Announcement for a Research Meeting on Saturday, December 1st, 2012

The 13th Korea-Japan Workshop on Linguistics and Language Processing: Corpora, Annotation and Human Language Processing

Announcements

Registration:
No registration process/fee required for participation.
Dates and Venue

Date	December 1st Saturday, 2012
Tentative Venue	14-603 on the sixth floor of building 14 in Waseda Campus of Waseda University

program

10:00-10:10

preparations

10:10-11:45

Open Lecture (pre-workshop session)

moderator Yasunari Harada

10:10-10:15	Yasunari Harada	Introduction
10:15-11:45	Kiyong Lee	Open Lecture: Annotating Semantic Information in Natural Language from Texts to Pictures or Maps [abstract] [slides]

11:45-13:15

lunch break / registration

13:15-15:50

session 1

moderator Ryo Otoguro

13:15-13:20	Jae-Woong Choe	workshop opening
13:20-14:10	Takenobu Tokunaga and Ryu Iida	Construction and Use of Multimodal Corpora of Referring Expressions in Collaborative Problem Solving Dialogues [abstract] [slides]
14:10-15:00	Kyoko Ohara	Japanese FrameNet: Lexicon and Constructicon Building for Japanese [abstract] [slides]
15:00-15:50	Takane Ito, Yuki Kobayashi, and Yoko Sugioka	ERP responses to violations in Japanese verb conjugation patterns [abstract] [slides] [pdf for: Yuki Kobayashi, Yoko Sugioka and Takane Ito (2012) "ERP Responses to Violations in Japanese Verb Conjugation Patterns" in N. Miyake, D. Peebles, & R. P. Copper (eds.) Proceedings of the 34th Annual Conference of the Cognitive Science Society, 611-616, Austin, TX: Cognitive Science Society.]

15:50-16:10

break

16:10-18:15

session 2

moderator Sachiko Shudo

16:10-16:30	Yasunari Harada, Mayumi Kawamura and Masanori Suzuki	Data Collection of Spontaneous Extended Utterances by Japanese Learners of English [abstract]
16:30-17:10	Gabor Pinter	Towards User Generated Speech Databases in Language Education [abstract] [slides] See also: http://www.pinlab.info/talks/20121201-kj13/20121201-kj13-v02.pdf
17:10-18:10	Jae-Woong Choe	Adnominal Case Constructions: a contrastive study in Japanese and Korean based on a parallel corpus [abstract] [slides]
18:10-18:15	Kiyong Lee	workshop closing

18:15-18:25

wrap up and clean up

19:00-21:00

further discussion

access maps

http://www.waseda.jp/jp/campus/index.html

http://www.waseda.jp/jp/campus/waseda.html
Topics

syntax, semantics and pragmatics

natural language processing

human sentence processing

corpus linguistics

Supported in part by:

Grants-in-Aid for Scientific Research (B)

21320109

Additional Data Collection and Preliminary Analysis of Elicited Extended Spontaneous Utterances by English Learners in Japanese Undergraduate Courses

Jointly hosted by:

Institute for the Study of Language and Information, Waseda University

Institute for Digital Enhancement of Cognitive Development, Waseda University

abstracts

Annotating Semantic Information in Natural Language from Texts to Pictures or Maps
For the past ten years or so, one of ISOfs working groups, namely ISO/TC 37/SC 4/WG 2, has been developing annotation schemes for marking up some relevant semantic information in natural language. Among them are two semantic annotation schemes, which were published as ISOfs international standards in early 2012: one is ISO-TimeML for the annotation of temporal and event-related information in text, and the other SemAF-DA for the annotation of dialogue acts in (transcribed) dialogues with multimodal information, namely information coming from gestures, facial expression, or tones. Besides these two published ISO standards, there is one newly proposed semantic annotation scheme, named ISO-Space, on spatial and motioninvolving information. Unlike the other two annotation schemes, ISO-Space attempts to extend the target of its annotation task from text to visual objects such as pictures, both still and moving, and maps. Todayfs talk focuses on ISO-TimeML and ISO-Space, showing how they can annotate written text as well as visual material such as pictures and maps. At this preliminary state, we should simply assume that annotators are humans, for it could be a totally different problem if we try to develop a program that enables machines to extract semantic information from multimedia data and annotate such information to the given data.

Construction and Use of Multimodal Corpora of Referring Expressions in Collaborative Problem Solving Dialogues
In the last couple of years, we have been developing a collection of multimodal corpora of referring expressions named the REX corpora. The corpora have two notable features, namely (1) they include time-aligned extra-linguistic information such as participant actions and eye-gaze on top of linguistic information, (2) dialogues were collected with various configurations in terms of the puzzle type, hinting and language.
In this talk, after describing how the corpora were constructed and sketching out each, I will present an analysis of various statistics for the corpora with respect to the various configurations mentioned above.
Time permitting, I will present some research results using these corpora, including, resolution and generation of referring expressions and comparable study of usage of referring expression in English and Japanese.

Japanese FrameNet: Lexicon and Constructicon Building for Japanese
This talk will discuss Japanese FrameNet, an online Japanese lexical resource created in the style FrameNet, i.e. based on Frame Semantics and supported by corpus evidence. The talk will focus on what the lexical resource is about, what kinds of theories are behind it, and how we can use it for what purposes.
So far, the Japanese FrameNet project has been building a frame-based lexicon of Japanese through two annotation modes, namely, lexicographic annotation and full text annotation. In addition, through our full text annotation work, we have started to describe the relations between sentential, lexical, and constructional meanings. We are currently investigating whether the semantic notions for building a frame-based lexicon are also independently needed to account for the meaning of constructions in Japanese, by trying to build a gconstructiconh, a registry of constructions, of Japanese (cf. Fillmore 2008, Fillmore & Baker 2010, Fillmore, Lee-Goldman, & Rhodes 2012, Lee-Goldman & Rhodes 2009).

ERP responses to violations in Japanese verb conjugation patterns
ERP (Event Related Potential) measurement using the violation paradigm of Japanese verb conjugation was conducted in order to investigate the mental and neural mechanisms involved in the processing of different conjugation patterns. A LAN-like component followed by a P600 was elicited for the anomaly of using a tense-bearing form with the negative ending, while only P600 was observed for the anomaly of using an infinitive form in the same environment. The non-application of morpho-phonological changes of verb roots (gonbinh) yielded an N400 component and a P600. The P600 components observed in all types of errors reflect the cost of processing morphological and/or syntactic anomalies, while the difference in the negativities suggest that two different mechanisms of rule-based computation and lexical memory are involved in the processing of Japanese verb conjugation.

Data Collection of Spontaneous Extended Utterances by Japanese Learners of English
In this presentation we will report on our efforts on collecting spoken utterances by English learners studying in Japanese university English courses. Most of our efforts so far have focused on data collection, with some preliminary manual transcription of a limited part of the collected utterances. For this purpose, we developed a distributed data handling system utilizing web-browsers for data input. The transcribed portion represents about 1/20 of the core data to be compiled and the transcription so far came to roughly 150,000 words including 10% of filler-like expressions, suggesting that when we have transcribed all of the core spoken utterances for the eight-year project, it will come to something like 3,000,000 words. This is a very rough estimat but will give some indication of the scale of the project under way.
All of the learners participating in this project are taking Versant English Test, which is an automated spoken English test, four times a year. This test utilized speech recognition technology and automatically reports test takers English proficiency in oral interactions in terms of a total score plus four sub-scores of vocabulary, sentence mastery, pronunciation and fluency.

Towards User Generated Speech Databases in Language Education
The present research aims to offer an alternative to the conventional, top-down approach to speech data collection in language education settings by putting forward a bottom-up, usage-oriented design. A web-based software platform is being developed with the primary goal of enhancing speech recording activities in language classes. The platform offers a convenient, non-constrained method of collecting and evaluating learners' speech, which can greatly enhance the compilation of learners' speech databases in addition to its educational value.
The system is not ready for public release yet, still there are various aspects of the project that can raise the interest of both developers and users of spoken language databases starting with the highly technical details of sound library integration and continuing up to the practical questions of interface design.

Adnominal Case Constructions: a contrastive study in Japanese and Korean based on a parallel corpus
The adnominal case constructions in Korean and Japanese seem identical in terms of their syntactic configuration, and yet the Japanese adnominal marker '-no' is used more than twice as much as its counterpart in Korean '-ui'. The question we address in this presentation is what is the nature of the difference, namely, how can we properly describe the difference in distributions. For example, is the difference due to certain lexical properties of some Japanese nouns? It is possible that there is a certain limited set of nouns in Japanese that might induce the '-no' construction in an unusually productive way. Or is the difference part of some general tendency in the use of the construction in Korean and Japanese, not particularly tied up with lexical properties? In order to answer some of these questions, we extracted relevant constructions ('N1-ADN N2') and nouns from the Sejong Korean-Japanese Parallel corpus and then analyzed them focusing on the similarities and differences between the two languages. In the presentation we will focus on the similar/different distributional patterns between the two languages.

authors, moderators and organizers

Jae-Woong Choe received his Ph. D. in linguistics from University of Massachusetts, Amherst in 1987, and is currently Professor of Linguist ics at Korea University. He is a former President of the Korean Society for Language and Information and served as a Steering Committee member of PACLIC for the period of 2005-2012. His areas of interest include Theoretical Linguistics, mainly on Semantics and Pragmatics, and Computational/Corpus-based Linguistics.

Yasunari Harada is Professor at Faculty of Law and Director at the Institute for Digital Enhancement of Cognitive Development of Waseda University. He is a former President of the Logico-Linguistics Society of Japan and a Steering Committee member of PACLIC, With Jong-Bok Kim as his partner and counterpart, he has been organizing Korea-Japan Workshop on Linguistics and Language Processing since 2004.

Ryu Iida is at Department of Computer Science, Tokyo Institute of Technology.

Takane Ito is Professor at the University of Tokyo, Komaba.

Mayumi Kawamura is chief Language Annotator in Project VALIS.

Jong-Bok Kim is Professor at School of English and Director at Center for the Study of Language and Information of Kyung Hee University in Seoul. He received his Ph.D. in Linguistics from Stanford University in 1996 and has been working on syntax, semantics, and computational linguistics for Korean and English. His current research topics include comparative constructions in English and Korean, inversion phenomena in English, complex predicates in Korean, cleft constructions, focus projections in English and Korean, just to name a few. In addition to publishing numerous papers in prestigious domestic and international journals, he is the co-author of the textbook ``English syntax: an introduction" (CSLI Publications), which is now worldwide used as an undergraduate and graduate textbook. Since 2003, he has also turned his attention to computational implementation too and developed a computationally feasible grammar for Korean, Korean Resource Grammar.

Yuki Kobayashi is at Center for Evolutionary Cognitive Sciences, the University of Tokyo, Komaba.

Kiyong Lee is Professor Emeritus of Linguistics at Korea University and has been serving as ISO/TC 37/SC 4/WG 2 Convenor for the past ten years or so.

Kyoko Ohara is Professor at Keio University.

Ryo Otoguro is Associate Professor at Faculty of Law of Waseda University in Tokyo.

Gabor Pinter is an associate professor at School of Languages and Communication, Kobe University. He received his PhD in linguistics (perceptual phonology) from Kobe University in 2008. His research interest includes perceptual phonology, speech recognition, machine learning and theoretical linguistics.

Sachiko Shudo is Associate Professor of English and Linguistics in Waseda University in Tokyo. She received her Ph.D. (1998) in linguistics from Georgetown University. Her dissertation, The Presupposition and the Discourse Function of the Japanese Particle mo, was published in Routledge's Dissertation Series. Her main research interests include pragmatic issues involving presupposition.

Masanori Suzuki is Principal Test Development Manager of Pearson Knowledge Technologies.

Yoko Sugioka is Professor at Keio University.

Takenobu Tokunaga (a.k.a. Tokunaga, Takenobu) is Professor at Department of Computer Science, Tokyo Institute of Technology. His research interests include building and managing language resources, application of human language technologies to intelligent information access, and dialogue systems.

notices

First drafted September 13th, 2012. Last revised December 3rd, 2012.
The meeting and talks announced in this web page are subject to change without prior notice. The organizers should not be held responsible for any purported or actual damages by prospective participants due to those changes.