《计算机科学学术论文写作 (6).pdf》由会员分享，可在线阅读，更多相关《计算机科学学术论文写作 (6).pdf（6页珍藏版）》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、Theses67argument,then make it.Be thorough.A Ph.D.is an opportunity to do research indepth;shortcuts and incomplete experiments suggest shoddy work.For an extended research degree such as a Ph.D.,another difference between athesis and a paper is that the former may report on a series of more or less
2、indepen-dent research discoveries.In contrast,a typical paper concerns a single consistentinvestigation.A thesis may,moreover,include work drawn from multiple papers.For this reason,there is more variation in structure from thesis to thesis than frompaper to paper.An example of the problems faced in
3、 organizing a thesis is how toconsolidate descriptions of new algorithms.It may make sense to bring all of theminto a single chapter and then evaluate and compare them in subsequent chapters,orit may be preferable to describe them one by one,evaluating each in turn.Factors toconsider in choosing an
4、organization include how cohesive the algorithms are(forexample,whether they address the same problem)and whether an explanation ofone algorithm is meaningful if the previous one has not yet been evaluated.As the scope of a thesis is more substantial than that of a paper,the introductionshould be br
5、oad in topic and conversational in tone.It could introduce a whole arearather than a single problem.Another reason to develop a substantial introduction isthatathesisisamorethorough,detaileddocumentthanisapaper.Whywastheprob-lem worth investigating in depth?How do the parts of the investigation rela
6、te to eachother?What are some practical,concrete ways in which the outcomes of the workmightbeused?Runningexamplesmaybeoutlinedintheintroduction,togiveunitytothethesisoverall.Theroleofathesissintroductionis,however,muchthesameasinapaper.Asintheintroductionofapaper,theory,jargon,andnotationareinappro
7、priate.Take the time to learn about the challenges that are specific to thesis writing.3Browse other theses,from your own institution,other institutions,and other disci-plines.Form views about the strengths and weaknesses of these theses;these viewswill help to shape your own work.Critically,remembe
8、r that an examiner may onlyhave hours to read your workyou need to help them to spend that time well.Getting It WrongOver the next few chapters I look at the details of how to write well and also somecommon mistakes that researchers make;these largely concern details,that is,indi-vidual elements tha
9、t are poor.Some problems in papers,however,are at a higher level,and concern the qualityoftheworkasawhole.Asajournaleditor,conferencechair,andreferee,Iseedefectsof this kind again and againproblems that make it certain that the paper will berejected,and which in some cases are obvious to the referee
10、 in the first few momentsof reading.Common ways in which authors“get it wrong”are below.Many of these issuesare also discussed elsewhere in this book,but it is valuable,I think,to consider3There are plenty of good textbooks on this topic,and a couple by me.685Writing a Paperthem together.Something t
11、hat can be confronting about these issues is that,often,the authors appear to have worked hard over a long period of time to produce asubstantial document;and yet it immediately obvious that there is no chance ofthe work being accepted.An experienced researcher may feel baffled that this hasoccurred
12、as the work progressed,did no one see that it was going badly?Be alert to the potential faults in your own work,and have the courage to abandonor refocus activity that has little chance of leading to a valuable outcome.And,while the examples below are in some respects extremewhich makes them easy to
13、understandthey do highlight the kinds of issues that readers become alert to,andwhich authors should therefore avoid.IrrelevanceWhen I first see a paper,impressions form in a minute or two,influenced by lay-out,readability,and so on.With some papers,though,a positive initial response isgradually fol
14、lowed by a sinking feeling:I cannot figure out what this paper is about.Something elementary is utterly missing.What that“something”is can vary.Sometimes there is a lack of connection tothe literature on any particular topic,and thus no sense of what the author is tryingto achieve.In some cases the
15、author has proposed an elegant solution,but it is notobvious what the problem is,or the problem is so unrealistic that it is impossible tograsp.4Itmaybethattheauthorhasgivenaclearmotivationforthework,butthebulkofthepaperconcernssomethingelseentirely;anexamplewasapaperwhosestartingpoint was the chall
16、enge faced by teachers who wish to ensure that Web searchesonly return pages that are appropriate for children,but the contribution concernedmechanisms for selectively highlighting passages that were relevant to the query.Another form of this are those papers that are submitted to an inappropriateve
17、nue:5work on file compression submitted to a conference on database modelling,or work on face recognition submitted to a journal on data visualization.6Even more4I once struggled with a paper that concerned relational databases,but in which each recordandI do mean record,not tablehad an arbitrary nu
18、mber of fields.So,not relational then,but in someplaces relational properties were assumed.(Like many of the examples in this text,this instance is“real”but altered to disguise its origins,and also to make it easy to explain in a sentence or two.)AndanotherinwhichtheauthorsassumedthatWebqueriesareth
19、eresultofarandom walkthrougha weighted graph representing mental representations of related concepts,and wished to use a logof queries to infer the graph.Some rather arbitrary use of terminology(“actuation maps can bemade explicit through provocation by deliberative stimuli”)was at first intimidatin
20、g,until I realisedthe authors were using it to disguise the fact that they hadnt figured out how to achieve anything.5Which is not the same thing as venues that are inappropriate.A consequence of publicationpressureshasbeentheriseofjournalsandconferencesthatseemlittlemorethanopportunistic,withglossy
21、 web presences,plausible editorial boards or program committees,and even affiliations withmajor professional societiesbut with low standards of refereeing,high publication or registrationcosts,and,ultimately,no citations.6At a journal where most of the submissions were on Web search,I received respe
22、ctable paperson automated migration of software between operating systems and on a method for evaluating aGetting It Wrong69surprising are papers where the authors have utterly misunderstood the norms ofresearch or presentation for the field,such as papers where the authors have madeno use of standa
23、rd resources such as data sets,or,for example,a paper on searchtechnology written as a narrative from the imagined perspective of a document.Most curiously of all,in some papers there is no obvious research question,nostatement of aims or goals,and no claimed contribution.A more subtle problemof thi
24、s kind is when a paper appears to tell a coherent story,but on inspection itbecomes clear that,say,the experimental results are unrelated to the conclusions.Insome cases they seem to be on a different topic altogether.An example was a paperthat gave results for the efficiency of a string search meth
25、od but drew the conclusionthat the method enhanced data privacy.Stated so concisely,the paper sounds absurd!And yet such problems are not rare.Inconsistency,Inadequacy,and IncompletenessSome papers seem reasonable in parts,but the parts dont belong in the same docu-ment.A sensible,well-organised pap
26、er may be framed in terms of grandiose,ambi-tious claims that can only be described as ridiculous.7Or there may be a detailed,insightful literature review,but it is either disconnected from the contribution,or,bizarrely,thecontributionislessinterestingthanthepreviousworkthatwasdescribedso well.For p
27、apers that are overall at a high standard,perhaps the single commonestproblem that leads to rejection is that the experiments are inadequate.There may bean interesting method,but the experiments are trivial or uninformative,and fall farshort of supporting the claims;often,in these cases,the problem
28、is that the data setused is too artificial to allow any interesting conclusion to be drawn.Or a small dataset may be used to support claims for applications at an entirely different scale,suchas a set of a few thousand documents being used to make claims about Web search.Or the data set may not be r
29、elevant to the problem at all.It is as if the researchers(Footnote 6 continued)dictionary for medical practitioners,among others.And many that were not so respectable;topicsincluded image enhancement for ancient rock carvings(evaluated on a single image),use of XMLfor storing machine maintenance log
30、s(utterly trivial),automated translation of eighteenth-centuryEnglish text into modern usage(only arguably modern,but unarguably garbled and harder to read),and a tool for distinguishing between kinds of spider(use of a computer for a task does not meanthe task is computer science).7An example was a
31、 Ph.D.thesis that concerned how to develop software specifications in termsof a particular way of describing assertions and tests.The work was ambitious,but did appear toachieve reasonable initial outcomes.However,the motivation was that the work would ultimatelymake it unnecessary to write programs
32、,and that the specifications could be automatically inferredfrom transcripts of human conversation.(This condensation of several pages of rambling text intoa single sentence doesnt convey the full eccentricity of these claims.)No connection was madebetween the claims and the actual contribution.705W
33、riting a Paperare so excited about the ideas that they fail to see the need for validation,and offerresults that have no plausible relevance to the papers claims.Another variety of inadequacy is when parts of the paper are missing,or dealtwith in a few brief lines rather than pages.Strikingly,some p
34、apers have no literaturereview,or are based on a single out-of-date textbook,as if previous or recent workwas of no relevance.But if the author cannot be troubled to properly place the workin context of what is already known,a reader cannot learn what the contribution is.Another common failing is pa
35、pers where the reader cannot identify what the datais(there may be 200 documents,but where from?what content?what size?and soon),or who ran the experiments,or what techniques were tried.In the mind of thereader,moreover,there may not be much of a distinction between information thatis missing and in
36、formation that is concealeda line of thought that is unlikely tolead to belief that the work was done well.And some papers just arent ready to berefereed;the underlying work is unfinished and the paper is incomplete.IncomprehensibilityIn the cases above,the shortcomings are not always immediately ap
37、parent.In othercases,the papers problems are obvious straight away.For example,when presentedwith an incoherent abstract or introduction,the reader immediately feels that theworkcannotbeofvalue.Areadercanhavenohopeforapaperwhoseabstractbeginswith the sentence:“Internet supports all type of the forms
38、 of information that aredigital,suchasthepagesoftheWebeverywhereandalsolibrariesandemail,soitisalanguageofallourinformationsourcesinaworldrepositorythatisourknowledge.”8In such cases,there seems to be a wide gap between what the writer wants to sayand the actual words on the page.Ive observed simila
39、r writing in students who areperfectly clear in conversation,but in documents seem to want to pour all of theirthoughtsintoafewsentences.Butincomprehensibilitystemsfrommanycauses,andtakes many forms.Regardless of cause,if the result is a document that cannot beread,it wont be.UglinessThe look of a d
40、ocument is another respect in which problems can be immediatelyobvious.I suspect that many authors do not realise how much impact defects in8Orthisopeningsentence:“Wehavenewnetworksofwirelesslikewirednetworksthatourmethodmakes use of in computers connected to each other but heterogeneously and distr
41、ibuted.”Or this one:“With the explosioning of documents on the internet,systems of finding documentsthat are the answers of users their queries have become important in the recent years.”Or this:“An ideal vector space is the base of IR research,so the basic problem of IR is to set upa suitable vecto
42、r space,in this suitable vector space,query and document can be represented wellby vectors.”Getting It Wrong71presentation have on readers,but the message is clear:if something looks terrible,then the author doesnt care about the content;and if the author doesnt care,thenthe reader certainly shouldn
43、t.Thereareseveralcommonformsofthisugliness.Oneisinillustrationsandtables:graphs that are badly designed or badly rendered,tables that are irregular or chaotic,diagrams in which the parts are unrelated,and so on.Another form is inlayout,with,for example,absurdly sized headings or columns that overlap
44、.A third form is thepresence of dramatic formatting glitches,such as font and font size changing fromparagraph to paragraph.Each of these conveys an impression of laziness.Another kind of mistake conveys an impression of bad judgement:the decisionto use inappropriate styles of presentation.The comic
45、 sans font has been widelymocked for its use in slides;it is even more mockable in a paper.Other examplesincludeuseofcoloursinsteadofitalicsforemphasis,comicaldrawings,9andpeculiarover-the-top jokes.10A more subtle form of ugliness is when a paper is dense with errors.These maybe errors of fact,spel
46、ling errors,garbled citations,incomplete sentences,or any ofa range of such things.They show that the author is indifferent to the work,and thereader will respond likewise.IgnoranceAll of the issues noted in this section make it difficult to see a paper as being ofvalue,but,as a way of persuading th
47、e reader that a paper is worthless,nothing ismore certain than a display of ignorance.An example of this is when much of a paper is spent explaining an elementaryconcept that will be familiar to any likely reader and maybe even to undergraduates.Whileafewlinesofreviewmaybeappropriate(toensurethatter
48、minologyiscorrectlyunderstood,for example),why spend six pages of an algorithms paper explainingthe difference between random-access memory and hard disks?Moreover,when theauthor gets the details wrongand uses 1980s literature on memory technology ina 2000s paper,to consider one particular paperthe
49、main effect is to reveal that thework is unreliable.A similar example is when the author discusses at great length a statement that iseither blindingly obvious or,worse,clearly false.“Web pages from a single websitemaybemorelikeeachotherthanpagesdrawnfromdifferentlocations”,besidesbeing9Particularly
50、 memorable(not in a good way)was a submission that included a photograph ofBritains Queen Elizabeth II,on which the author had superimposed a cartoon of a smiling mouthand a thumbs-up,to illustrate the Queens happiness at the result of a successful Web search for“corgi”.There was no mention of royal