{"id":138,"date":"2006-09-13T03:11:00","date_gmt":"2006-09-12T18:11:00","guid":{"rendered":"https:\/\/fugutabetai.com\/blog\/2006\/09\/13\/information-processing-society-of-japan-natural-language-meeting\/"},"modified":"2006-09-13T03:11:00","modified_gmt":"2006-09-12T18:11:00","slug":"information-processing-society-of-japan-natural-language-meeting","status":"publish","type":"post","link":"https:\/\/fugutabetai.com\/blog\/2006\/09\/13\/information-processing-society-of-japan-natural-language-meeting\/","title":{"rendered":"Information Processing Society of Japan Natural Language Meeting"},"content":{"rendered":"

Notes from the 2006-09-12 to 13 Information Processing Society of Japan meeting. The Information Processing Society of Japan Special Interest Group on Natural Language Processing<\/a> holds bi-monthly meetings all around Japan. Two months ago, the meeting was in Hakodate. This time, the meeting was in Shinjuku, very close to where I live, so I decided it would be a good chance to attend and see what research is going on in the field in Japan. <\/p>\n

It was really interesting. All but two of the presentations were in Japanese, which was a very nice chance to get up to speed on technical Japanese, and to see how presentations here go. It was pretty tiring too though. I also had a chance to talk with some of the member of the \u60c5\u5831\u7206\u767a\u4e16\u754c\u30cb\u30e5\u30fc\u30b9 group that I’m involved with. <\/p>\n

If you are interested in reading some very surface comments about the papers that I saw on Tuesday’s session, click to read more…<\/p>\n

<\/p>\n

The first presentation was about concepts that can not be found in
\ndictionaries or wikis. How do these concepts operate like technical
\nterms, and how are they related? It looks like the input is books:
\nthey use scanners and OCR over books. He builds a matrix of technical
\nterms over the publications. Given a term, you can then pick out the
\ncontexts that it can appear in, and find related terms. <\/p>\n

—<\/p>\n

Second Presentation is from New York University – I met him at SIGIR –
\nNobuyuki Shimizu – “Knowledge Frame Extraction From Navigational Route
\nInstruction” <\/p>\n

Giving directions to a robot or something, how do you plan the path?
\nThere are three actions, turn left\/right, go straight, enter a door.
\nTakes those directions and fills a frame with them. <\/p>\n

—<\/p>\n

Shimomura-san from Okayama daigaku. <\/p>\n

He is trying to determine the agent, theme, and goal of sentences. So
\nthe problem is to assign parts, roles to the actors in the sentence.
\nSo it turns out they’ve been studying many of these, and now today’s
\npresentation is on “Reason” and “Opponent”. They use EDR to get
\ninformation on nouns, and have some interesting rules that use EDR
\ncategories and particles for information on how to classify nouns
\n(actor, agent, theme, goal, etc.) <\/p>\n

—<\/p>\n

Tokyo Daigaku – Hiroko Ishida-san – “Study of sensory classification
\nas co-occurring terms with imitative words” <\/p>\n

This is about retrieving information from medical type databases using
\nnormal Japanese, and the giongo \/ gitaigo associated with describing
\nsounds or feelings in organs. <\/p>\n

They used goo \u30d8\u30eb\u30b9\u30b1\u30a2 as a resource, as well as a giongo\/gitaigo
\ndictionary, and a medical dictionary. They hand-classified about 2300
\nterms from goo into about 8 categories on what part of the body they
\nrelate to. <\/p>\n

—<\/p>\n

I had lunch with Yoshioka-sensei, Mori-sensei, and others.<\/p>\n

—<\/p>\n

Yoshioka’s presentation: Classification of Anchor Text for Web
\nInformation Applications<\/p>\n

He looked at the anchor text in 100gigabytes of web pages (from the
\nNTCIR web test collection) and how inter- and intra- site link anchor
\ntext differs. About 100,000,000 links. There is some interesting
\nanalysis of the link content. He wants to classify link text into 8
\ntypes of categories. <\/p>\n

—<\/p>\n

Hirao Kazuki – “Web Search Result Clustering Based on Structure of
\nCompound Nouns” – Okayama Daigaku<\/p>\n

The idea here is to use the compound nouns in Japanese documents to
\ncluster them. Also, since compound noun composition is fairly easy to
\nunderstand, we can use that to build a hierarchy of clusters. \u8907\u5408\u540d
\n\u8a5e is compound noun in Japanese. You can break them up into a
\nsupplemental concept and main concept (\u88dc\u8db3\u90e8 \u4e3b\u8981\u90e8) That can be used
\nto build a concept hierarchy. The build their clusters, and then have
\na nice way of labeling the clusters. In particular, if the cluster
\nwas generated by a keyword that is a sub-concept that they all share,
\nthen they make a label like “XX Recipes”. <\/p>\n

—<\/p>\n

Hiroyuki Sakai – “Estimation of Impact Contained in Articles about
\neach Company in Financial Areas” – Toyoshahi University<\/p>\n

This also looks like a very interesting and relevant (to my work)
\npaper. He is talking about identifying documents and sentences that
\nwill have an “impact” on business. I’m having a really hard time
\nunderstanding this presentation (the presenter speaks quickly) but it
\nlooks like they calculate the stock price the day before and the day
\nafter the date of the news article, and look at that to calculate the
\nimpact. In general, I think this is a very good approach (and for
\nlonger periods of terms, like over a week, look at the week-behavior
\nof the stock price and the centroid of articles over that week.) <\/p>\n

They extract some terms that have a big impact (via entropy) on the
\ncompany. <\/p>\n

—<\/p>\n

Daisuke Matsuzaki, Toshinori Watanabe, Hisashi Koga, Nuo Zhang –
\n“Document Relation Analysis by Data Compression” – University of
\nElectro-Communications <\/p>\n

I wasn’t particularly interested in this paper, and I’ve seen work
\nbefore on using compression to do document categorization.<\/p>\n

—<\/p>\n

Testuya Sakai – “Controlling the Penalty on Late Arrival of Relevant
\nDocuments in Information Retrieval Evaluation with Graded Relevance” –
\nKnowledge Media Laboratory, Toshiba R&D Center.<\/p>\n

I’ve read one of his papers on Q-measure before, and it seems quite
\nsimilar. <\/p>\n

—<\/p>\n

Hiroka Kimura, Toshinori Watanabe, Hisashi Koga, Nuo Zhang – “A New
\nDocument Retrieval Method Using LZ78 Compression Function” <\/p>\n

They did some summarization experiments using artificial documents.
\nThey had a few strategies: centroid summarization, most similar
\nsummarization, and most unique summarization. I didn’t really follow
\nwhat the documents were like and what they evaluated against. They
\nalso used Japanese news documents to do real summarization with human
\nmodels to compare against. I don’t understand how the summaries were
\nscored either actually… People liked centroid and most novel types
\nof summarization most. It looked like it was more of a human scoring
\nevaluation than anything else. <\/p>\n

This particular paper was the focus of a lot of questions. The main
\nthrust of the questioning was “how is this different from a
\nvector-space model for summarization?” for the most part. The
\ninteresting part of the answer is that this approach works really with
\nbinary data, so application to text is a first step, but conceivably
\nthese approaches can be applied to pictures or video as well.
\nCertainly I agree that is true, but I’m not sure if there is a strong
\nargument to be made for how to interpret the results over those kinds
\nof media. <\/p>\n

—<\/p>\n

Junichi Fukumoto, Tsuneaki Kato, Fumito Masui, Tatsunori Mori, Noriko
\nKando – “An Automatic Evaluation of Question Answering using Basic
\nElements” – Ritsumeikan University, University of Tokyo, Mie
\nUniversity, Yokohama National University, National Institute of
\nInformatics<\/p>\n

An overview of the BE evaluation method for Question and Answering.
\nIt will be used in the QAC at NTCIR this year. They have also created
\na Japanese version of a BE creator that runs over Kabochya output. <\/p>\n","protected":false},"excerpt":{"rendered":"

Notes from the 2006-09-12 to 13 Information Processing Society of Japan meeting. The Information Processing Society of Japan Special Interest Group on Natural Language Processing holds bi-monthly meetings all around Japan. Two months ago, the meeting was in Hakodate. This time, the meeting was in Shinjuku, very close to where I live, so I decided […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[10],"tags":[],"_links":{"self":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/posts\/138"}],"collection":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/comments?post=138"}],"version-history":[{"count":0,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/posts\/138\/revisions"}],"wp:attachment":[{"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/media?parent=138"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/categories?post=138"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fugutabetai.com\/blog\/wp-json\/wp\/v2\/tags?post=138"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}