45.5
For the SVD baselines, we generate a truncated
matrix Xtrunc which retains the information of how L model on this larger corpus. The fact that this
frequently each word occurs with only the top basic SVD model does not scale well to large cor-
10,000 most frequent words. This step is typi- pora lends further evidence to the necessity of the
cal of many matrix-factorization-based methods as type of weighting scheme proposed in our model.
the extra columns can contribute a disproportion-
Table 3 shows results on five different word
ate number of zero entries and the methods are
similarity datasets. A similarity score is obtained
otherwise computationally expensive.
from the word vectors by first normalizing each
The singular vectors of this matrix constitute
feature across the vocabulary and then calculat-
the baseline “SVD”. We also evaluate two related
ing the cosine similarity. We compute Spearman’s
baselines:
√ “SVD-S” in which we take the SVD of
rank correlation coefficient between this score and
Xtrunc , and “SVD-L” in which we take the SVD
the human judgments. CBOW∗ denotes the vec-
of log(1+ Xtrunc ). Both methods help compress the
tors available on the word2vec website that are
otherwise large range of values in X.7
trained with word and phrase vectors on 100B
4.3 Results words of news data. GloVe outperforms it while
using a corpus less than half the size.
We present results on the word analogy task in Ta-
Table 4 shows results on the NER task with the
ble 2. The GloVe model performs significantly
CRF-based model. The L-BFGS training termi-
better than the other baselines, often with smaller
nates when no improvement has been achieved on
vector sizes and smaller corpora. Our results us-
the dev set for 25 iterations. Otherwise all config-
ing the word2vec tool are somewhat better than
urations are identical to those used by Wang and
most of the previously published results. This is
Manning (2013). The model labeled Discrete is
due to a number of factors, including our choice to
the baseline using a comprehensive set of discrete
use negative sampling (which typically works bet-
features that comes with the standard distribution
ter than the hierarchical softmax), the number of
of the Stanford NER model, but with no word vec-
negative samples, and the choice of the corpus.
tor features. In addition to the HPCA and SVD
We demonstrate that the model can easily be
models discussed previously, we also compare to
trained on a large 42 billion token corpus, with a
the models of Huang et al. (2012) (HSMN) and
substantial corresponding performance boost. We
Collobert and Weston (2008) (CW). We trained
note that increasing the corpus size does not guar-
the CBOW model using the word2vec tool8 .
antee improved results for other models, as can be
The GloVe model outperforms all other methods
seen by the decreased performance of the SVD-
on all evaluation metrics, except for the CoNLL
7 We also investigated several other weighting schemes for
test set, on which the HPCA method does slightly
transforming X; what we report here performed best. Many better. We conclude that the GloVe vectors are
weighting schemes like PPMI destroy the sparsity of X and
therefore cannot feasibly be used with large vocabularies. useful in downstream NLP tasks, as was first
With smaller vocabularies, these information-theoretic trans-
formations do indeed work well on word similarity measures, 8 We use the same parameters as above, except in this case
but they perform very poorly on the word analogy task. we found 5 negative samples to work slightly better than 10.
Semantic Syntactic Overall
Table 4: F1 score on NER task with 50d vectors. 85
Accuracy [%]
and CW. See text for details. 70
65
Model Dev Test ACE MUC7
60
Discrete 91.0 85.4 77.4 73.4
55
SVD 90.8 85.7 77.3 73.7 50
SVD-S 91.0 85.5 77.6 74.3 Gigaword5 +
Wiki2010 Wiki2014 Gigaword5 Wiki2014 Common Crawl
SVD-L 90.5 84.8 73.6 71.5 1B tokens 1.6B tokens 4.3B tokens 6B tokens 42B tokens
4.4 Model Analysis: Vector Length and 4.6 Model Analysis: Run-time
Context Size The total run-time is split between populating X
In Fig. 2, we show the results of experiments that and training the model. The former depends on
vary vector length and context window. A context many factors, including window size, vocabulary
window that extends to the left and right of a tar- size, and corpus size. Though we did not do so,
get word will be called symmetric, and one which this step could easily be parallelized across mul-
extends only to the left will be called asymmet- tiple machines (see, e.g., Lebret and Collobert
ric. In (a), we observe diminishing returns for vec- (2014) for some benchmarks). Using a single
tors larger than about 200 dimensions. In (b) and thread of a dual 2.1GHz Intel Xeon E5-2658 ma-
(c), we examine the effect of varying the window chine, populating X with a 10 word symmetric
size for symmetric and asymmetric context win- context window, a 400,000 word vocabulary, and
dows. Performance is better on the syntactic sub- a 6 billion token corpus takes about 85 minutes.
task for small and asymmetric context windows, Given X, the time it takes to train the model de-
which aligns with the intuition that syntactic infor- pends on the vector size and the number of itera-
mation is mostly drawn from the immediate con- tions. For 300-dimensional vectors with the above
text and can depend strongly on word order. Se- settings (and using all 32 cores of the above ma-
mantic information, on the other hand, is more fre- chine), a single iteration takes 14 minutes. See
quently non-local, and more of it is captured with Fig. 4 for a plot of the learning curve.
larger window sizes.
4.7 Model Analysis: Comparison with
4.5 Model Analysis: Corpus Size word2vec
In Fig. 3, we show performance on the word anal- A rigorous quantitative comparison of GloVe with
ogy task for 300-dimensional vectors trained on word2vec is complicated by the existence of
different corpora. On the syntactic subtask, there many parameters that have a strong effect on per-
is a monotonic increase in performance as the cor- formance. We control for the main sources of vari-
pus size increases. This is to be expected since ation that we identified in Sections 4.4 and 4.5 by
larger corpora typically produce better statistics. setting the vector length, context window size, cor-
Interestingly, the same trend is not true for the se- pus, and vocabulary size to the configuration men-
mantic subtask, where the models trained on the tioned in the previous subsection.
smaller Wikipedia corpora do better than those The most important remaining variable to con-
trained on the larger Gigaword corpus. This is trol for is training time. For GloVe, the rele-
likely due to the large number of city- and country- vant parameter is the number of training iterations.
based analogies in the analogy dataset and the fact For word2vec, the obvious choice would be the
that Wikipedia has fairly comprehensive articles number of training epochs. Unfortunately, the
for most such locations. Moreover, Wikipedia’s code is currently designed for only a single epoch:
Training Time (hrs) Training Time (hrs)
1 2 3 4 5 6 3 6 9 12 15 18 21 24
72 72
70 70
Accuracy [%]
Accuracy [%] 68 GloVe 68
CBOW
66 66
64 64 GloVe
Skip-Gram
62 62
5 10 15 20 25 20 40 60 80 100
60 60
Iterations (GloVe) Iterations (GloVe)
1 3 5 7 10 15 20 25 30 40 50 1 2 3 4 5 6 7 10 12 15 20
Negative Samples (CBOW) Negative Samples (Skip-Gram)
it specifies a learning schedule specific to a single methods or from prediction-based methods. Cur-
pass through the data, making a modification for rently, prediction-based models garner substantial
multiple passes a non-trivial task. Another choice support; for example, Baroni et al. (2014) argue
is to vary the number of negative samples. Adding that these models perform better across a range of
negative samples effectively increases the number tasks. In this work we argue that the two classes
of training words seen by the model, so in some of methods are not dramatically different at a fun-
ways it is analogous to extra epochs. damental level since they both probe the under-
We set any unspecified parameters to their de- lying co-occurrence statistics of the corpus, but
fault values, assuming that they are close to opti- the efficiency with which the count-based meth-
mal, though we acknowledge that this simplifica- ods capture global statistics can be advantageous.
tion should be relaxed in a more thorough analysis. We construct a model that utilizes this main ben-
In Fig. 4, we plot the overall performance on efit of count data while simultaneously capturing
the analogy task as a function of training time. the meaningful linear substructures prevalent in
The two x-axes at the bottom indicate the corre- recent log-bilinear prediction-based methods like
sponding number of training iterations for GloVe word2vec. The result, GloVe, is a new global
and negative samples for word2vec. We note log-bilinear regression model for the unsupervised
that word2vec’s performance actually decreases learning of word representations that outperforms
if the number of negative samples increases be- other models on word analogy, word similarity,
yond about 10. Presumably this is because the and named entity recognition tasks.
negative sampling method does not approximate
the target probability distribution well.9 Acknowledgments
For the same corpus, vocabulary, window size, We thank the anonymous reviewers for their valu-
and training time, GloVe consistently outperforms able comments. Stanford University gratefully
word2vec. It achieves better results faster, and acknowledges the support of the Defense Threat
also obtains the best results irrespective of speed. Reduction Agency (DTRA) under Air Force Re-
search Laboratory (AFRL) contract no. FA8650-
5 Conclusion 10-C-7020 and the Defense Advanced Research
Recently, considerable attention has been focused Projects Agency (DARPA) Deep Exploration and
on the question of whether distributional word Filtering of Text (DEFT) Program under AFRL
representations are best learned from count-based contract no. FA8750-13-2-0040. Any opinions,
findings, and conclusion or recommendations ex-
9 In contrast, noise-contrastive estimation is an approxi-
pressed in this material are those of the authors and
mation which improves with more negative samples. In Ta-
ble 1 of (Mnih et al., 2013), accuracy on the analogy task is a do not necessarily reflect the view of the DTRA,
non-decreasing function of the number of negative samples. AFRL, DEFT, or the US government.
References Word Representations via Global Context and
Multiple Word Prototypes. In ACL.
Tom M. Apostol. 1976. Introduction to Analytic
Number Theory. Introduction to Analytic Num- Rémi Lebret and Ronan Collobert. 2014. Word
ber Theory. embeddings through Hellinger PCA. In EACL.
Marco Baroni, Georgiana Dinu, and Germán Omer Levy, Yoav Goldberg, and Israel Ramat-
Kruszewski. 2014. Don’t count, predict! A Gan. 2014. Linguistic regularities in sparse and
systematic comparison of context-counting vs. explicit word representations. CoNLL-2014.
context-predicting semantic vectors. In ACL. Kevin Lund and Curt Burgess. 1996. Producing
Yoshua Bengio. 2009. Learning deep architectures high-dimensional semantic spaces from lexical
for AI. Foundations and Trends in Machine co-occurrence. Behavior Research Methods, In-
Learning. strumentation, and Computers, 28:203–208.
Yoshua Bengio, Réjean Ducharme, Pascal Vin- Minh-Thang Luong, Richard Socher, and Christo-
cent, and Christian Janvin. 2003. A neural prob- pher D Manning. 2013. Better word represen-
abilistic language model. JMLR, 3:1137–1155. tations with recursive neural networks for mor-
phology. CoNLL-2013.
John A. Bullinaria and Joseph P. Levy. 2007. Ex-
tracting semantic representations from word co- Tomas Mikolov, Kai Chen, Greg Corrado, and Jef-
occurrence statistics: A computational study. frey Dean. 2013a. Efficient Estimation of Word
Behavior Research Methods, 39(3):510–526. Representations in Vector Space. In ICLR Work-
shop Papers.
Dan C. Ciresan, Alessandro Giusti, Luca M. Gam-
bardella, and Jürgen Schmidhuber. 2012. Deep Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg
neural networks segment neuronal membranes Corrado, and Jeffrey Dean. 2013b. Distributed
in electron microscopy images. In NIPS, pages representations of words and phrases and their
2852–2860. compositionality. In NIPS, pages 3111–3119.
Ronan Collobert and Jason Weston. 2008. A uni- Tomas Mikolov, Wen tau Yih, and Geoffrey
fied architecture for natural language process- Zweig. 2013c. Linguistic regularities in con-
ing: deep neural networks with multitask learn- tinuous space word representations. In HLT-
ing. In Proceedings of ICML, pages 160–167. NAACL.
Ronan Collobert, Jason Weston, Léon Bottou, George A. Miller and Walter G. Charles. 1991.
Michael Karlen, Koray Kavukcuoglu, and Pavel Contextual correlates of semantic similarity.
Kuksa. 2011. Natural Language Processing (Al- Language and cognitive processes, 6(1):1–28.
most) from Scratch. JMLR, 12:2493–2537. Andriy Mnih and Koray Kavukcuoglu. 2013.
Scott Deerwester, Susan T. Dumais, George W. Learning word embeddings efficiently with
Furnas, Thomas K. Landauer, and Richard noise-contrastive estimation. In NIPS.
Harshman. 1990. Indexing by latent semantic Douglas L. T. Rohde, Laura M. Gonnerman,
analysis. Journal of the American Society for and David C. Plaut. 2006. An improved
Information Science, 41. model of semantic similarity based on lexical
John Duchi, Elad Hazan, and Yoram Singer. 2011. co-occurence. Communications of the ACM,
Adaptive subgradient methods for online learn- 8:627–633.
ing and stochastic optimization. JMLR, 12. Herbert Rubenstein and John B. Goodenough.
Lev Finkelstein, Evgenly Gabrilovich, Yossi Ma- 1965. Contextual correlates of synonymy. Com-
tias, Ehud Rivlin, Zach Solan, Gadi Wolfman, munications of the ACM, 8(10):627–633.
and Eytan Ruppin. 2001. Placing search in con- Fabrizio Sebastiani. 2002. Machine learning in au-
text: The concept revisited. In Proceedings tomated text categorization. ACM Computing
of the 10th international conference on World Surveys, 34:1–47.
Wide Web, pages 406–414. ACM. Richard Socher, John Bauer, Christopher D. Man-
Eric H. Huang, Richard Socher, Christopher D. ning, and Andrew Y. Ng. 2013. Parsing With
Manning, and Andrew Y. Ng. 2012. Improving Compositional Vector Grammars. In ACL.
Stefanie Tellex, Boris Katz, Jimmy Lin, Aaron
Fernandes, and Gregory Marton. 2003. Quanti-
tative evaluation of passage retrieval algorithms
for question answering. In Proceedings of the
SIGIR Conference on Research and Develop-
ment in Informaion Retrieval.
Erik F. Tjong Kim Sang and Fien De Meul-
der. 2003. Introduction to the CoNLL-2003
shared task: Language-independent named en-
tity recognition. In CoNLL-2003.
Joseph Turian, Lev Ratinov, and Yoshua Bengio.
2010. Word representations: a simple and gen-
eral method for semi-supervised learning. In
Proceedings of ACL, pages 384–394.
Mengqiu Wang and Christopher D. Manning.
2013. Effect of non-linear deep architecture in
sequence labeling. In Proceedings of the 6th
International Joint Conference on Natural Lan-
guage Processing (IJCNLP).