tag:blogger.com,1999:blog-77322264772536864682024-03-13T08:26:05.181-07:00Misc Research StuffAn online notebook for my jottings on NLP and machine learning.Unknownnoreply@blogger.comBlogger59125tag:blogger.com,1999:blog-7732226477253686468.post-68771439549343924182009-07-03T13:01:00.000-07:002009-07-04T13:37:15.564-07:00Dealing with large scale graphs<div>To a hammer everything looks like a nail but one great hammer to have in your toolbox is the graph. The <a href="http://www.aclweb.org/anthology/index.html">ACL anthology</a> alone lists more than 300 results for the query <a href="http://www.google.com/custom?q=%22graph+based%22&btnG=Search&hl=en&client=google-coop-np&cof=AH%3Aleft%3BCX%3AACL%2520Anthology%2520search%3BL%3Ahttp%3A%2F%2Fwww.google.com%2Fcoop%2Fintl%2Fen%2Fimages%2Fcustom_search_sm.gif%3BLH%3A65%3BLP%3A1%3BVLC%3A%23551a8b%3BGFNT%3A%23666666%3BDIV%3A%23cccccc%3B&cx=011664571474657673452%3A4w9swzkcxiy&adkw=AELymgVkmTTk4qTXJrDVPNTR6g4ViEj-nAg-Nqo7jvuyoligGcMvib0rqxKsTBfQt6QMeJ0oC2s2Qq0e-eV8IKdKmlbX_YDsfpSHlZuUWX9Baq88Tjxz24BaobmQZZo2_wTS3EFlrDDBAX9FfPCf-vKxUqOHxyN5yUZpAGfJtl8SdTQaJ0Kj02E&boostcse=0&sa=2">"graph based"</a>. Graphs based formalisms allow us to write down solutions in succinct linear algebra representation. However implementation of such solutions for large problems, or even for small datasets with blown-up graph representations can be challenging in limited resource environments. While some go for interesting <a href="http://snowbird.djvuzone.org/2007/abstracts/139.pdf">approximate solutions</a>, an alternative solution is to pool in several limited resource nodes into a map-reduce cluster and design a parallel algorithm to conquer scale with concurrency. This is easier said than done since designing some parallel algorithms requires a different perspective of the problem. This is well worth the effort as the new insights gained will reveal connections between things you already knew. For instance, in our <a href="http://www.textgraphs.org/ws09/index.html">TextGraphs 2009</a> paper we started out scaling up <a href="http://learning.eng.cam.ac.uk/zoubin/papers/zgl.pdf">Label Propagation</a> but eventually the connection to <a href="http://google.stanford.edu/~backrub/pageranksub.ps">PageRank</a> became obvious. To me this was a bigger learning moment than getting Label Propagation work for large graphs. [<a href="http://www.clsp.jhu.edu/~delip/nocrawl/textgraphs09.pdf">Preprint Copy</a>]</div><div><br /></div><div>For the actual implementation, we used <a href="http://hadoop.apache.org/">Hadoop</a> (surprise!) although <a href="http://googleresearch.blogspot.com/2009/06/large-scale-graph-computing-at-google.html">bulk synchronous parallel models</a> make more sense given the locality of the operations in most graph algorithms.</div>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-33464497985138385892009-03-31T15:04:00.000-07:002009-03-31T15:26:24.994-07:00Sentiment Analysis is AI-HardIn a <a href="http://mags.acm.org/communications/200904/?pg=16">breezy article</a> on sentiment analysis, Alex Wright quotes Bo Pang saying:<br /><blockquote>We are dealing with sentiment that can be expressed in subtle ways.</blockquote>This is so true with the examples I've encountered while working and my favorite is this one I saw on iTunes recently.<br /><br /><a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjH5PcS2INajpnQcv6P9bJnQWd7P6KgLfxUi9BvUAtASQeSa1CTK1hL_O790ElfwtEwsyWg6ljy6ClBMEue_XfjOn_7Bf5lbR9wBQ8zRsrszeQn6WFKXeo4UomflxYTNR8z0OTxcnlFqSI0/s1600-h/Picture+4.png"><img style="margin: 0px auto 10px; display: block; text-align: center; cursor: pointer; width: 400px; height: 61px;" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjH5PcS2INajpnQcv6P9bJnQWd7P6KgLfxUi9BvUAtASQeSa1CTK1hL_O790ElfwtEwsyWg6ljy6ClBMEue_XfjOn_7Bf5lbR9wBQ8zRsrszeQn6WFKXeo4UomflxYTNR8z0OTxcnlFqSI0/s400/Picture+4.png" alt="" id="BLOGGER_PHOTO_ID_5319478826930339346" border="0" /></a><br />While I commend Alex for writing an informative yet accessible article on the topic, I disagree with the article's opinion that sentiment analysis is a series of "filters". That is clearly an euphemism. Any working sentiment analysis system is actually an engineering feat often consisting of a series of hacks duct-taped by a glue handling special cases.<br /><br />The article also seems to suggest that extracting factual information is somehow easier than opinions. I invite them to participate <a href="http://apl.jhu.edu/%7Epaulmac/kbp.html">here</a>.Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-57938500175512770842009-02-28T09:02:00.000-08:002009-02-28T09:09:42.562-08:00On the way to Brewer's Art<div>Never mind how we got to this topic:</div><div><br /></div>me: Parsing is for fogies.<div>Markus: What?</div><div>Jason: I think he means crusty old linguists.</div><div>Markus: You should probably use a shallow parser.</div><div>me: I'm shallower than that; I use n-grams.</div>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-91639279983365918322008-12-19T11:45:00.000-08:002008-12-19T11:47:38.501-08:00EACL ReadingEACL 2009 <a href="http://www.eacl2009.gr/conference/acceptedpapers">accepted paper list</a> is up. Here's my reading list:<br /><br />WEAKLY SUPERVISED PART-OF-SPEECH TAGGING FOR RESOURCE-SCARCE LANGUAGES<br />Kazi Saidul Hasan and Vincent Ng<br /><br />USING CYCLES AND QUASI-CYCLES TO DISAMBIGUATE DICTIONARY GLOSSES<br />Roberto Navigli<br /><br />SYNTACTIC AND SEMANTIC KERNELS FOR SHORT TEXT PAIR CATEGORIZATION<br />Alessandro Moschitti<br /><br />SENTIMENT SUMMARIZATION: EVALUATING AND LEARNING USER PREFERENCES<br />Kevin Lerman, Sasha Blair-Goldensohn and Ryan McDonald<br /><br />PERSON IDENTIFICATION FROM TEXT AND SPEECH GENRE SAMPLES<br />Jade Goldstein-Stewart, Ransom Winder and Roberta Sabin<br /><br />OUTCLASSING WIKIPEDIA IN OPEN-DOMAIN INFORMATION EXTRACTION: WEAKLY-SUPERVISED ACQUISITION OF ATTRIBUTES OVER CONCEPTUAL HIERARCHIES<br />Marius Pasca<br /><br />GROWING FINELY-DISCRIMINATING TAXONOMIES FROM SEEDS OF VARYING QUALITY AND SIZE<br />Tony Veale, Guofu Li and Yanfen Hao<br /><br />GENERATING A NON-ENGLISH SUBJECTIVITY LEXICON: RELATIONS THAT MATTER<br />Valentin Jijkoun and Katja Hofmann<br /><br />CONTEXTUAL PHRASE-LEVEL POLARITY ANALYSIS USING LEXICAL AFFECT SCORING AND SYNTACTIC N-GRAMS<br />Apoorv Agarwal, Fadi Biadsy and Kathleen Mckeown<br /><br />COMPANY-ORIENTED EXTRACTIVE SUMMARIZATION OF FINANCIAL NEWS<br />Katja Filippova, Mihai Surdeanu, Massimiliano Ciaramita and Hugo Zaragoza<br /><br />ANALYSING WIKIPEDIA AND GOLD-STANDARD CORPORA FOR NER TRAINING<br />Joel Nothman, Tara Murphy and James R. CurranUnknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-4330628867690146572008-12-02T16:08:00.000-08:002008-12-05T20:19:12.083-08:00And we're back ...Sometime back I wrote about <a href="http://resnotebook.blogspot.com/2008/07/quick-scan-at-acl.html">Wordle</a> to visualize textual information using frequency counts. Change.gov, Obama's transition team website <a href="http://change.gov/newsroom/entry/join_the_discussion_daschles_healthcare_response/">uses it</a> on the comments in response to their health care system. This is very interesting but I think Wordle should display top 100 collocations instead of top 100 words. But oh, we also learnt at last ACL how to <a href="http://www.blogger.com/www.aclweb.org/anthology-new/P/P08/P08-1075.pdf">learn collocation information from unigram frequencies</a>.Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-47325914264867682482008-07-17T05:10:00.000-07:002008-07-17T05:28:42.926-07:00Too many cooks?Computational Linguistics is becoming like the <a href="http://www.sciencemag.org/current.dtl">Science</a> or <a href="http://www.nature.com/nature/index.html">Nature</a>. For instance, see <a href="http://www.mitpressjournals.org/doi/abs/10.1162/coli.2008.07-055-R2-06-29">this paper</a> in the current issue: (In this case, the broth wasn't spoiled ;-)<br /><br /><a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhp6AJMW8_ERpPkTIWdUnSwjocKnJK3sAlW6c1iFjXEwX7bE33yljuT91EVhB9lb91k-jLj2JyH8qcdQ-Hcep08SwtRyPtwyBNyKzrzS0UDyDCKqCd-pTwo8x6ZEsXlQQ9SNWV-RHifkbPP/s1600-h/Picture+2.png"><img style="margin: 0px auto 10px; display: block; text-align: center; cursor: pointer;" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhp6AJMW8_ERpPkTIWdUnSwjocKnJK3sAlW6c1iFjXEwX7bE33yljuT91EVhB9lb91k-jLj2JyH8qcdQ-Hcep08SwtRyPtwyBNyKzrzS0UDyDCKqCd-pTwo8x6ZEsXlQQ9SNWV-RHifkbPP/s400/Picture+2.png" alt="" id="BLOGGER_PHOTO_ID_5223954801996340482" border="0" /></a>Guess which paper has the largest number of authors on the <a href="http://aclweb.org/anthology-new/">ACL anthology</a>?Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-15396811357609484292008-07-08T10:55:00.000-07:002008-07-08T11:34:33.508-07:00To theory or not to theoryI stumbled upon this paper "Reflections after Refereeing Papers for NIPS" by <a href="http://www.stat.berkeley.edu/%7Ebreiman/">Leo Breiman</a> that gives some really candid insights into theory papers. (Unfortunately, I could not find a soft copy to share, except <a href="http://direct.bl.uk/bld/PlaceOrder.do?UIN=026632341&ETOC=EN&from=searchengine">this link</a>.) Some noteworthy observations:<br /><blockquote></blockquote><blockquote>"No theorems" implies "No theory"<br /><br />"... more than 99% of the published papers are useless exercises."<br /><br />"Mathematical theory is not critical to development of machine learning."<br /><br />"Our fields would be better off with far fewer theorems, less emphasis on faddish stuff, and much more into scientific inquiry and engineering."<br /></blockquote><br />I really liked this article, especially coming from someone who has been working in theory all his life but I would still prefer reading papers giving theoretical insight, however useless, than pages and pages of feature engineering & experimentation using classifier X on problem Y -- the current trend at ACL.Unknownnoreply@blogger.com1tag:blogger.com,1999:blog-7732226477253686468.post-12649756335123690982008-07-07T06:49:00.000-07:002008-07-07T06:58:00.163-07:00A quick scan at ACL<a href="http://mendicantbug.com/">Mendicant Bug</a> informs about a new tag-cloud service called <a href="http://wordle.net/">Wordle</a>. Here is a look at this year's ACL. Gives a clear idea of what is going on! A larger image is available <a href="http://wordle.net/gallery/wrdl/55900/ACL2008">here</a>.<br /><br /><a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEitj9EeBl-rJA4PyTYuZuspOB0mAGxFcDM5W0-lCn-7ijlw3jV2VkTLAG94knRiGqlo4bHj9-rTU3jUHkIgyUPFbt7LxrMlC9I9moWg_EeZfVT3paHFD1XKXTcdL6k1l1UXTctHL7jvqMnx/s1600-h/Picture+1.png"><img style="margin: 0px auto 10px; display: block; text-align: center; cursor: pointer;" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEitj9EeBl-rJA4PyTYuZuspOB0mAGxFcDM5W0-lCn-7ijlw3jV2VkTLAG94knRiGqlo4bHj9-rTU3jUHkIgyUPFbt7LxrMlC9I9moWg_EeZfVT3paHFD1XKXTcdL6k1l1UXTctHL7jvqMnx/s400/Picture+1.png" alt="" id="BLOGGER_PHOTO_ID_5220269682864239874" border="0" /></a>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-30294586171448516752008-05-11T22:43:00.000-07:002008-05-11T22:51:39.251-07:00Powerset Natural Language Search<a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjl5n67cOOrzweKimjrQLwqqJtUeI7j98KrwNygcFfZ3E25OmgRg6bYmx8in5gTpX3-QeZyu3Th-_dbRzpos2pvFmH6UMUlQOraXDNI5Cat9iTpKIK6lTzpiheC9_FiXdR2Hj2-Kknbllcd/s1600-h/Picture+1.png"><img style="margin: 0px auto 10px; display: block; text-align: center; cursor: pointer;" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjl5n67cOOrzweKimjrQLwqqJtUeI7j98KrwNygcFfZ3E25OmgRg6bYmx8in5gTpX3-QeZyu3Th-_dbRzpos2pvFmH6UMUlQOraXDNI5Cat9iTpKIK6lTzpiheC9_FiXdR2Hj2-Kknbllcd/s400/Picture+1.png" alt="" id="BLOGGER_PHOTO_ID_5199364090099267794" border="0" /></a><br /><a href="http://www.powerset.com">Powerset</a>, a company we only remember seeing as conference sponsors, now actually has something working. After receiving an email from them, I tried out several queries. At best, it seems to answer most Wh-questions and certain whole-part relations.<br /><br />Try out the <a href="http://www.google.com/search?q=Who+is+Bart+Simpson%27s+father%3F">same query on Google</a>.Unknownnoreply@blogger.com1tag:blogger.com,1999:blog-7732226477253686468.post-84278480248921354272008-04-03T20:35:00.000-07:002008-04-03T20:38:09.460-07:00Writing styleThe sweetest thing ever written in a paper: "The reader who is unfamiliar with this field or who has allowed his or her facility with some of its concepts to fall into disrepair may profit from a brief perusal of Feller (1950) and Gallagher (1968)."<br /><br /> - Brown et. al., "Class based n-gram Models of Natural Language.", Computational Linguistics, 1990Unknownnoreply@blogger.com1tag:blogger.com,1999:blog-7732226477253686468.post-18820412657734511652008-03-28T11:54:00.000-07:002008-03-28T12:05:14.441-07:00Searching ACL anthologyIf you are looking up the <a href="http://acl.ldc.upenn.edu/">ACL anthology</a> regularly, my friend <a href="http://www.clsp.jhu.edu/%7Emarkus/">Markus</a> has a nice firefox search plugin to do that. You can get that and others from <a href="http://mycroft.mozdev.org/download.html?category=14&country=WW&language=all">this page</a>.<br /><br /><a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEitwfITu5NMCnNh7N9xZnwjCrc1xdtawib4wOhKHSzZyTutmidahf4j8fBQ0ammlE-KjXUkl2-658tjRLTu4Dpi5269kNbaAP643KBoIeR3luBcXs930BZzbgFqTsho30-GRVt1tXrq8VWA/s1600-h/Picture+1.png"><img style="margin: 0px auto 10px; display: block; text-align: center; cursor: pointer;" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEitwfITu5NMCnNh7N9xZnwjCrc1xdtawib4wOhKHSzZyTutmidahf4j8fBQ0ammlE-KjXUkl2-658tjRLTu4Dpi5269kNbaAP643KBoIeR3luBcXs930BZzbgFqTsho30-GRVt1tXrq8VWA/s400/Picture+1.png" alt="" id="BLOGGER_PHOTO_ID_5182870229223618370" border="0" /></a>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-28488634928152455772008-03-27T21:26:00.000-07:002008-03-27T22:20:52.966-07:00ACL accepted papers<span style="font-size:100%;"><span style="font-family:georgia;">Hal posted a while back about the </span><a style="font-family: georgia;" href="http://nlpers.blogspot.com/2008/03/acl-papers-up.html">ACL accepted papers</a><span style="font-family:georgia;"> that I just read now -- I've been living under a rock for some time. You can get a printer friendly version </span><a style="font-family: georgia;" href="http://cs.jhu.edu/%7Edelip/misc/acl08.html">here</a><span style="font-family:georgia;">. I know, my paper did not make it to that list :(</span><br /><br /><span style="font-family:georgia;">New additions to my reading list:</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Distributional Identification of Non-Referential Pronouns</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Shane Bergsma, Dekang Lin and Randy Goebel</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >An Unsupervised Approach to Biography Production using Wikipedia</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Fadi Biadsy, Julia Hirschberg and Elena Filatova</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Resolving Personal Names in Email Using Context Expansion</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Tamer Elsayed, Douglas Oard and Galileo Namata</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Mining Wiki Resources for Multilingual Named Entity Recognition</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Alexander Richman and Patrick Schone</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Inducing Gazetteers for Named Entity Recognition by Large-scale Clustering of Dependency Relations</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Jun'ichi Kazama and Kentaro Torisawa</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Name Translation in Statistical Machine Translation - Learning When to Transliterate</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Ulf Hermjakob, Kevin Knight and Hal Daume</span><br /><br /><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >The Tradeoffs Between Open and Traditional Relation Extraction</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Michele Banko and Oren Etzioni</span><br /><br /><span style="font-family:georgia;">(Longest paper title)</span><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Unsupervised Discovery of Generic Relationships Using Pattern Clusters and its Evaluation by Automatically Generated SAT Analogy Questions</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Dmitry Davidov and Ari Rappoport</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Finding Contradictions in Text</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Marie-Catherine de Marneffe, Anna Rafferty and Christopher Manning</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Extracting Question-Context-Answer Triples from Online Forums</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Shilin Ding, Gao Cong, Chin-Yew Lin and Xiaoyan Zhu</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >EM Can Find Pretty Good HMM POS-Taggers (When Given a Good Start)</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Yoav Goldberg, Meni Adler and Michael Elhadad</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Extraction of Entailed Semantic Relations Through Syntax-based Comma Resolution</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Vivek Srikumar, Roi Reichart, Mark Sammons, Ari Rappoport and Dan Roth</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Learning Bigrams from Unigrams</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Xiaojin Zhu, Andrew Goldberg, Michael Rabbat and Robert Nowak</span><br /><br /></span><span style="font-style: italic;font-family:georgia;font-size:100%;" >Evaluating Roget's Thesauri</span><span style="font-size:100%;"><br /><span style="font-family:georgia;">Alistair Kennedy and Stan Szpakowicz</span><br /><br /><span style="font-style: italic;font-family:georgia;" >Randomized Language Models via Perfect Hash Functions</span><br /><span style="font-family:georgia;">David Talbot and Thorsten Brants</span><br /><br /><span style="font-style: italic;font-family:georgia;" >Solving Relational Similarity Problems Using the Web as a Corpus</span><br /><span style="font-family:georgia;">Preslav Nakov and Marti Hearst</span><br /></span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-49589678048146375322008-02-24T17:50:00.001-08:002008-02-24T17:50:28.820-08:00What do you do?<span style="font-family: trebuchet ms;font-family:georgia;" >As a grad student working on NLP how do you explain what you are working on, to friends and family? I inevitably end up referring to the Google search engine even though what I do is quite far from IR. Actually, thats not true. These days IR seems to consume everything but thats another story.</span><br /><br /><span style="font-family: trebuchet ms;font-family:georgia;" >This reminds me of a funny conversation at CLSP recently:</span><br /><br /><span style="font-family: trebuchet ms;font-family:georgia;" >Sanjeev is telling us about an incident where a concerned parent of a young child with a speaking disability is asking him for his opinion. Apparently, she is confused about "Language and Speech Processing" in CLSP.</span><br /><br /><span style="font-family: trebuchet ms;font-family:georgia;" >Keith butts in: "Run a few more iterations of EM and he'll be fine."</span>Unknownnoreply@blogger.com1tag:blogger.com,1999:blog-7732226477253686468.post-25118047547079075212008-02-14T01:23:00.000-08:002008-02-14T01:59:11.602-08:00A song on parsingWe all know <a href="http://cs.jhu.edu/%7Ejason/">Jason</a>'s love for parsing from <a href="http://cs.jhu.edu/%7Ejason/research.html">his work</a> but it takes a different level of dedication to write a Valentine's Day <a href="http://cs.jhu.edu/%7Ejason/fun/grammar-and-the-sentence/">song about parsing</a>.<br /><br />As Jason says, "Parsers just want to be appreciated, like everyone else."Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-64745632603460813232007-10-17T17:07:00.000-07:002007-10-17T17:12:39.058-07:00Funny bone<span style="font-family:trebuchet ms;">The frequentist exclaimed, "All your Bayes are belong to us!" to which the Bayesian responded, "Well, it depends."</span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-34974172762781140802007-09-20T22:58:00.000-07:002007-09-20T23:09:21.150-07:00NIPS papers are out<span style="font-family: verdana;">For a full list see </span><a style="font-family: verdana;" href="http://nips07.stanford.edu/accepted_papers.html">here</a><span style="font-family: verdana;">. Some papers I want to read based on my current interests:</span><br /><br /><span style="font-family: verdana;">Random Projections for Manifold Learning</span><br /><span style="font-family: verdana;">Chinmay Hegde, Michael Wakin, Richard Baraniuk</span><br /><br /><span style="font-family: verdana;">The Distribution Family of Similarity Distances</span><br /><span style="font-family: verdana;">Gertjan Burghouts, Arnold Smeulders, Jan-Mark Geusebroek</span><br /><br /><span style="font-family: verdana;">Manifold Sculpting</span><br /><span style="font-family: verdana;">Michael Gashler, Dan Ventura, Tony Martinez</span><br /><br /><span style="font-family: verdana;">A learning framework for nearest neighbor search</span><br /><span style="font-family: verdana;">Lawrence Cayton, Sanjoy Dasgupta</span><br /><br /><span style="font-family: verdana;">Learning Bounds for Domain Adaptation</span><br /><span style="font-family: verdana;">John Blitzer, Koby Crammer, Alex Kulesza, Fernando Pereira, Jennifer Wortman</span><br /><br /><span style="font-family: verdana;">Convex Relaxations of EM</span><br /><span style="font-family: verdana;">Yuhong Guo, Dale Schuurmans</span><br /><br /><span style="font-family: verdana;">A Randomized Algorithm for Large Scale Support Vector Learning</span><br /><span style="font-family: verdana;">Krishnan Kumar, Chiru Bhattacharya, Ramesh Hariharan</span><br /><br /><span style="font-family: verdana;">Bundle Methods for Machine Learning</span><br /><span style="font-family: verdana;">Alex Smola, S V N Vishwanathan, Quoc Le</span><br /><br /><span style="font-family: verdana;">Regularized Boost for Semi-Supervised Learning</span><br /><span style="font-family: verdana;">Ke Chen, Shihai Wang</span><br /><br /><span style="font-family: verdana;">Learning the structure of manifolds using random projections</span><br /><span style="font-family: verdana;">Yoav Freund, Sanjoy Dasgupta, Mayank Kabra, Nakul Verma</span><br /><br /><span style="font-family: verdana;">A complexity measure for intuitive theories</span><br /><span style="font-family: verdana;">Charles Kemp, Noah Goodman, Joshua Tenenbaum</span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-36972594875799718022007-08-18T00:40:00.000-07:002007-08-18T00:54:25.570-07:00NLP and Global Warming<span style="font-family:verdana;">Those of us who were at EMNLP-CONLL 2007 remember the "NLP and Global Warming" exchange between James Clarke, Jason Eisner, and Dan Bikel at the Q/A session of the </span><a style="font-family: verdana;" href="http://acl.ldc.upenn.edu/D/D07/D07-1001.pdf">Clarke and Lapata paper</a><span style="font-family:verdana;">. The transcript of this funny conversation is now </span><a style="font-family: verdana;" href="http://www.cs.jhu.edu/%7Ejason/advice/conf/NLP-and-global-warming.html">online</a><span style="font-family:verdana;">, thanks to Jason.</span><br /><br /><span style="font-family:verdana;">I really liked Hal's ending remark.</span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-65873532280439025042007-08-15T16:16:00.001-07:002007-08-18T01:03:14.191-07:00People Search on the Web<span style="font-family:verdana;">Wired has an </span><a style="font-family: verdana;" href="http://www.wired.com/techbiz/startups/news/2007/08/spock_reputation">article</a><span style="font-family:verdana;"> about </span><a style="font-family: verdana;" href="http://www.spock.com/">spock.com</a><span style="font-family:verdana;">, a people search engine that combines crawled and user added content. From the few searches I did, looks like this is good for celebrity names than a regular person with web content. For instance, searching a name like "David Smith" produces these <a href="http://www.spock.com/q/David-Smith">results</a>. Of the top 10 results, only 3 of them actually have the name "David Smith" or something closer and the first result is not one of them. Compare this with a general purpose search engine like </span><a style="font-family: verdana;" href="http://www.google.com/search?source=ig&hl=en&q=David+Smith">Google</a><span style="font-family:verdana;">. Among a dozen random NLP/ML academic names (professors) I tried, it only got Jason Eisner and Tom Mitchell correct. One reason for this poor recall is probably they don't get content from user home pages.</span><br /><span style="font-family:verdana;">(Some sites where this data is derived from include MySpace, Friendster, IMDB, Wikipedia, ratemyprofessors.com, etc.)</span><br /><br /><span style="font-family:verdana;">Nevertheless, this website is a representative of interesting KDD-style problems that one could do with people names. It is also interesting as people names that we look for fall in the "long tail" without sufficient data to support calling for clever machine learning techniques.</span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-22681725071322342452007-08-12T12:30:00.000-07:002007-08-12T13:33:40.681-07:00Digital Reasoning awarded contextual similarity patent?<span style="font-family:verdana;">I was lead to </span><a style="font-family: verdana;" href="http://www.forbes.com/businesswire/feeds/businesswire/2007/07/31/businesswire20070731005886r1.html">this article</a><span style="font-family:verdana;"> on Forbes via </span><a style="font-family: verdana;" href="http://www.inma.ucl.ac.be/%7Efrancois/blog/entries/entry_594.php">Damien's post</a><span style="font-family:verdana;">. The article is about a company Digital Reasoning getting patent on what sounded to me as contextual similarity. Their "white paper" makes reference to a </span><a style="font-family: verdana;" href="http://tinyurl.com/2h6nz4">patent number 7249117</a><span style="font-family:verdana;"> (via USPTO). Unlike research papers, reading the patent document was so difficult. Will get to it sometime later but here is an extract from their press release about what their technology can do.</span><br /><br /><blockquote>* Learn the meanings of words, classes of words, and other symbols based on how they are used in context in natural language<br />* Create and manipulate models of this "meaning" - i.e. the mathematical patterns of usage - including the detection of groups or similar categories of words or development of hierarchies or creation of relationships between words<br />* Improve the models based on human feedback or using other structured information after model construction <br />* The representation or sharing of this model or learning in an ontology, graph structure, or programming languages</blockquote><br /><br /><span style="font-family:verdana;">Anyone from the ACL/ML/AI community can immediately recognize this and start citing their favorite papers on these topics starting from at least a decade ago. A promotional video from the company on YouTube can be found </span><a style="font-family: verdana;" href="http://www.youtube.com/watch?v=R5ihr4kx3dQ">here</a><span style="font-family:verdana;">. Excerpt from the video: "... We treat the text representation of human language as a signal ... ". </span><br /><br /><span style="font-family:verdana;">I think everyone should stop taking patents seriously. Wishful thinking?</span>Unknownnoreply@blogger.com1tag:blogger.com,1999:blog-7732226477253686468.post-587666097409617462007-08-02T19:40:00.000-07:002007-08-02T21:25:18.661-07:00Recommending scientific papers<div style="font-family: verdana;">I noticed a new feature in Citeseer which tries suggest an "alternate document" for a paper.<br /><div><img id="BLOGGER_PHOTO_ID_5094312320294261490" style="margin: 0px auto 10px; display: block; text-align: center; width: 463px; height: 94px;" alt="" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh_5QM7WPCmAo-7zKV1omxVaPkvyPolTemm4729Wv-8lQMVBI9dPyRDc_Jqq3wX1rY_GcurcgFc1bD3IOxOizkSq2I9-lzcYjtVJqIxSmDGUkdUdb6OWfjSLElCFdh6INfxRS5Ca6SncBqO/s400/recommendation.jpg" border="0" />Clearly it does not do what it implies to do and it doesn't show up for all papers. (Experimental?) So, an interesting question is how does one recommend scientific papers? Something more than mere document similarity is required. If I am reading a CRF paper then there is no point in listing all papers containing similar words. Just listing nodes connected to inward and outward links of the paper in the citation graph wont suffice either. Ideal recommendations for a paper would depend on the role the user is playing. When I am reading a paper about some new topic, I would like to get pointed to original papers on the topic, some recent papers on the topic, and may be some survey papers or books. On the other hand when I am writing a paper, I would like to be pointed to all papers related to the topic (recall important than precision here to avoid reviewer comments on "missing reference") in some magical order that puts papers more relevant to your work above. Also these papers might not be related in directly through citations. If there is a recent related work in the Annals of Statistics, for instance, then it should show up when I am working on, say, approximate inference methods for graphical models. (Possible to deduce this from my previous queries?)<br /><br />In spite of more information being present in a scientific paper than its text, recommending or ranking papers appears to be quite challenging.<br /><br /></div></div>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-36854876136378991152007-07-26T22:14:00.000-07:002007-07-26T23:31:05.068-07:00Google allows data binging for researchersGoogle now opened access to university researchers to its search and MT systems in <a href="http://googleresearch.blogspot.com/2007/07/drink-from-firehose-with-university.html">today's announcement</a> on their <a href="http://googleresearch.blogspot.com/">research blog</a>. The search API documentation does not mention any restriction on the number of queries that can be posted for search (The earlier limit was 1000). Whatever the number is I am guessing it will be large (<span style="font-style: italic;">Drinking from the firehose?</span>). However, the MT API allows 1000 queries per day with the documentation hinting that this need not be a hard limit.<br /><br />Looking at the search API output, two things I really miss is the number of hits and the snippet for each search result. The number of hits has been used in several papers for <a href="http://portal.acm.org/citation.cfm?id=1073153">interesting</a> <a href="http://www.cwi.nl/%7Epaulv/papers/amdug.pdf">results</a>. The other useful feature is snippets. Every search result from Google is accompanied by a small snippet extracted from the page, as shown below for an example query "Dekang Lin".<br /><br /><a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhRqzPY64M46DrD_bGAUjtxQCLmHLU-QuEsSJHRLDedEOubRlEIbaRBoP-ptX-CrhspNpQO0ljp3KJBxo47UHWHCkLskacNKMxPx7UcaTGRvLRgkBMAXmzKVl9wriceRRIvhF3siGecU438/s1600-h/snippet.jpg"><img style="margin: 0px auto 10px; display: block; text-align: center; cursor: pointer;" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhRqzPY64M46DrD_bGAUjtxQCLmHLU-QuEsSJHRLDedEOubRlEIbaRBoP-ptX-CrhspNpQO0ljp3KJBxo47UHWHCkLskacNKMxPx7UcaTGRvLRgkBMAXmzKVl9wriceRRIvhF3siGecU438/s400/snippet.jpg" alt="" id="BLOGGER_PHOTO_ID_5091749865496056530" border="0" /></a>The information in the snippets can be used as informative features in different tasks like this one in <a href="http://www.cs.jhu.edu/%7Engarera/publications/snippetsSEMEVAL07.pdf">person name disambiguation</a>. (BTW, Dekang is now at Google)<br /><br />Despite these minor quibbles, these new APIs will be quite useful to all of us and will certainly result in more papers on <a href="http://portal.acm.org/citation.cfm?id=1245144">Googleology</a>.<br /><br /><span style="font-style: italic;">Later addition:</span> Turns out we can sort of get the counts by simply counting the number of search results by repeatedly executing the request (only ten results per request) but the API caps this limit to 100. That means you could get a maximum of 1000 results. Which is not quite the same as "<span style="">Results <b>1</b> - <b>10</b> of about <b>779,000,000</b>". </span>Though that number is approximate, it is still indicative of how strong the query is w.r.t the web. For example GoogleCount("Horse+animal") >> GoogleCount("Horse+truck").Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-29368467247481523572007-07-24T09:53:00.000-07:002007-07-24T09:58:26.666-07:00Readings from SIGIR 2007<span style=";font-family:trebuchet ms;font-size:100%;" ><a href="http://www.sigir2007.org/">SIGIR 2007</a> is happening now at Amsterdam!<br /><br />Latent Concept Expansion Using Markov Random Fields, Donald Metzler, Bruce Croft<br /><br />Random Walks on the Click Graph, Nick Craswell, Martin Szummer<br /><br />Towards Automatic Extraction of Event and Place Semantics from Flickr Tags, Tye Rattenbury, Nathaniel Good, Mor Naaman<br /><br />Clustering of Documents with Local and Global Regularization, Fei Wang, Changshui Zhang, Tao Li<br /><br />Detecting, Categorizing and Clustering Entity Mentions in Chinese Text, Wenjie Li, Donglei Qian, Chunfa Yuan, Qin Lu<br /><br />Principles of Hash-based Text Retrieval, Benno Stein<br /><br />DiffusionRank: A Possible Penicillin for Web Spamming, Haixuan Yang, Irwin King, Michael R. Lyu<br /><br />Context Sensitive Stemming for Web Search, Fuchun Peng, Nawaaz Ahmed, Xin Li, Yumao Lu<br /><br />Combining Content and Link for Classification using Matrix Factorization, Shenghuo Zhu, Kai Yu, Yun Chi, Yihong Gong<br /><br />ARSA: A Sentiment-Aware Model for Predicting Sales Performance Using Blogs, Yang Liu, Jimmy Huang, Aijun An, Xiaohui Yu<br /><br />Heavy-Tailed Distributions and Multi-Keyword Queries, Arnd Konig, Surajit Chaudhuri, Liying Sui, Kenneth Church</span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-72726243983972133952007-07-23T15:56:00.000-07:002007-07-23T22:17:27.903-07:00Readings from AAAI 2007<a style="font-family: trebuchet ms;" href="http://www.aaai.org/Conferences/AAAI/2007/aaai07program.php">AAAI 2007</a><span style="font-family:trebuchet ms;"> is now going on at </span>Vancouver<span style="font-family:trebuchet ms;">. Here is my selection of NLP and Learning papers I would like to know more about.</span><br /><br /><span style="font-size:100%;"><span style="font-family:trebuchet ms;">Deriving a Large-Scale Taxonomy from Wikipedia, </span><span style="font-family:trebuchet ms;">Simone Paolo Ponzetto, Michael Strube</span><br /><br /><span style="font-family:trebuchet ms;">Relation Extraction from Wikipedia Using Subtree Mining, </span><span style="font-family:trebuchet ms;">Dat P. T. Nguyen, Yutaka Matsuo, Mitsuru Ishizuka</span><br /><br /><span style="font-family:trebuchet ms;">Finding Related Pages Using Green Measures: An Illustration with Wikipedia, </span><span style="font-family:trebuchet ms;">Yann Ollivier, Pierre Senellart</span><br /><br /><span style="font-family:trebuchet ms;">Graph Partitioning Based on Link Distributions, </span><span style="font-family:trebuchet ms;">Bo Long, Mark (Zhongfei) Zhang, Philip S. Yu</span><br /><br /><span style="font-family:trebuchet ms;">Semi-supervised Learning by Mixed Label Propagation, </span><span style="font-family:trebuchet ms;">Wei Tong, Rong Jin</span><br /><br /><span style="font-family:trebuchet ms;">Semi-Supervised Learning with Very Few Labeled Training Examples, </span><span style="font-family:trebuchet ms;">Zhi-Hua Zhou, De-Chuan Zhan, Qiang Yang</span><br /><br /><span style="font-family:trebuchet ms;">Clustering with Local and Global Regularization, </span><span style="font-family:trebuchet ms;">Fei Wang, Changshui Zhang, Tao Li</span><br /><br /><span style="font-family:trebuchet ms;">Isometric Projection, </span><span style="font-family:trebuchet ms;">Deng Cai, Xiaofei He, Jiawei Han</span><br /><br /><span style="font-family:trebuchet ms;">Improving Similarity Measures for Short Segments of Text, </span><span style="font-family:trebuchet ms;">Wen-tau Yih, Christopher Meek</span><br /><br /><span style="font-family:trebuchet ms;">Topic Segmentation Algorithms for Text Summarization and Passage Retrieval: An Exhaustive Evaluation, </span><span style="font-family:trebuchet ms;">Gaël Dias, Elsa Alves. José Gabriel Pereira Lopes</span><br /><br /><span style="font-family:trebuchet ms;">Robust Estimation of Google Counts for Social Network Extraction, </span><span style="font-family:trebuchet ms;">Yutaka Matsuo, Hironori Tomobe, Takuichi Nishimura</span><br /><br /><span style="font-family:trebuchet ms;">Harvesting Relations from the Web - Quantifiying the Impact of Filtering Functions, </span><span style="font-family:trebuchet ms;">Sebastian Blohm, Philipp Cimiano, Egon Stemle</span><br /><br /><span style="font-family:trebuchet ms;">Template-Independent News Extraction Based on Visual Consistency, </span><span style="font-family:trebuchet ms;">Shuyi Zheng, Ruihua Song, Ji-Rong Wen</span><br /><br /><span style="font-family:trebuchet ms;">Comprehending and Generating Apt Metaphors: A Web-driven, Case-based Approach to Figurative Language, </span><span style="font-family:trebuchet ms;">Tony Veale, Yanfen Hao</span><br /><br /><span style="font-family:trebuchet ms;">Mobile Service for Reputation Extraction from Weblogs - Public Experiment and Evaluation, </span><span style="font-family:trebuchet ms;">Takahiro Kawamura, Shinichi Nagano, Masumi Inaba, Yumiko Mizoguchi</span><br /><br /><span style="font-family:trebuchet ms;">The Impact of Time on the Accuracy of Sentiment Classifiers Created from a Web Log Corpus, </span><span style="font-family:trebuchet ms;">Kathleen T. Durant, Michael D. Smith</span><br /><br /><span style="font-family:trebuchet ms;">Nectar: Learning by Combining Observations and User Edits, </span><span style="font-family:trebuchet ms;">Vittorio Castelli, Lawrence Bergman, Daniel Oblinger</span><br /><br /><span style="font-family:trebuchet ms;">Multi-Label Learning by Instance Differentiation, </span><span style="font-family:trebuchet ms;">Min-Ling Zhang, Zhi-Hua Zhou</span><br /><br /><span style="font-family:trebuchet ms;">Extracting Influential Nodes for Information Diffusion on a Social Network, </span><span style="font-family:trebuchet ms;">Masahiro Kimura, Kazumi Saito, Ryohei Nakano</span><br /><br /><span style="font-family:trebuchet ms;">Temporal and Information Flow Based Event Detection from Social Text Streams, </span><span style="font-family:trebuchet ms;">Qiankun Zhao, Prasenjit Mitra, Bi Chen</span><br /><br /><span style="font-family:trebuchet ms;">Analyzing Reading Behavior by Blog Mining, </span><span style="font-family:trebuchet ms;">Tadanobu Furukawa, Mitsuru Ishizuka, Yutaka Matsuo, Ikki Ohmukai, Koki Uchiyama</span><br /></span><br /><span class="down" style="display: block;font-family:trebuchet ms;" id="formatbar_CreateLink" title="Link" onmouseover="ButtonHoverOn(this);" onmouseout="ButtonHoverOff(this);" onmouseup="" onmousedown="CheckFormatting(event);FormatbarButton('richeditorframe', this, 8);ButtonMouseDown(this);" ></span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-35817764114562991112007-07-18T12:57:00.000-07:002007-07-18T13:06:55.109-07:00Reading List from KDD 2007<span style="font-family:trebuchet ms;"><a href="http://www.kdd2007.com/">KDD 2007</a> will be on Aug 12-15 in the neighborhood at San Jose. Here is my selection:<br /><br />"Extracting Semantic Relations from Query Logs", Ricardo Baeza-Yates and Alessandro Tiberi<br /><br /></span><span style="font-family:trebuchet ms;">"Efficient Incremental Clustering with Constraints", Ian Davidson, S.S. Ravi, and Martin Ester<br /><br /></span><span style="font-family:trebuchet ms;">"A Probabilistic Framework for Relational Clustering", Bo Long, Zhongfei Zhang, and Philip S. Yu<br /><br /></span><span style="font-family:trebuchet ms;">"Tracking Multiple Topics for Finding Interesting Articles", Raymond Pon, Alfonso Cardenas, David Buttler, and Terence Critchlow<br /><br /></span><span style="font-family:trebuchet ms;">"Feature Selection Methods for Text Classification", Anirban Dasgupta, Petros Drineas, Boulos Harb, Vanja Josifovski, and Michael Mahoney<br /><br /></span><span style="font-family:trebuchet ms;">"Hierarchical Mixture Models: a Probabilistic Analysis", Mark Sandler<br /><br /></span><span style="font-family:trebuchet ms;">"Information distance from a question to an answer", Xian Zhang, Yu Hao, Xiaoyan Zhu, and Ming Li<br /><br /></span><span style="font-family:trebuchet ms;">"Statistical Change Detection for Multi-Dimensional Data", Xiuyao Song, Mingxi Wu, Chris Jermaine, and Sanjay Ranka<br /><br /></span><span style="font-family:trebuchet ms;">"Constraint-Driven Clustering", Rong Ge, Martin Ester, Wen Jin, and Ian Davidson<br /><br /></span><span style="font-family:trebuchet ms;">"Enhancing Semi-Supervised Clustering: A Feature Projection Perspective", Wei Tang, Hui Xiong, Shi Zhong, and Jie Wu</span><br /><br /><span style="font-family:trebuchet ms;"></span>Unknownnoreply@blogger.com0tag:blogger.com,1999:blog-7732226477253686468.post-42244232500716925502007-07-17T18:40:00.000-07:002007-07-18T00:09:04.862-07:00NLP in India?I was surprised to see <span class="entry-author-name"><a href="http://blog.outerthoughts.com/">Alex</a>'s <a href="http://blog.outerthoughts.com/2007/07/link-nlp-the-indian-perspective/">post</a> to which I don't agree fully.</span><br /><blockquote><span style="font-style: italic;">... because NLP is so underdeveloped in India, even undergraduate-level projects may be contributing to the cutting edge of research.</span><br /></blockquote>Turns out he was referring to <a href="http://technigal.wordpress.com/2007/07/16/natural-language-processing-the-indian-perspective/">this post</a> from an undergrad which tries to give "the Indian perspective", rather inaccurately. Having worked on NLP at one of the <a href="http://www.iitm.ac.in/">IIT</a>s I am compelled to write from a grad student perspective. Sunayana's post is interesting as it brings out several issues in Indic computing.<br /><br />1. Lack of annotation data - corpora, treebanks, and aligned texts which are sinews and bones of any language processing system. Resources exist, largely due to the efforts of <a href="http://www.ciil.org/">CIIL</a>, various universities and other government agencies but these are dwarfed compared to resources that exist for other languages, like English or the European languages.<br /><br />However, the rich morphology in Indian languages can be exploited to mitigate the amount of annotation data required for certain tasks, for instance <a href="http://www.cse.iitb.ac.in/%7Epb/papers/ACL-2006-Hindi-POS-Tagging.pdf">POS tagging</a>.<br /><br />2. Encoding issues - As rightly pointed by Sunayana, before the adoption of unicode, several data sources were locked up in the fonts they use. But things are changing, there is more and more Indian language content in unicode today than ever. Websites like BBC and Wikipedia are spewing out a lot of content in unicode for those interested in collecting monolingual, comparable corpora. A cursory glance at <a href="http://en.wikipedia.org/wiki/Wikipedia:Multilingual_statistics">Wikipedia statistics</a> shows the number of articles in, say <a href="http://stats.wikimedia.org/EN/ChartsWikipediaHI.htm">Hindi</a> or <a href="http://stats.wikimedia.org/EN/ChartsWikipediaTA.htm">Tamil</a> for example, has more than doubled in the past six months.<br /><br />3. Visibility - While there has been an increasing trend to publish in reputed conferences like ICML or ACL, more participation is certainly desirable. <a href="http://www.ijcai-07.org/">IJCAI 2007</a> was held in India and I highly recommend, if you are around, to submit (sub. deadline: Jul 31st) and/or attend <a href="http://www.ijcnlp2008.org/">IJCNLP 2008</a>.<br /><br />This is an exciting time to do NLP research on Indian languages. There is both corporate as well as government motivations which translate to grants and support to universities. The group at IIT Bombay, for example, implemented and deployed, local language based systems for helping farmers. Similar efforts have been taken by other institutes. Microsoft research at Bangalore, and IBM research at New Delhi and Bangalore are working on various projects on Indian Languages, including speech recognition.<br /><br />At the end of all this, I must partially agree with the quote I made from Alex's blog. Yes, <span style="font-style: italic;">some</span> undergrads do make brilliant contributions which is just because of what they have in their bones. This is true for any country or university.Unknownnoreply@blogger.com2