Duy Le <ledduy@gmail.

com>

Writing Technical Papers in English

21 messages

Duy-Dinh Le <ledduy@gmail.com> Thu, Nov 17, 2005 at 1:06 PM

To: "The CVPR Group at HCMUNS, VN" <cvpr-hcmuns-vn@googlegroups.com>
Cc: ledduy@grad.nii.ac.jp, thnhi@fit.hcmuns.edu.vn

Ti c may mn c c hc mt lp v k nng vit paper bng ting Anh Todai (Uni. of

Tokyo). Lp ny dnh cho SV nm cui ca Todai nhng vi ti th cc k b ch. Ging vin l
ngi Anh, c thm nin trong vic rewriting cc paper ca lnh vc Tin hc. Rewriting l vic
c v chnh li cc paper cho cc tc gi, c bit l cc tc gi m ting Anh khng phi l
ting m trc khi submit i journal, conference. NII ch ti hc, Todai v ngay c INRIA,
vic nghin cu Php cng u c dch v ny.

C nhng iu m ti tng phi vt ln khi t mnh vit cc paper, v d: t tn tiu lm sao,

vit abstract, introduction th no, cch dng th (tense), ... u c cp hu ht trong mn
hc ny. Ti rt mun chia s nhng iu hc c vi cc bn v ti bit cc bn chc cng
gp khng t kh khn nh ti.

Ti rt s c cc email di d vn trong c th v n u i na, cho nn chn, ti

s gi tng phn nh trong tng email. C nhiu ngi c email bng ting Anh khng cng
chn nn ti s vit gii thiu bng ting Vit trc ri mi phn ting Anh nguyn gc vo.

By gi, xin c bt u bng cu chuyn: Vit g trong paper?

Paper y gii hn li trong cc paper m nhng ngi lm khoa hc chng ta mun gi

(submit) vo cc conference hay journal. Mc tiu khi gi cc paper thng l cng b cc kt
qu nghin cu. Chnh v vy m cn phi lm ni bt trong paper ci m chng ta cn cng b.
Ting Anh gi my th ny l thesis, argument hay claim.

Remember that your main aim is usually to persuade your readers of something. For this reason,
a paper should be based on a thesis, or main argument. A paper that only states facts is not a
good paper . You must draw some conclusions from those facts.

Vy hiu th no l thesis?

A thesis is a claim or proposition that you try to prove. It tells the reader why the work you did is
orginal, useful, or important. Generally, it is your answer to a question you explore in your paper.
It could be an application, a comparative evaluation, a unique feature, a proposal, a trend, or an
analytical finding, for example.

c v d sau y s thy kh nhiu trng hp, khi chng ta bt tay vo nghin cu, hay vit
paper, hu nh chng ta ko n thesis

For example, "Our new face recognition algorithm can improve the efficiency of police work" is a
thesis, but "we designed a face recognition algorithm" is not. "Our flying-height tester is more
accurate than existing testers" is a thesis, but "Our flying-height tester uses an adaptive detection
algorithm" is not. In other words, a thesis should explain the significance of your work.

The thesis should be stated explicitly in one or more parts of the paper (particularly in the
abstract and conclusion). It should also control the structure of the rest of the paper; in other
words, the paper should consist of a series of steps showing why your thesis is correct.
 t tn paper (title) nh th no? Build your title and abstract around the thesis

Xem trong email ti!

L nh Duy

PS: Mi ngi c t nhin tho lun nh!

Thanh V. Nguyen <ngvietthanhvn.uom@gmail.com> Thu, Nov 17, 2005 at 2:14 PM

Reply-To: cvpr-hcmuns-vn@googlegroups.com
To: cvpr-hcmuns-vn@googlegroups.com

Hi all, nhn tin thy Duy nhc n thesis, tontot tranh th hi lun :-D
y l mt rank nh gi ca mt hi ngh,
[ ] Very original
[ ] Moderately original
[ ] Derivative
[ ] Very derivative

Trong cc trng hp sau th thng n c gn vo mc no

+ Apply mt(hoc nhiu) cch tip cn bi ton ny vo bi ton khc
+ Thay i mt phn trong hng tip cn c cho kt qu tt hn
+ Hng tip cn c ng l ra tt nhng v cha c th nghim y , nay a ra mt th
nghim y v chng minh n tt
...

V thng thng th nh th no s c chp nhn ti mt conf tt

VThnh

Duy-Dinh Le <ledduy@gmail.com> Thu, Nov 17, 2005 at 2:23 PM

To: cvpr-hcmuns-vn@googlegroups.com

Nhn tin Dun ni v thesis statement, cng nn ni r hn v t thesis.

Trc gi ti vn ngh thesis l lun vn, ci ny cng ko sai. Nhng ng l ngi ta dng
thesis trong nhiu ng cnh khc na. Hy xem Wiki ni g v thesis:
http://en.wikipedia.org/wiki/Thesis

In academia, a thesis or dissertation is a document that presents the author's research and
findings and is submitted in support of candidature for a degree or professional qualification. A
thesis statement is the statement that begins a formal essay or argument.

At Canadian universities, writings presented in fulfillment of undergraduate coursework

requirements are normally called papers, term papers or essays. A long paper presented for
completion of an honours degree is sometimes called a major paper, or, more rarely, an
undergraduate thesis or honours thesis. Major papers presented as the final project for a masters
degree are normally called theses; and major papers presenting the student's research towards a
doctoral degree are called dissertations.

At UK universities, the term thesis is usually associated with a Ph.D. (doctoral ) or M.Phil. degree,
while dissertation is the more common term for the research project required for an M.A. or other
postgraduate degree. The equivalent for undergraduate degrees is usually referred to as a
 research project.

In many US doctoral programs, the term dissertation can refer to the major part of the student's
total time spent (along with 2-3 years of classes), and may take years of full-time work to
complete. At some universities, dissertation is the term for the required submission for the
doctorate and thesis refers only to the master's degree requirement. At many others, the word
thesis is used for both.

L nh Duy

Duy-Dinh Le <ledduy@gmail.com> Thu, Nov 17, 2005 at 2:40 PM

To: cvpr-hcmuns-vn@googlegroups.com

H h, Vit Thnh hi cu ny kh qu. Ti cng i tm cu tr li cho cu hi ny rt lu ri.

Nhng rt tic vn cha ra v ti ch bit da vo kinh nghim bn thn. C l ai c Prof. xn th
hi trc tip i.

Theo ti bit th c accept bi mt conf. tt (rank 1) th ngoi yu t Original ra, cn nhiu

yu t khc quyt nh na, v d nh Evidence, Theoretical Background, ... Nhng cc bi m
c oral th chc l cc yu t quan trng trn u t.

V cc cu hi ca VThnh, nu hiu ng ngha original - mi. Nh Wiki ln na:

Originality refers to something being new or novel. It is not received from others nor copied from
the creations of others. The word is often applied in an admiring fashion to the creations of artists,
writers and thinkers.

Theo ti ngh, nu tng ("thesis") ca mnh m cha c ai ng ti th c xem l original.

C nhng ci cha ai ng ti nhng m n c v d suy ra c t nhng ci ngi ta lm
th ko c xem l original.

+ Apply mt(hoc nhiu) cch tip cn bi ton ny vo bi ton khc

Nu bi ton khc y l bi ton mi m v cc cch tip cn lin quan trc ko cho kqu
tt bng cch tip cn trong paper ca mnh th vn c th xem l mi. Ly v d bi ton face
detection ca Viola. tng l xi AdaBoost, Haar wavelet v Integral Image. AdaBoost c
ncu bn Machine Learning, th nghim dng cho text categorization, Haar wavelet dng
trong Computer Vision nhng cha dng cho face, Integral Image dng trong Computer
Graphics. Tuy nhin khi Viola dng 3 mn ny cho bi ton face detection, tt nhin l c hiu
chnh cht t, th mang li hiu qu cc ln. l chy nhanh gp c trm ln cc cch tip cn
c, trong khi kt qu vn tng ng, thm ch hn. Hn na, tng ny c th m rng ra
cho cc bi ton khc ko ch cho face. Ti ngh, nu l ti, ti s cho l Very original.

+ Thay i mt phn trong hng tip cn c cho kt qu tt hn

+ Hng tip cn c ng l ra tt nhng v cha c th nghim y , nay a ra mt th
nghim y v chng minh n tt

Nu kt qu khng mang li vt bc th ko c nh gi l original.

Ni chung, mt khi mnh thng tho (master) c mt lnh vc hp no , bit ht cc

chiu s v ngun gc ca n ri th lc s t nhn ra c chiu no l chiu mi c gi tr,
chiu no l chiu tm tm.

L nh Duy
Duan Tran <tdduan@gmail.com> Thu, Nov 17, 2005 at 3:00 PM
Reply-To: cvpr-hcmuns-vn@googlegroups.com
To: cvpr-hcmuns-vn@googlegroups.com

Nhn tin bn n paper, Bico cng a lun 1 vn kh nhy cm l Plagiarism - tm dch

l "o vn". Phn ny chc l lp writing ca anh Duy cng c hc ch h (lp writing em ang
hc cng c hc).

Mi ngi c th tham kho thm trang ACM

http://www.acm.org/pubs/plagiarism%20policy.html mi va c cp nht vo cui thng 10, v
va ri c mt s trng hp paper o ACM b pht hin Plagiarism. Tt c cc prof bn UIUC
u c gi mail yu cu cnh bo SV ca mnh nn trnh.

Mt im cng hi mi l sefl-plagiarsm:
Self-plagiarism is a related issue. In this document we define self-plagiarism as the verbatim or
near-verbatim reuse of significant portions of one's own copyrighted work without citing the
original source2. Note that self-plagiarism does not apply to publications based on the author's
own previously copyrighted work (e.g., appearing in a conference proceedings) where an explicit
reference is made to the prior publication.3 Such reuse does not require quotation marks to
delineate the reused text but does require that the source be cited.

Nu p dng cht ch self-plagiarism th chc kh nhiu papers b xem nh l plagiarism.

Dun

Duy-Dinh Le <ledduy@gmail.com> Tue, Nov 22, 2005 at 10:59 AM

To: cvpr-hcmuns-vn@googlegroups.com
Cc: thnhi@fit.hcmuns.edu.vn, hn220371@yahoo.com

Tranh th vit tip phn t tn papers.

Build your title and abstract around the thesis

t tn paper (title) cng l bc quan trng khng km, bi v n l thng ip u tin m cc

tc gi mun gi n ngi c. Hy tng tng chng ta nhn vo hng trm papers ca mt
proceeding/journal hay hng chc trang tr ra ca Google search, mt title n tng s rt quan
trng v n gip thu ht c gi ngay ln p vo mt u tin. Vy lm th no?

Try to emphasize your thesis in the title and abstract. Here's a very short example that shows
how you can do this:

Demonstrational Interfaces: A Step beyond Direct Manipulation

Abstract: With demonstrational interfaces, the use applies direct-manipulation techniques to

abstractions by operating on example values. Applications range from text editing to visualization.

This title tells us the advantage of demonstrational interfaces: they are better than direct
manipulation. Without even reading the abstract, we already know the thesis. In the abstract, we
learn from just two sentences what demonstrational interfaces are and what kind of applications
they have.

Hy th phn tch mt vi poor examples m chng ta c th tng tri qua.

Development of the Autonomous Vehicle PVS (Personal Vehical System)

Abstract: The PVS (Personal Vehical System) travels to a destination guided by detecting white
lines and any obstacle on the road from image data and by referring to the map data. Fuzzy
inferences are used for steering control, in order to achieve human-like maneuvers. The PVS can
move at 60km/h on a straight road and at 30km/h on a curved road .

The title tells us only that the PVS was developed, not why it was developed. The abstract tells
us what it can do and how it does it, but does not compare the PVS with other PVSs or
discuss its advantages or applications . The reason for the human-like maneuvers is also not
given. In short, there is no thesis

Another problem is that, except for the mention of fuzzy inferences, there is no information about
the principles on which the system is based. A technical paper, unlike a technical report,
should generally emphasize theoretical rather than practical considerations .

C th thy rng, ch cn nhn gc writing, cha cn n ngi rnh v chuyn mn

(experts), cng c th nh gi rng title v abstract ny cha t yu cu. Chng ta khng bit
contribution ca paper ny l g. Nhng con s 60km/h hay 30km/h khng ni ln c ngha
g nu chng ta khng c s so snh vi cc h thng PVS khc.

A Fuzzy-Based Expert System Building Tool

Abstract: In this paper, an integrated expert system building tool is proposed. The characteristics
of this tool are:
1. The incorporation of rule-based and fuzzy-based inference.
2. Support tools for building fuzzy knowledge bases, such as a specialized editor and a fuzzy
debugger.
3. A self-tuning algorithm for fuzzy membership functions.
A fuzzy-based technical analysis support system is described as a typical application of this tool.

This time, the title is a bit more informative: it tell us that the tool uses fuzzy logic. However, it
doesn't tell us why it uses fuzzy logic, or whether any other tools do so. The abstract tells us
something abount the principles used, and about an application, but again, there is no real
thesis-no claim that requires any proof.

Cu cui bnh lun rt ng, nhn vo abstract ta khng tm c tc gi mun chng minh ci g
c.

The title of a paper should be

- Attractive
- Descriptive:
If possible, refer to the aim of the paper, the approach, and any special features. For instance,
instead of
A Novel Handwriting OCR System
you could write
Improving Automatic Handwriting Recognition through Stroke Matching

Descriptive words such as new, novel, fast, unique, accurate, efficient, reliable and simple
are useful for focusing the title.

- Short: Ideally, the title should fit into one line.

- Consistent with the text
- Easy to understand
- Easy to retrieve from a database: Try to include one or more keywords in the title, so that
someone looking for papers on, say, data hiding will be able to find your paper easily.

Mt s lu lin quan n vic trnh by title:

- Vit hoa cc k t u mi t,
- Khng vit hoa cc gii t (v d: of, for, in, at, to, from), conjunctions (v d nh and, but, or,
so), infinitve (v d to), article (v d a, an, the). Ngoi l nu l t u ca title.
 - Nu h thng c tn vit tt, vit Name: Description

Xem mt s v d sau hiu hn:

Do Object-Oriented Languages Need Special Hardware Support?

Information Flocking: Data Visualization in Virtual Worlds Using Emergent Behaviours
CliPPS: A Clinical Planning Support System
Discourse as a Knowledge Resource for Sentence Disambiguation
A Survey of Pre-electronic Calculation Devices.

K ti, Vit abstract nh th no?

L nh Duy

Duy-Dinh Le <ledduy@gmail.com> Wed, Nov 30, 2005 at 11:33 AM

To: cvpr-hcmuns-vn@googlegroups.com
Cc: thnhi@fit.hcmuns.edu.vn

Nhn dp tm ti liu v plagiarism, tm thy c mt s ti liu ny rt hay ca Indiana

University http://www.indiana.edu/~wts/pamphlets.shtml . Ni dung kh y t cc vit thesis
statement n cc li ng php, trnh plagiarism, ...

WTS Pamphlets
Each pamphlet deals with a specific writing situation. Pamphlets are available in html and pdf
versions. To read the pdf versions, you will need Adobe Acrobat Reader . Students may also pick
up paper versions of pamphlets at WTS. Faculty may contact WTS to order paper copies for their
classes.

"Citing Sources in MLA Style": html, pdf

"Citing Sources in APA Style": html, pdf
"How to Use Evidence": html, pdf
"How To Write a Thesis Statement": html, pdf
"Paragraphs and Topic Sentences": html, pdf
"Personal Statements and Application Letters": html, pdf
"Plagiarism: What It Is and How To Avoid It": html, pdf
"Proofreading for Common Surface Errors": html, pdf
"Proofreading for Spelling Errors": html, pdf
"Taking an Essay Exam": html, pdf
"Using Outlines": html, pdf
"Writing Book Reviews": html, pdf
"Writing Resumes and Cover Letters": html, pdf

L nh Duy

Duy-Dinh Le <ledduy@gmail.com> Sat, Mar 11, 2006 at 11:54 AM

To: cvpr-hcmuns-vn@googlegroups.com

Hm qua tnh c chat vi MSn tm c mt s guidelines cho vic vit paper ca MMSP. Post
ln y mi ngi tham kho

For authors:

CONTENT REQUIREMENTS
The papers submitted to MMSP'2006 MUST contain the sections below (in the same or in a
different order). A Background section may be added if needed.

1) Abstract

up to 150 words
State the proposed method and the advantages as reflected in the experimental
results
DO NOT describe the method
DO NOT list here experimental values, comparisons, etc.

2) Introduction

State the problem

Discuss previous work if any (if not, state that, to the best of your knowledge, there
are no other attempts to solve the problem )
State briefly how the proposed method solves the problem
State how the proposed method is different than previous works

3) Proposed method

Enumerate what this section contains

Describe each component one by one
If appropriate, include a flowchart, table, or algorithm

4) Experimental results and discussion

State the implementation details first (system and test set characteristics, criteria for
experimental evaluation, etc).
Present the experimental results and discuss them
Compare your results with at least one competing, state-of-art method

5) Conclusions and future work

Similarly to a patent application, enumerate the "claims" (aka, contributions) that the
paper makes
Outline the main directions for future improvements

Nh vy c th thy rng:

- Trong phn abstract, chng ta ch yu ni v proposed method (phng php m chng ta

xut) ng thi nu r n c thun li hn ngi khc ch no

- Phn Introduction l phn rt quan trng, trong phi lm ni bt c hai chnh: th nht
l work ca mnh khc work ca ngi ta nh th no. Cho d cc work khc ko lin quan nhng
cng phi tho lun l ti sao ko lin quan ( chng minh work ca mnh ang lm l mi).
th hai l ngoi vic nu proposed method, cn phi gii thch ti sao method c th gii
quyt c problem. Ngi ta s khng thch th lm khi nghe bn report rng system ca ti c
chnh xc 99% m ch thch th khi bn gii thch c ti sao system ca bn t c kt
qu cao nh vy.

- Phn Experiments l phn chng minh nhng g mnh trnh by trn l ng. lm vy,
i hi phi c comparison. Chng ta phi show c s vt tri ca chng ta so vi t nht
mt state-of-the-art method. Lu l phi so vi state-of-the-art ch khng phi so snh vi cc
 phng php tm thng. Chnh v vy m vic tm c cc paper cc top conf, top journal rt
quan trng v cc state-of-the-art thng c publish .

Phn ti s bn v Reviewers

L nh Duy

Duy-Dinh Le <ledduy@gmail.com> Sat, Mar 11, 2006 at 12:06 PM

To: cvpr-hcmuns-vn@googlegroups.com
 For reviewers

The goal of the guidelines included next is to aid the review process by providing a set of
compact, consistent, and clear rules regarding the content and format requirements of the
submitted papers. We ask that the reviewers:

1) verify that each paper under review contains the sections and information listed under "Content
requirements" in the instructions "For authors":

Any paper that does not meet these requirements should be rejected, regardless of
the topic discussed. Please mention the missing section and/or missing information in
the comments to the technical chairs that are part of the review form.

2) be very critical in evaluating the papers:

DO accept papers only if they are good and very good in terms of the method proposed,
originality, organization, soundness of the experimental evaluations, clarity of the
presentation and language
DO accept papers only if they are either original or, are non-original but have
a different point of view on existing work
DO accept papers only if they have thorough experimental evaluations
DO accept papers only if they contain comparisons of the experimental results with
at least one of the state-of-art methods
DO accept papers only if they DO NOT require major re-work in terms of the
presentation and language

C th thy rng:
- Phn experimental evalution l mt trong cc phn quan trng nht m reviewers s da vo
nh gi papers. Cng d hiu thi, mt khi bn mun claim rng phng php ca ti l xn,
l ngon th bn phi c evidence chng minh. V d bn mun claim rng VN c th sn
xut mt chic xe my tt. Vy th bn t nht phi chng minh c thng qua 2 : Mt l cc
thng s k thut lin quan n vic chng minh tnh "xn" ca xe my phi c evaluate thng
qua thc nghim mt cch cn thn, chi tit v y (thorough evaluations). Hai l bn phi so
snh n vi t nht mt loi xe my c cho l xn hin c trn th trng, v d nh Honda,
Suzuki. Nu bn i so xe ca bn vi xe Trung Quc th s chng c ngha g c.

- th hai ti thy rt hay, l bn v originality. Ci ny c th tr li phn no thc mc ca

Vit Thnh k trc. l ngi ta chp nhn nhng ci: non-original but have a different
point of view on existing work. Ti s c gng kim v d minh ha v ny sau. Ai c paper
thy th post ln chia s nh.

- V phn writing, h ch reject khi m chng ta vit qu t m thi. Ti ngh rng chng ta cng
ko n ni.

L nh Duy

Duy-Dinh Le <ledduy@gmail.com> Fri, Mar 31, 2006 at 12:02 AM

To: cvpr-hcmuns-vn@googlegroups.com

@Tn mn v chuyn vit lch

Vit lch l chuyn khng th trnh khi trong cuc sng ca chng ta. Ty tnh cht cng vic v
ngh nghip m mc i hi k nng vit khc nhau. V d nu chng ta sut i ch lm Ton
th vit lch n gin hn nhiu, trong khi nu lm nh bo, nh vn th vit lch phi cn hc
 trng lp hn hoi. Chng ta, nhng ngi lm Tin hc, nu mun c cng ng quc t
cng nhn th cng phi hc vit lch publish papers. Mc d khng n ni phc tp nh vit
cc tiu thuyt, truyn ngn nhng chuyn vit lch cho cc conferences, journals cng cn phi
hc bi v n cng c nhng chun mc ring. Lot bi Writing Technical Papers in English (
c phi login vo GMail) cng phn no phn nh mc phc tp ca chuyn vit lch ny
ri.

Mi ngi chng ta ai cng t nht mt ln vit lun vn tt nghip ri, chc cng thm tha vic
vit c ci g coi c i hi rt nhiu cng sc v thi gian. Cng ging nh trong
bng , chng ta c th d dng i ny vit km, i kia hay, i n lc hay lc d.
Nhng phi nm i bng ri th mi thy chuyn hay, d n phc tp n mc no.
V trn ht, l vn v ng cp. Vit lch cng vy, ngoi ti nng bm sinh, t n
mt ng cp no , khng c g khc hn l phi rn luyn v c xt thng xuyn. chnh
l l do m ti ang c gng duy tr vic vit lch ca mnh trn groups, trn blogs, d rng n tn
rt nhiu thi gian.

Ti ngh vic chng ta cng c cng s thch nghin cu v CVPR, NLP, ... v c cc ch
ncu trong qu trnh hc tp ca mnh cng c cm hng cho rn luyn vic vit lch. C
th cc bn s thy ngi ngn khi ngh rng mnh vit ra s ngi khc ch l vit v vn, l
ko ng vit, v.v..., ti ngh cm gic t ti ai cng c c. Tuy nhin chng ai c th mt lc
m thnh "cao th v lm" c c, ng d ngn nhng nu khng i th khng
th n c.

Hi vng mi ngi s vit bi nhiu hn ng gp cho group.

L nh Duy

Duy-Dinh Le <ledduy@gmail.com> Tue, Apr 4, 2006 at 3:14 PM

To: cvpr-hcmuns-vn@googlegroups.com
Cc: nhtanh@fit.hcmuns.edu.vn, mvcuong@fit.hcmuns.edu.vn, tmtriet@fit.hcmuns.edu.vn

@Avoiding Rejection
Avoding Rejection l mt bi vit rt hay ca J. Hendler lin quan n vic publish papers cc
conf, journal. Trong bi vit ny tc gi lit k ra mt s li thng gp trong cc submission m
c th dn n vic paper b reject, ng thi gii thch l do ti sao ngi ta li reject cc
submission nh vy. y l t liu rt ng qu cho chng ta khi bt tay vo vit papers.

Mi ngi s c cm nhn ring ca mnh khi c bi vit ny. y ti list ra mt s m ti

cm thy tm c v n gip mnh verify/modify li li cc suy ngh m mnh vn thc mc khi xa
m cha c li gii p.

Cite the right literature

Nh ti thng hay cp, trc khi bt tay vo hin thc mt vn g, ta nn look around xem
ngi ta lm c n u, kt qu th no, trnh pht minh li ci bnh xe! Qu trnh ny
trong NCKH gi l review state of the art (state of the art c th hiu l nhng k thut mi nht,
tin tin nht). Ngi c rt mun bit mi lin quan gia vn chng ta ang lm v cc vn
c. Cng c mt s ngi lp lun, ci ti lm l mi hon ton, trc gi cha ai
lm nn khng cn thit phi so snh vi nhng ci c bin minh cho vic khng
tm c cc papers lin quan . c on sau, chng ta s thy Hendler tr li nh th no v
lp lun trn.

First and foremost, many papers simply don't do a good job of situating the work in the greater
research milieu. This can be something as egregious as having no references to other work
at all or as subtle as missing a key reference. In the former case, authors have replied to my
rejections complaining that, as one author put it, "comparison with existing literature also
wouldn't help much because the approach of this model is very different from those
discussed in the literature."

Even if this is true of a piece of novel work, the author still has the responsibility to help
the reader understand why . As I responded to this author (text changed slightly to provide
anonymity),

How would our readers who aren't experts in the field know this? For example, suppose one of
them has heard a talk by <a researcher> about his <related> model and wants to know what's
different between that work and yours. You know it's different, but on the surface, there's
much that a reader could confuse. So, you can help readers understand how your work
compares and convince them that you're aware of the state of the art, so that they know you're
not just reinventing something out there (it's your responsibility as author, not theirs as reader, to
place the work in this context).

Do the right evaluation

Thng thng, ni dung mt paper s xoy quanh mt thesis m trong tc gi s claim nhng
g thuc v original contribution ca mnh. Mt phn khng th thiu trong paper l phi c
nhng bng chng (evidence) support cho cc claims. Claim nh th no th evidence phi
consistent (nht qun) th y. Nu chng ta cho rng cch tip cn ca chng ta l mi th phi
demo cho ngi ta thy n mi ch no, ch no nhng ci ang c khng c. Nu chng ta
cho rng cch tip cn ca chng ta hn cch tip cn trc th phi c so snh cho thy
iu !

A close contender for the most common cause of rejection, narrowly trailing literature review
flaws, is lack of evaluation. There are many different ways to evaluate a piece of research and
no one-size-fits-all solution to ensuring that the work is sound. Some cases require theoretical or
mathematical analysis, others require an experimental result or a user study, and sometimes all
that's needed might be just a strong demonstration. Deciding which approach to use, and the
key to getting your paper accepted, is simple justify your claims.

In deciding how to evaluate your research for the paper, ensure that the paper shows you can
achieve the justification you've asserted for the research. Publishing an empirical graph showing
how fast your system works or delivering a proof that the mathematics is correct isn't only
sometimes unnecessary but also is often insufficient. It all depends on what you're claiming
your new approach can achieve.

If you're claim claiming that your approach does something new, then all you need is a good
strong demonstration that your approach can do it.

If you claim your approach is superior to previous approaches, then your evaluation must
prove this.

(Hai ny rt hay, kim chng li cc paper c, chc nm ch yu trong 2 ny, hoc

l something new hoc l superior. Suy ra, lc mnh bt tay vo ncu, mnh cng th
lng sc xem ci mnh lm ra n nm loi no nh?)

Designing an appropriate evaluation is part of the art of good science and isn't always
easybut it's always needed. The heuristic of tailoring your evaluation to your claims (or
tailoring your claim to your evaluation design), however, is usually a good one. Let's take a
somewhat artificial casesuppose an author claimed a major breakthrough in knowledge
representation. How in the world might you evaluate such a thing?
 l 2 quan trng nht trong bi vit ca Handler. Cc bn c bi vit c cm nhn th
no th cng post ln chia s nh!

L nh Duy

PS: Ti bit c bi vit ny qua Hunh Trung ng. Bi vit ny c Prof. N. Jenning l
supervisor ca ng khuyn cc SV ca ng nn c.

Duy-Dinh Le <ledduy@gmail.com> Thu, Apr 13, 2006 at 10:00 PM

To: cvpr-hcmuns-vn@googlegroups.com

Ti sp phi submit thesis (hi vng vy) vo gia thng 5 ti. Tnh c tm kim trn Internet thy
c 2 hay:

1. Mt s kinh nghim trong vic vit lch, c ti liu, etc ca Minh (sinh nm 1974, tt
nghip PhD EPFL, hin l Ass. Prof ca UIUC) dnh cho GRA (Graduate Research Assistant)
ti UIUC.

Xem online y: http://www.ifp.uiuc.edu/~minhdo/teaching/GRAadvice.pdf hay trong tp tin nh

km.

2. Cc thesis template bng LaTex ca mt s trng H nh UC, UIUC, GaTech, etc c th

download c ti y: http://www.ctan.org/tex-archive/macros/latex/contrib/

L nh Duy

MinhDo06-GRAadvice.pdf
39K

Duy-Dinh Le <ledduy@gmail.com> Thu, May 4, 2006 at 5:56 PM

To: cvpr-hcmuns-vn@googlegroups.com

@ Art of Writing (Ngh thut vit lch)

Mt trong nhng cng on cui cng ca qu trnh research l publication trong writing ng
vai tr rt quan trng. Cng l mt truyn tnh tay ba o le mun thu nhng vo tay cc nh lm
phim Hn Quc th coi c my chc tp khng thy chn, nhng vo tay cc nh lm phim VN th
coi tp u tin l thy khng cn g coi na. Vit paper cng vy, c nhiu vn chng ta
hiu rt r nhng khi vit ra, c ln chng thy c cht g cm hng v hp dn c, trong khi
cng l vn nhng m do cc i ao cao th vit th n li thy hay ho khng k xit.
Trong thread ny, ti mun dn chng mt s v d minh ha cho trn.

Trc ht, cng nn nhc li y v ci c gi l gi tr ca mt research work (nhn t lnh

vc CVPR). l phng php xut (proposed method) cn phi general, ngha l n
c th p dng cho nhiu loi bi ton khc nhau . Mt phng php dnh cho face detection
s th v v c mi ngi quan tm nhiu hn nu n tng qut pht trin cho car
detection, animal detection, etc. Cng chnh v vy m c th mt phng php dng cho face
detection c report cho kt qu n 99% s khng c ch bng mt phng php ch t
kt qu hn 90% nu phng php u dng qu nhiu heuristic specific to face trong khi
phng php th hai li tng qut hn.

Cng chnh v l do ny m trong cc paper, bn cht vn l face detection nhng tiu

v cch vit li rt general, bn v object detection. iu ny s d lm cho cc bn mi bt
tay vo nghin cu nhm ln khi suy ngh v chn ti v tng nghin cu. Hi ti cn
VN, cng hay m mng kiu ny, ngha l m mng mnh c th develop methods m c th
detect general objects. V lc suy ngh v cc thut ton, bao gi cng ngh v general object
ch ko gn vi mt loi object c th no, v d nh face hay car

Thc t th li khc hon ton. Thng thng ngi ta s bt u bng mt loi object c th no
, v d nh face, car ri t khm ph ra cc vn /problems lin quan n n. T cc
problems , h s generalize thnh cc bi ton nghin cu/research problems. Sau khi c cc
research problems, h s tm cch a ra solutions gii bi ton . chng minh solution
l ng n, h s p dng solution tr li bi ton cho object l face hay car ban u. Ti ly
mt v d minh ha nh sau:
Real world problem: To detect face from images
==> Research problem: How to extract features efficiently to model object classes (one of them is
face)?
==> Solution: Using Haar-wavelet
==> Evidence: Using Haar wavelet to detect face

C th thy rng, cng mt real world problem, c th hnh thnh nn nhiu research problems
khc nhau, v cc solutions tng ng cng ch gip gii quyt mt phn no ca real world
problem m thi. cng chnh l l do m cng l bi ton face detection, nhng m papers v
n th v s k. (Khc vi VN trc y, nu c mt lun vn H ca SV v face detection
ri th nm tip theo, khng ai dm bn v n na). Ngoi ra, iu quan trng nht nn rt ra l
chng ta nn bt u research t cc bi ton, ng dng c th ch khng th ngh chung
chung, tng qut c. Bi v c bi ton c th, c ng cnh, chng ta mi c iu kin
kim chng cc l thuyt, cc phng php m chng ta pht trin cho n .

di y, ti dn chng ra 3 papers bn v object detection minh ha cho cc trnh by

trn.

[1] T. Evgeniou et al, Image Representations and Feature Selection for Multimedia Database
Search , TKDE, 2003
[2] S. Argawal et al, Learning to Detect Objects in Images via a Sparse, Part-Based
Representation , PAMI 2004
[3]. H. Schneiderman and T. Kanade, Object Detection Using the Statistics of Parts , IJCV 2004

paper [1], bn cht vn l face/people detection nhng c tc gi khi qut ha thnh bi

ton rt tng qut cho Multimedia Database Search. Chng ta s thy s ti tnh ca tc gi khi
m u bi vit bng multimedia data analysis nhng sau li focus vo detecting real-world
objects in images v cui cng l face and people detection

An important research issue in the field of multimedia data analysis is that of choosing the right
representations for the data (images, sounds, video, etc). Whether it is for searching, indexing,
comparison, etc., it is clear that the way the multimedia data are represented can significantly
influence the performance for the various data analysis tasks. In this paper, we focus on a
particular task, namely, that of detecting real-world objects in images for example for the
purpose of searching in image databases, and we study the problem of choosing image
representations for this task. Although we focus on object detection in images, a difficult
multimedia analysis problem, the methods and experiments we present can be applied to any
multimedia analysis task.
...
As already mentioned, a major issue in such systems is choosing an appropriate image
representation. We investigate the role of image representation for object detection using kernel
machine classifiers such as SVM. In particular, we present experimental results comparing
different image representations for both face and people detection.

Hay trong paper [2], bi ton nguyn thy l car detection c tng qut ha thnh cho bi ton
object detection ni chung, y l phn abstract:

We study the problem of detecting objects in still, gray-scale images. Our primary focus is
the development of a learning based approach to the problem that makes use of a sparse, part-
 based representation. A vocabulary of distinctive object parts is automatically constructed from a
set of sample images of the object class of interest; images are then represented using parts from
this vocabulary, together with spatial relations observed among the parts. Based on this
representation, a learning algorithm is used to automatically learn to detect instances of the object
class in new images. The approach can be applied to any object with distinguishable parts in a
relatively fixed spatial configuration; it is evaluated here on difficult sets of real-world images
containing side views of cars, and is seen to successfully detect objects in varying conditions
amidst background clutter and mild occlusion. In evaluating object detection approaches, several
important methodological issues arise that have not been satisfactorily addressed in previous
work. A secondary focus of this paper is to highlight these issues and to develop rigorous
evaluation standards for the object detection problem. A critical evaluation of our approach under
the proposed standards is presented.

Paper [3] cng vy, cng xut pht t face, car detection m pht trin thnh general object
detection. Nu c k phng php, chng ta s thy d tc gi c cho rng method l general
nhng vn c khng t ch rt specific to car or face.

Object detection is a big part of people's lives. We, as human beings, constantly "detect" various
objects such as people, buildings, and automobiles. Yet it remains a mystery how we detect
objects accurately and with little apparent effort. Comprehensive explanations have defied
psychologists and physiologists for more than a century.
Our goal in this research is not to understand how humans perceive, but to create computer
methods for automatic object detection.
...
Our method for object detection factors out variation in the pose of the object. Our object
detector uses a set of classifiers, each of which determines whether the object is present at a
specific pose in a fixed-size rectangular image window. For faces, the detector uses classifiers
for three discrete poses: front, left profile, and right profile. Taking advantage of facial
symmetry, we only needed to train classifiers for the frontal and right profile viewpoints shown in
Fig. 1(a), and we built a left profile detector by reflecting the right profile detector. For cars, we
use 15 discrete viewpoints, and by exploiting symmetry again, we only trained classifiers for the
eight viewpoints as shownin Fig. 1(b).

S d ti ni y l art of writing bi v nu vit theo nhng g ta hiu v bn cht th c s thy

ko c g hp dn c. Xem 3 phn introduction ca cc paper ti trch dn trn, s thy cch
ngi ta li cun ngi c t nhng dng u tin nh th no.

Trong nhng ln tho lun vi nhau v vn ny, t-san l ngi lun cm thy t hi lng vi
chnh bn thn v thy mnh vit lc no cng direct, c khng thy so vi ngi ta. Vi kinh
nghim ca mnh, ti cng ng nh vy. Nhng cng khng nn qu bi quan v vit hay hay
d cn ph thuc vo tm nhn (cng ging nh nu bit c c c cu kim ri th nhn ra
ht c cc bin ha ca cc kim php hin c). M tm nhn cao hay thp ph thuc vo qu
trnh rn luyn theo thi gian m thi.

L nh Duy

Thuy Ngan <nltngan@gmail.com> Sun, May 7, 2006 at 5:19 PM

Reply-To: cvpr-hcmuns-vn@googlegroups.com
To: cvpr-hcmuns-vn@googlegroups.com

Cho mi ngi,

c my bi vit ca Thy Duy gi v chuyn vit lch thy hay qu! Ngn thy mnh cng
khng c thi quen vit lch, cn phi rn luyn thm . C l by gi phi tp thi quen ging
Thy Duy c c ci g hay phi ghi ch li, vit bi tng hp,...

Kin thc khng c h thng li th rt ln xn v n khi mun bt tay vit s cm thy kh

khn.
 Hc k ny Ngn hc nhiu course nn thi gian cho research khng nhiu lm nhng k hoch
l xc nh r ti Master sm. Khng bit trong nhm ngoi anh t ra c nhng bn no
quan tm n NLP khng?

THY NGN
[Quoted text hidden]

Anh Duc Duong <anhduc.duong@gmail.com> Sun, May 14, 2006 at 11:42 PM

Reply-To: cvpr-hcmuns-vn@googlegroups.com
To: cvpr-hcmuns-vn@googlegroups.com

Qu tht, bi ca Duy vit rt hay.

C iu, thy c vn kh, lm c cn kh hn.
Cng nh u phi ai cng hc c c c cu kim nh Lnh H Xung; bao nhiu ngi
luyn nhng chi c Trng V K luyn thnh Cn Khn i n di.

Ti thng hay c mt thi quen: ban u c gng tm hiu nhng ci khc nhau ca cc vn
ring bit m mnh ang nghin cu.
Khi thy r s khc nhau ri, ti li c gng tm ci chung ca chng. Ri li c gng tip tc tm
ci khc,... (nu cn c diu kin tm hiu, cng nh kh nng). Thng sau khi ti lm c
t nht 1 vng nh vy, ti cm thy mnh nm r hn vn rt nhiu.

Ti hc iu ny t mt ngi thy dy ton ca ti t nhng nm cp 2. Vi cch ny, khi hoc

cao hc, ti nut tri cc kin thc v CSDL, thit k HTTT, ... ca c Thy kh d dng v ti
nhn thy n hnh dng ca i s quan h, ci m ti quen thuc khi cn hc i hc. Khi lm
bi tp v tm kha cc quan h, ti nhn thy l thuyt th trong , ...

Va ri, khi ln Lt dy cao hc, ti c trao i vi mt thy trn ang day mn ha

my tnh. Chng ti c trao i s v cc thut ton v ng. Khi nghe ti ni KHTN dy v
nguyn l Bresenham v thut ton Bresenham (thc ra mnh thng dy MidPoint), anh bn
khoe l dy mt thut ton khc tt hn thut ton Bresenham nhiu. Ti bit ngay l
MidPoint (Duy chc cn nh r), nhng phi mt kh nhiu sc ti mi thuyt phc c anh
bn ny l MidPoint chng qua cng chnh l mt th hin ca nguyn l Bresenham thi. Nu
nhn theo cch , mnh c th ng dng Bresenham rng hn nhiu.

Hm nay ni chuyn hi lan man.

Cho cc bn nh.

Duy-Dinh Le <ledduy@gmail.com> Thu, Oct 26, 2006 at 8:37 PM

To: cvpr-hcmuns-vn@googlegroups.com

@ Presentation in English - Technical Writing

Presentation in English l tn gi mt mn hc mi trong chng trnh o to PhD ca NII. Bi
u tin do Prof. Houle trnh by lin quan n Technical Writing cng kh hay. Ti gi slide ln
y mi ngi cng tham kho.

Cc vn c tm tt kh c ng. Nu cc bn theo di thread ny t u s thy ti cng

cp phn no.

Cn nh c mt ln Prof. Houle sa gip ti paper ca CIVR. ng l hc c nhiu th. V d

khi c 2 cu m khng c lin quan g (connect) vi nhau, Prof. s hi ti v sao? Hay c nhng
cu Prof. ni rng ti vit ng php th khng sai nhng m khng th gi l readable v native
 speaker ngi ta khng dng kiu din t . Hay mt v d khc, khi ti dng hard ch
kh khn, Prof. li ni dng hard nhiu khi b confused v n cng c ngha l cng, ... cho nn
ch kh khn ng lun dng difficult. Hay ng khng dng because din t bi v m dng
since, bi v ng cho rng ngi ta dng because qu thng xuyn trong vn ni ri. Hay c
mt s t ng ni l trong vn ni ngi ta dng nhng trong vn vit formal t dng, v d nh a
lot of or lots of ko nn dng trong vn vit m thay vo dng many hay much. ng l face-to-
face discussion s gip ci thin k nng vit rt ng k.

Ti s tham d lp ny vo th hai tun sau. Hi vng s c g hay ho cng chia s vi mi

ngi.

L nh Duy

Houle06-TechnicalWriting.ppt
242K

Duy-Dinh Le <ledduy@gmail.com> Fri, Nov 24, 2006 at 11:16 AM

To: cvpr-hcmuns-vn@googlegroups.com

@ Sentences and Patterns for Writing

Trong cc sch lin quan n dy writing, ngi ta thng c lp mt danh sch cc t, cm t
thng dng. V d ch mi quan h thi gian, ngi ta dng, at the same time, subsequently,
initially, after that, etc; hay ch s tng phn, ngi ta dng, however, even so, in
constrast; ...

Nu bn l ngi khng thng xuyn vit paper, lc bt tay vo vit chc l s thy mnh vit t
qu, t vng nh bin i u ht, bao nhiu cm t hay ho hc c trong cc paper khc qun
note li nn gi mun xi kim cng kh.

Tnh c ti tm c trang web ny (http://www.comp.nus.edu.sg/~ligl/writing/ ) ca mt PhD

student NUS, ghi li cc sentences and patterns for writing rt hay. Cn nh hi trc trong lc
tho lun, Tt t cng c tng v vic pht trin mt thut ton rt ra nhng patterns kiu
ny. Ko bit gi cn hng th lm khng nh? :p

Ti paste li y nguyn vn chng ta tin tham kho

Sentences useful in writing

1. now more than before, the exponential increase of electronic data nowadays is available

2. see [..] and references therein

3. from the best of our knowledge

4. a collection of algorithms is currently known to solve this problem efficiently (possibly

optimally) on ...

5. Our algorithm is three to five times more space efficient than any reasonable ***

algorithms.
6. In the last decades IT has been investigated from different points of view and its basic

ideas enriched by new functionalities with the aim to cope with the features of the peculiar

setting of the use: ..., just to cite a few.

7. Large collections of full-text documents are now commonly used in automated information

retrieval.

8. One suggestion that may represent a step in the right direction is based on the use of ...,

rather than ....

9. The survey covers the problems in the following way.

For each problem and variations of it we review results for both the *** and *** version.

10. This will in most cases include a formal definition of the problem, a comparison of the

available results and a description of the techniques used to obtain the results.

11. More importantly, we will also pick one or more of the central algorithms for each of the

problems and present it in almost full detail.

12. Specifically, we will describe the algorithm, prove that it is correct and analyze its time

complexity.

13. For brevity, we will omit the proofs of a few lemmas and skip over some less important

details.

14. Common for the algorithms presented in detail is that, in most cases, they are the basis

for more advanced algorithms.

15. Typically, most of the algorithms for one of the above problems are refinements of the

same dynamic program.

16. *** rules could be written and exception lists used to disambiguate the difficult cases

described above. However, the lists will never be exhaustive, and multiple rules may

interact badly since....

17. while the rich literature provides valuable information about individual methods, clear

conclusions about cross-method comparison have been difficult because often the

published results are not directly comparable. Since the data used in different

publications are different, and no statistical significance analysis was conducted to verify

the impact of the difference in data on the performance variation of these methods.
18. In common with general scientific investigation, new ideas, concepts and interpretations

emerge quite spontaneously and these are then discussed, used, discarded or subsumed

into the prevailing subject paradigm. Sometimes these innovative concepts coalesce into

a new sub-discipline within the broad subject tapestry of ***.

19. We propose solutions for the first two problems, and a partial remedy for the third.

20. Now a new iteration can take place, and so on until the last of the centroid pairs has

settled down.

21. Supporting evidence has accumulated at a slower rate than one might have wished, but a

trickle sustained over 10 years forms a decent-sized pond.

22. Much like the decomposition of periodic signals into the sum of sinusoids, the probability

density of a data source can be estimated as the finite sum of standard distributions,

known as a finite mixture density.

23. Gaussian mixture models are a semi-parametric alternative to non-parametric histograms

(which can also be used to approximate densities) and it has greater flexibility and

precision in modeling the underlying distribution of sub-band coefficients.

24. Mixture models are a semi-parametric alternative to non-parametric histograms [Bishop:

Neural networks for pattern recognition] (which can also be used as densities) and

provide greater flexibility and precision in modeling the underlying statistics of sample

data.

25. Exhaustive enumeration is of academic interests only - in practice one would embed

different structure scoring methods in a greedy model search outer loop (Friedman, 1998)

to find probable structures.

26. Our methodology provides many advantages and some of them are as follows: 1) it is

based on a larger training set with diverse *** features and hence is expected to cover

greater *** space; 2) it is also based on the largest number of *** properties reported as of

date and hence can provide an optimal solution; 3) we have set the inter-correlation

coefficient between features as 0.9 and thereby the selected descriptors are expected to

be independent; 4) the utility of our models are demonstrated for in silico screening of

new chemical classes, an important criterion for any predictive model to have impact on
 drug discovery and development.

27. The problem is especially difficult since the outliers contain useful information that cannot

be discarded through filtering.

28. parsimonious means that it has the fewest parameters and greatest number of degrees of

freedom among all models that fit the data.

29. Four main problems in current data: curse of dimensionality, noise, incomplete, sparse

30. Characteristics of biological systems: non-linear, dynamic, with feedback, stochastic

31. Requirement for the model of a biological system: modularity, adaptive, dealing with

hidden effects, scalable

32. This list of applications is by no means complete and does not include ...

33. There is no substitute for computing chi-square by hand to understand the statistic.

34. ***1 theory does not provide a full answer to this problem since the choice of ***2 is not

part of the theory. It is, however, a step in the right direction.

35. because the various outcome values are mutually exclusive and collectively exhaustive.

36. Small scale experiments show that the proposed schemes are attractive.

37. comparison made by human evaluators between manually and system generated

timelines showed that although manually generated timelines are on average more

preferable, system generated timelines are sometimes judged to be better than manually

constructed ones.

38. This paper gives a personal view of the state of affairs in data analysis. That means that

inevitably I will be making general statements, so that most of you will be able to disagree

on some details. But I am trying to paint a broad picture, and I hope that you will agree

with the overall picture.

39. In this age of the Internet, opportunities to profit are as ubiquitous as your chances to

seize them. With more knowledge and power, literally at your fingertips, you can make

swifter and wiser decisions in your path towards achieving sterling portfolio performance.

But how do you get the right tools to sharpen your online edge? ==== The solution? Go

direct to the professionals.

40. The only exception of which I am aware is []: ....

 41. (and I am actually reproducing the quotation given by Suppes in [96])

42. This should by no means be viewed as a comprehensive survey. The topics covered

clearly reflect my own biases.

43. The primary disadvantage is increased computational complexity. However, with the

advent of (Markov chain Monte Carlo methods for Bayesian analysis and the widespread

availability of immense computing power), this problem is somewhat mitigated.

44. xxx is a matter that has attracted the attention of the scientific community during the last

decade. Among ....

45. for analysis purpose, one is not so much concerned with numerical or transmission

efficiency or representation compactness, but rather with accuracy and adaptive

properties of the analyzing tool.

46. Our results show that *** are the best performing algorithms over the experiments we

have run.

47. A variety of algorithms have previously been reported in the literature and their promising

performance has been evaluated empirically [***].

48. This was an entirely novel point of view at the time.

Patterns for writing

First sentence in a paper with the optional patterns for the second sentence

1. XXX has attracted the attention of the scientific community during the last decade.
(Among these ...
2. XXX is essential/important for YYY. (It does ....
3. One of the most fascinating and at the same time XXX is YYY. (The focus is YYY). YYY
is/does ....
4. XXX is ....
5. The use of XXX dates back to YYY and has been the focus of considerable activity in
recent years. (In particular, ...)
6. Advances in XXX along with the emergence of new and more YYY have stimulated a
resurgence in ZZZ. XXX is ...
7. Since the advent of the theory of XXX, developed by YYY, countless investigations with
these formalisms have been performed in a variety of fields.
8. XXX have been successfully used to do ....
 9. XXX is of great interest today because of ...
10. XXX is an active area of research.

First sentence in the literature review part of a paper

1. Previous studies have shown that ...

2. XXX can be accomplished by .... YYY is a relatively recent technique.
3. The earliest scientific reference to XXX is found in YYY who proposed that ....
4. Many researchers proposed various kinds of methods or algorithms to .... (In the
beginning, these methods are ...)
5. Several recent studies indicate that ....
6. Two different approaches have been used to do ..... The first one is/does .... The second
one is/does ...
7. Normally, ...
8. Generally / In general, ...
9. A review of the literature indicates that ....
10. A number of studies have been conducted using ....
11. A recent survey suggested that ...
12. Today, ...
13. One of the best known studies ...

New citations in literature review

1. XXX is a relatively recent technique. It is/does ....

2. Recently, ....

2.1 Recently, there has been a growing interest in XXX.

3. Rather less well-known are the XXX, which ....

4. With reference to the combinatorial exposion of ..., more / new .... methods have emerged
5. With maturation of XXX techniques, some researchers have applied XXX to YYY.

Scoping

1. Among them / these methods / the techniques /

2. In particular,
3. using adjectives

First sentence in the gap part of a paper

1. Despite XXX of this technique, ...

 2. However, these reported findings are ....

verbs in the literature review

proposed / hypothesized / reported / suggested / elaborated / claimed / explored / find / show /

L nh Duy

Duy-Dinh Le <ledduy@gmail.com> Tue, Dec 12, 2006 at 2:18 PM

To: cvpr-hcmuns-vn@googlegroups.com

@ The Task of the Referee

Trong bi ny, ti mun ni s qua mt cht v cng vic ca cc Reviewers. Hai bi ti tm thy
c rt hay l (xem trong file nh km):

1. How to write a review - Boi Faltings - EPFL, 2004

2. Task of the Referee, Alan Jay Smith , 1990.

C mt s im th v sau:

1. Paper checklist dng ngi gi kim tra xem mnh c th tha mn reviewer hay
khng?

Every paper must state:

the problem addressed

the solution or insight proposed
an example that shows how it works
an evaluation, ideally a comparison with existing techniques
Easy to check
Many papers fail this test!

Phn evaluation l phn rt quan trng cho hu ht cc paper, v a s paper b reject l do

nguyn nhn ny.

2. Ngh thut vit lch

Many authors are good salespeople:

hiding assumptions
using unrealistic examples
comparing with old or wrong versions of existing work
providing incorrect summaries of experimental results
This is where we need your intelligence!

Phn ny kh th v, ti s c minh ha trong phn Reading sau. Tuy nhin cp n vn

ny y thy nu reviewer non tay, s c mt s paper vn c accept.

L nh Duy
2 attachments
STAIRS-2004-talk.pdf
96K
taskoftheferee.pdf
88K