IWLeL 2004: Interactive Workshop on
Language e-Learning 2004
-
Date and time: 15:00-18:00 December 10th Friday, 2004
-
Venue: Conference Rooms #1 and #2, International Conference Center, Waseda University, Tokyo
-
Registration: Pre-registration by email is requested but not required. Attendance is free of charge.
-
Co-organized by
-
Sponsored by
-
PACLIC 18 Organizing Committee
-
Media Network Center, Waseda University
UPDATES and NOTICE
TIME TABLE (TENTATIVE)
- 15:00-15:15 Preparations / Opening / Warming-up
- 15:15-16:00 1st Round (presentations by odd numbers in the list)
- 16:00-16:45 2nd Round (presentations by even numbers in the list)
- 16:45-17:30 3rd Round (optional discussions and networking)
- 16:45-17:00 Early Removal
- 17:30-18:00 Cooling-down / Closing / General Removal
- 18:15-19:30 Optional Networking Get-Together elsewhere around campus
LIST of PRESENTATIONS (as of November 10th: subject to change without prior notice)
- Academic Presentations (in alphabetic order of the first author)
- AntConc: A Learner and Classroom Friendly, Multi-Platform Corpus Analysis Toolkit
Laurence ANTHONY (Waseda University)
- Computer Assisted Language Learning Based on Corpora and Natural Language Processing: The Experience of Project CANDLE
Jason S. CHANG (National Tsing Hua University)
- Extensive Reading with Guidance
Chin-chuan CHENG, Chu-ren HUANG, Xiang-yu CHEN, Yu-chun HUANG, Joyce HAN (Academia Sinica) & Feng-ju LO (Yuan-Ze University)
- Creating E-Learning Material to Teach Essential Vocabulary for Young EFL Learners
Kiyomi CHUJO (Nihon University), Chikako NISHIGAKI (Chiba University)
- A Web-Based Japanese Language Learning Project
Shinichi FUJITA, Toshiko HOSAKA, Akiko SWAN, Akihiro MASUOKA, Hiroaki OKUMOTO, Akira MORITA, Seinosuke NARITA (Waseda University) & Yong-Ju AN (Sunmoon University)
- Teaching tonal discrimination based on statistical properties and acoustical characteristics of the Chinese Four Tones: With regard to the contrast between tone-2 and tone-3
Shizuo HIKI, Kazuya IMAIZUMI, Kazuko Sunaoka & Liming YANG (Waseda University)
- Investigation into Language Learners' Acquisition Order Based on the
Error Analysis of the Learner Corpus
Emi IZUMI & Hitoshi ISAHARA (NICT)
- Evaluation of Reading Support Tools by Reading Comprehension Tests and Reading Speed Tests
Katsunori KOTANI (NICT), Takehiko YOSHIMI (Ryukoku University / NICT), Takeshi KUTSUMI (SHARP Corporation), Ichiko SATA (SHARP Corporation) & Hitoshi Isahara (NICT)
- Mobile CALL-drill and Web-test Tools for the Communicative Chinese Conversation Lessons
Song LIU, Kazuko SUNAOKA & Yoshiyori URANO (Waseda University)
- Pronunciation Portfolio: How were, are, and will be you?
Nobuaki MINEMATSU, Satoshi ASAKAWA, Koji OKABE, Shyu HENRICK, and Keikichi HIROSE (University of Tokyo)
- Pronunciation Understood: How intelligible do you think you are?
Nobuaki MINEMATSU, Satoshi ASAKAWA, Koji OKABE, Shyu HENRICK, and Keikichi HIROSE (University of Tokyo)
- Pronunciation Diagnosis: What to correct at first in YOUR case?
Nobuaki MINEMATSU, Satoshi ASAKAWA, Koji OKABE, Shyu HENRICK, and Keikichi HIROSE (University of Tokyo)
- Automatic Evaluation on Web-Based Foreign Language Learning System Tenjin
Kumiko TANAKA-Ishii (University of Tokyo)
- A System for Extensive Slash Reading Using Web
Shosaku TANAKA, Yoichi TOMIURA & Kensei YUKINO (Kyushu University)
- From Learners' Corpora to Expert Knowledge Description: Analysing Prepositions in the NICT JLE (Japanese Learner English) Corpus
Midori TANIMURA (Kinki University), Kazuhiro TAKEUCHI (NICT), Maruf HASAN (Shinawatra University) & Hitoshi ISAHARA (NICT)
- Corporate Presentations (in alphabetic order)
- eSurvey Institute, Ltd.
- GlobalEnglish
http://www.globalenglish.com
- Ordinate Corporation
http://www.ordinate.com
- QE Tech KK
http://qetec.com
- reallyenglish
http://www.reallyenglish.com
- SpeakESL
http://speakesl.com/
APPLICATION for WORKSHOP PRESENTATIONS
- Pre-registration for PACLIC 18 is suggested but not required.
- Please send email to IWLeL2004@decode.waseda.ac.jp if you are interested in giving an interactive presentation (system demonstration or poster) at the workshop.
- Please mention "APPLICATION for WORKSHOP PRESENTATIONS" in your email subject.
- Please include the following in your application:
You will be giving us implied consent to publish those information on our web page and elsewhere.
- Title of your presentation:
- Names and affiliation of authors, each in the following format:
Given_name FAMILY_NAME (Affiliation)
- Abstract of your presentation:
about 200-500 words
WORKSHOP PARTICIPATION and EMAIL (PRE-)REGISTRATION
- Pre-registration is requested but not required.
- Attendance is free of charge.
- Please send email to IWLeL2004@decode.waseda.ac.jp in advance (for pre-registration) or afterwards (walk-in), with participants' name, affiliation and contact informaition for our record.
- Please mention "WORKSHOP PARTICIPATION" in your email subject.
ABSTRACTS (as of November 30th: subject to change without prior notice)
- Academic Presentations (in alphabetic order of the first author)
- AntConc: A Learner and Classroom Friendly, Multi-Platform Corpus Analysis Toolkit
Laurence ANTHONY (Waseda University)
Using a corpus of texts as the basis for materials preparation in English language education has grown in popularity over the past few years. Corpora are now commonly used when selecting relevant vocabulary items, and are useful for finding common noun phrases and example sentences. They are also an invaluable resource when investigating word usage, especially in an English for Specific Purposes (ESP) context, where subject specific dictionaries and usage guides may be unavailable. On the other hand, the effectiveness of corpora in the classroom is less certain. In particular, teachers and learners have found that many of the available corpus tools are not appropriate for classroom use. Most of these tools have been designed for research purposes, and therefore have complicated interfaces and redundant features. They are also relatively expensive, especially if they need to be installed on all the learnersf computers. A further problem is that the most popular corpus tools only run on a Windows platform. For classrooms with Unix or Linux systems, there are few alternatives.
In this presentation, I will describe the development of AntConc, which is a freeware, multi-platform, multi-purpose corpus analysis toolkit, designed specifically for use in the classroom. AntConc started out as a simple concordance program for use by over 700 students in the technical reading and writing program at Osaka University, Japan. Feedback from the teachers and students in this program, and other users around the world, has led to many revisions and improvements. Now, AntConc offers a comprehensive set of tools including a powerful concordancer, word and keyword frequency generators, tools for cluster and lexical bundle analysis, and a word distribution plot. It can also handle html and xml formatted data, and allows users to choose from a range of search options, including simple wildcard searches and powerful regular expression searches.
After describing the background and development of AntConc, I will demonstrate some of its most important tools and features, and show how these can be easily accessed and used by learners. Finally, I will talk about the current limitations of the software, and how I hope to address these issues in the future.
- Computer Assisted Language Learning Based on Corpora and Natural Language Processing: The Experience of Project CANDLE
Jason S. CHANG (National Tsing Hua University)
This paper describes Project CANDLE, an ongoing 3-year project which uses various corpora and NLP technologies to construct an online English learning environment for learners in Taiwan. This report focuses on the interim results obtained in the first eighteen months. First, an English-Chinese parallel corpus, Sinorama, was used as the main course material for reading, writing, and culture-based learning courses. Second, an online bilingual concordance TotalRecall and a collocation reference tool TANGO were developed based on Sinorama and other corpora. Third, many online lessons, including extensive reading, verb-noun collocations, vocabulary were designed to be used alone or together with TotalRecall and TANGO. Fourth, an online collocation check program was developed for detecting V-N miscollocation and suggesting correct collocates in student text based on the theory of L1 interference and the database of BNC and the bilingual Sinorama Corpus. Other computational scaffoldings are under development. It is hoped that this project will help intermediate learners in Taiwan enhance their English proficiency with effective pedagogical approaches and versatile languagereference tools.
- Extensive Reading with Guidance
Chin-chuan CHENG, Chu-ren HUANG, Xiang-yu CHEN, Yu-chun HUANG, Joyce HAN (Academia Sinica) & Feng-ju LO (Yuan-Ze University)
A language learning mode called "word-focused extensive reading" has been proposed to facilitate word-usage learning. The user inputs a word to find its usage. The computer package then searches for the knowledge base for the word and provides guidance to show a summary of the salient features of word collocation. Then sentences with the word in question are displayed one at a time. For each sentence the relevant features of the collocating words are extracted and the number of their occurrence is tabulated. In this way the reader sees the quantitative information of word collocation so as to recognize salient features and thus participates in the building of the knowledge of word-usage. The program includes Chinese and English word collocation databases. For Chinese the knowledge was built on the basis of the Balanced Corpus of Academia Sinica. The English collocation data was collected from the British National Corpus.
- Creating E-Learning Material to Teach Essential Vocabulary for Young EFL Learners
Kiyomi CHUJO (Nihon University) & Chikako NISHIGAKI (Chiba University)
Increasingly, larger numbers of young learners in Japanese primary schools are starting to learn English. This expansion of Teaching English to Young Learners (TEYL) marks a major change that will affect secondary level teaching. Primary school teachers need effective and motivational teaching tools that will assist the development of TEYL successfully. In addition, these tools need to ensure that these early language-learning experiences will be built upon at the secondary level and beyond so that the benefits of an early start are not lost.
Theoretical and empirical research in EFL suggests that teaching essential, everyday words to elementary-aged children can be very beneficial for EFL learners in general, especially from the perspective of supplementing the vital domain of ELT vocabulary, which, in a Japanese EFL setting, is lacking in adequate teaching materials.
Specifically targeted for teaching essential vocabulary for TEYL, an effective and enjoyable e-learning material was developed for Japanese primary school children. Determining what vocabulary to include was done by collecting twenty picture dictionaries published overseas, and spoken data of infant to fourteen-year-old children taken from corpora such as CHILDES (Child Language Data Exchange System). From these databases, the 500-word base list was generated by using a combination of various statistical measures such as frequency, range, chi-square, and log-likelihood ratio. In addition to 'what to include', an equally important 'how to teach' aspect was incorporated into the program by including tasks integrating learning theory and information processing to ensure long-term word retention and a practical, real application to the students' daily lives. This was accomplished by using commercial application software called IBM Homepage Builder.
In this presentation, we will not only show the e-learning material, but also discuss the process used to produce it, and the rationale behind its creation.
- A Web-Based Japanese Language Learning Project
Shinichi FUJITA, Toshiko HOSAKA, Akiko SWAN, Akihiro MASUOKA, Hiroaki OKUMOTO, Akira MORITA, Seinosuke NARITA (Waseda University) & Yong-Ju AN (Sunmoon University)
This project has been developed a website that provides web-based multimedia materials for Japanese language learners in Japan and overseas. These materials include:
1) Distance learning using a Bulletin Board System (BBS) which aims at improving a learnerfs speaking ability by letting overseas students record their voice and upload this recorded voice and receive feedback through comments and corrections of the learnerfs pronunciation from a native speaker of Japanese. Korean students living in Korea have been studying beginner level Japanese through this distance learning program since March 2004.
2) Speech recognition system using short skits : A video based program with a speech recognition system was developed experimentally that enables students to use paired transitive and intransitive verbs instantaneously by pronouncing an appropriate verb in a given situation. After watching a short skit, learners have to choose which verb, transitive or intransitive, to use in a given situation. Learners pronounce the verb using a microphone, and this system can tell them whether their pronunciation is acceptable (recognizable) or not immediately after recording their voice.
The developers have also created other e-learning video materials for foreign students to help them understand cultural differences as well as other web based supplementary materials which show students their areas of weakness, so they can concentrate on studying those areas.
These materials are being used in Japanese language classes offered at Waseda and other universities.
- Teaching tonal discrimination based on statistical properties and acoustical characteristics of the Chinese Four Tones: With regard to the contrast between tone-2 and tone-3
Shizuo HIKI, Kazuya IMAIZUMI, Kazuko Sunaoka & Liming YANG (Waseda University)
Based on the statistical properties of occurrence frequency and transition probability of Chinese tones and the acoustical characteristics of their perceptual function and phonatory control, an effective method of teaching tonal discrimination to beginners was proposed, with special regard to the ontrast between Tone-2 and Tone-3. Referring to the difference from the Japanese word accent, intermediate tonal stimulus with synthetic speech and visual display of voice pitch are utilized in practicing listening, and the efficiency is improved by introducing the CAI algorithm. In practicing phonation, acceptable range is indicated according to the talker's voice range, which is derived from the extraction of fundamental frequency of speech. Possibility of transferring the teaching system to a self-learning program over the internet was also discussed.
- Investigation into Language Learners' Acquisition Order Based on the
Error Analysis of the Learner Corpus
Emi IZUMI & Hitoshi ISAHARA (NICT)
In second language education, it is important for teachers to know their students' acquisition order of major linguistic items in the target language. This leads them to teach these items more effectively in language classrooms. There have been a lot of studies done for revealing natural sequence in second language acquisition since 1970's, and it is an established idea that major grammatical morphemes are acquired in the common order by learners across different backgrounds such as their L1 or ages. In this paper, we tried to see what extent this idea could be applied in our corpus of Japanese leaner English named "The NICT JLE (Japanese Learner English) Corpus". In this corpus, the learners' grammatical and lexical errors have been annotated manually with 47 types of error tags, which is quite useful for investigating the acquisition order from the viewpoint of error analysis of learner language. In the experiment, we have found that there is no significant correlation between the long-held idea and the sequence extracted from our corpus. The most distinguished difference between these two is that articles and plural -s are acquired in the later stage by Japanese learners. This might arise from L1 transfer because Japanese language does not have any relevant markers with articles and plural -s. We assume that this result gives us some important clues for improving the current language teaching system.
- Evaluation of Reading Support Tools by Reading Comprehension Tests and Reading Speed Tests
Katsunori KOTANI (NICT), Takehiko YOSHIMI (Ryukoku University / NICT), Takeshi KUTSUMI (SHARP Corporation), Ichiko SATA (SHARP Corporation) & Hitoshi ISAHARA (NICT)
- Mobile CALL-drill and Web-test Tools for the Communicative Chinese Conversation Lessons
Song LIU, Kazuko SUNAOKA & Yoshiyori URANO (Waseda University)
Our presentation aims to introduce the construction of the Chinese Corpus WTCCiWaseda Tutorial Chinese Corpusjwhich supports our Chinese conversation classes and the development of the mobile CALL-drill and the Web-test that are based on this Corpus resources.
As of 2004 October, the Chinese language CCDL(Cultural Distance Learning) project at Waseda University affiliates with Peking University, National Taiwan Normal University and administrates twenty-five tutorial classes of four students each. We place native speakers in Japan, Beijing and Taipei as tutors to conduct Chinese conversation lessons by communicative approach. We have established an education management server on campus to administrate all the classes that stretch over multiple countries.
Chinese Corpus named WTCCiWaseda Tutorial Chinese Corpusjis located in this management server and it registers approx. 8000 words for the Tutorial Chinese lessons. This WTCC corpus information is based on The Grammatical Knowledge-base of Contemporary Chinese corpus by Peking University (approx. 10000 words), HSK (Hanyu Shuiping Kaoshi, approx. 8000 words) and Japanese Basic University Educational Vocabulary corpus (approx. 3000 words). We have added linguistic information such as Japanese translation, pronunciation signs in pinyin, part of speech, and the frequency level marks as tags. This Corpus also provides the database for prep/review drills for students, class report manuals for tutors, Web-based achievement tests and the CALL (computer assisted language learning) drills.
Furthermore, we are developing new functions as follows. Firstly, we are building an automatic registry system that adds new vocabularies used in the live tutorial lessons into the Corpus automatically. New vocabularies will be reported by the tutors after each class and will be registered in the server. The server refers to the WTCC Corpus database and if it is already registered, it will provide the educational data for further study. If it is not yet registered, it will be acknowledged as a new entry and the server makes an additional entry. This additional information will be circulated among the students through their computers and mobile phones. This whole process will be computerized.
Secondly, we are establishing the mobile education system by connecting the server and mobile phones to provide achievement tests and CALL drill functions. The current issues are the use of Chinese fonts on the Japanese mobile phone screens and the development of mobile technologies to carry the same multimedia information as the Web. Our goal is to adopt the Unicode system on mobiles to expand the usage of Web-tests and CALL drills.
- Pronunciation Portfolio: How were, are, and will be you?
Nobuaki MINEMATSU, Satoshi ASAKAWA, Koji OKABE, Shyu HENRICK, and Keikichi HIROSE (University of Tokyo)
No two students are the same. There are 2 billion students of English on this planet and each student is always changing through training. This means that there are more than 2 billion types of English pronunciations. But there has been no good method so far to represent the tremendous number of pronunciations individually. No writing, no civilization. It's chaos. In this situation, students have been asked for tens of years "How are you?" And they have been requested to answer "I'm fine." The authors believe that there are no more ridiculous conversations than that. The correct answer should be "I don't know how I am. You don't know either, do you!" This study introduces a very novel method to represent the more than 2 billion pronunciations individually. The method is based on physical implementation of structural phonology and the implementation can be regarded as mathematical interpretation of Saussure's claim that language is a system of conceptual differences and phonic differences. Each student's pronunciation is acoustically and entirely represented as phonemic structure. But the structure cannot have any dimensions to indicate the non-linguistic features like age, gender, speaker, microphone, room, line, etc. The presentation will show that the pronunciation represented by the proposed method is both purely acoustic and purely linguistic. I'm warning you that this is not magic. Message to teachers; Do you still continue saying "How are you?" or do you start saying "I konw you."
- Pronunciation Understood: How intelligible do you think you are?
Nobuaki MINEMATSU, Satoshi ASAKAWA, Koji OKABE, Shyu HENRICK, and Keikichi HIROSE (University of Tokyo)
Two goals are possible in pronuntiation training. Native-souding and intelligible pronunciations. In most of the cases, however, students speak English just to communicate with others. It is known that some Japanese accented pronunciations are accepted but that the others are not. What is accepted and what is not accepted. The answer strongly depends on hearing and perceptual characteristics of listeners. In this study, characteristics of ears of Americans is focused on. A large number of samples of Japanese English were presented to 6 American listeners who had lived in Japan for less than 1 year and were asked to type what they heard. After the typing experiments, each word of all the utterances had probability of being correctly identified. Speech and language technologies were used to calculate values of many acoustic/linguistic parameters of the invidual words. CART was done to examine what kind of pronunciation errors tend to be acceptable or unacceptable. Results showed that the most important factor of the pronunciaiton is its rhythm. The second one is use of easy and plain words. The third is adequate pronunciation of the individual phonemes. The correct identification rates of the individual words were predicted by American and Japanese teachers only by hearing. It is surprizing that Japanese teachers are very poor in indicating that the completely unintelligible pronunciations are unintelligible, implying that hearing characteristics of Japanese is quite different from that of Americans. It is much the case with Japanese students. Pronunciation training should be done based on this fact because the purpose of the training is smooth communication with listeners. Message to teachers; "Repeat after me." "Listen to me." They are the phrases repeated in classrooms billions of times. But it may be meaningless at all because students don't know how to repeat and how to listen. They don't know about ears of the listeners.
- Pronunciation Diagnosis: What to correct at first in YOUR case?
Nobuaki MINEMATSU, Satoshi ASAKAWA, Koji OKABE, Shyu HENRICK, and Keikichi HIROSE (University of Tokyo)
No two students are the same. There are 2 billion students of English on this planet. They need 2 billion different kinds of instructions on what to correct at first in their pronunciations. This study proposes a very novel method to automatically generate the 2 billion instructions optimally and individually devised for the 2 billion students. In the case of Japanese students of English, r & l confusion and s & th confusion are often observed. Can you answer the following question? "Which confusion should be corrected at first in my case?" It should be r & l for some students and s & th for the others. If you cannot answer adequately based on the strength of Japanese accentedness observed in the student's pronunciation, to be a good teacher, you must use the method we're proposing. The method is realized based on the new representation of the pronunciation; phonemic structure of the pronunciation. The phonemic structure of a student is linguistically matched with the lexical structure of the target language. The phonemic strucutre can be divided into some sub-strcutures. By considering which sub-structure causes the largest damage when communicating in the target language with the student's phonemic structure, the instruction will be automatically generated on what to correct at first in his/her case. I'm warning you that this is not magic. Message to teachers; Can you introduce every student to the target language, not introduce the target language to every student?
- Automatic Evaluation on Web-Based Foreign Language Learning System Tenjin
Kumiko TANAKA-Ishii (University of Tokyo)
We report on Tenjin, a web-based system supporting foreign language education. It provides an integrated environment enabling students to solve assignments and teachers to evaluate them. Tenjin automatically evaluates assignments at the time when they are submitted by students, which helps students to study individually and improve the quality of their final submission. This is performed by language independent scheme based on dynamic programming and pattern matching by regular grammar. Currently, Tenjin is run on 6 languages. We describe the design of Tenjin and report our classroom experience using it with a class from 2003 to 2004.
- A System for Extensive Slash Reading Using Web
Shosaku TANAKA, Yoichi TOMIURA & Kensei YUKINO (Kyushu University)
This presentation shows a system for "Extensive Slash Reading" using Web. Slash reading is one of popular methods for learning languages, where learners read sentences slashed at boundaries among sense groups. This method requires learner to read a lot of slashed sentences as extensive reading, then the method is called "extensive slash reading."
The followings are some of requirements for extensive reading in (Day&Bamford 1998), and the method should fulfill them as well.
1) A variety of materials on a wide range of topics is available so as to encourage reading for different reasons and in different ways.
2) Learners select what they want to read and have the freedom to stop reading material that fails to interest them.
3) Reading materials are well within the linguistics competence of the learners in terms of vocabulary and grammar.
4) Reading is individual and silent, at the student's own pace, and, outside class, done when and where the student chooses.
Additionally, materials for slash reading must satisfy the following character, of course.
5) Sentences in materials have slashes at boundaries among sense groups.
From the viewpoint of the volume, variation and topicality, texts on Web are better than authentic or published ones. However, some of them are not proper as materials because anyone can export his/her text on Web very easily. Though there are some interesting texts, they can be nongrammatical and informal such as private memorandums or informal messages. Besides, the learners cannot know the level of a text on Web before reading and very few texts on Web have been slashed.
This system proposed here tries to solve these problems using natural language processing techniques.
The system consists of the following three modules.
1) Filtering module: The first module calculates readability of a text as a material to ensure it is within the linguistics competence of learners.
2) Slashing module: The second module identifies sense groups in the sentences and slashes them.
An algorithm for identifying sense groups is based on chunking process (Tanaka and Tomiura 2004). We can get a slashed material from an arbitrary text on Web by this module.
The detail about the algorithm is described in the poster.
3) Reading module: This module is optional for learners.
If the learners use this module, it shows them the sentences per sense group. Consequently, the learners can read and understand the sentences automatically in a sequence order as its words appear.
This system enables learners to do an efficient slash reading using proper texts on Web. Furthermore, because this system is running as a Web server, if learners can access the Internet, they can use it with a Web browser wherever and whenever they want.
- Learners Corpora to Expert Knowledge Description: Analysing Prepositions of the NICT JLE (Japanese Learner English) Corpus
Midori TANIMURA (Kinki University), Kazuhiro TAKEUCHI (NICT), Maruf HASAN (Shinawatra University) & Hitoshi ISAHARA (NICT)
Annotated Learner Corpora are recent attempts to isolate or codify error-related knowledge. In general, they consist of at least 3 common attributes: incorrect instance, respective corrected instance, and annotation (Why-scheme). The third attribute, annotation, also called Why-scheme, generally describes errors in the following two major ways. Firstly, error-tag based on a predefined error-hierarchy. Secondly, narrative - based on natural language descriptions of the reasons for learners' errors (ex. "tense sequence error" or "past tense form of pay is not payed," etc.)
Each attribute provides us with useful information, for example, errors learners often tend to make, insights into causes of errors etc. However, for a particular error, the first two attributes, incorrect instance and respective corrected instance, identified by different annotators may not be unique (e.g., N-V agreement error: either N or V can be corrected to achieve conformity). For the third attribute, Why-scheme by error tags, it is easy to compile statistics but difficult to describe errors in detail, since there are various, unpredictable errors occurring in learners' English. The opposite is the case, in looking at Why-scheme by narrative.
Sharing and reusing corpora annotated with different error-scheme seems to be difficult. However, by sharing the terms to describe Why-scheme on the Internet and constraining them in some way, we can facilitate integration and reuse heterogeneous learner corpora.
In this presentation, we use Japanese Learner's Spoken English, NiCT JLE Corpus (formerly SST corpus, Tono et al., 2001) which contains 9 different proficiency levels (level 1 the most elementary and level 9 the most advanced). Although this corpus has already been error-tagged on the basis of 45 kinds of grammatical terms, it still has more room to improve the error tagging system. In order to explore the possibilities of the error-tagged corpus, further investigation into details is required. To that end, we are now analyzing learners' errors focusing in particular on prepositions, one of the most difficult word classes to acquire, and describing their features in more formal way. Using these features, we examine preposition error tendencies of Japanese English learners, considering teaching and learning in the classroom. Not only would this work provide us with notable features of the JLE corpus, but would contribute to the development of general error description.
PROGRAM COMMITTEE MEMBERS
- Program Chair
- Yasunari HARADA (Waseda University)
- Members
- Yoshinori SAGISAKA (Waseda University)
- Hitoshi ISAHARA (NICT)
- Masao UTIYAMA (NICT)
- Kiyomi CHUJO (Nihon University)
- Nobuaki MINEMATSU (University of Tokyo)
- Proceedings Editors
- Laurence ANTHONY (Waseda University)
- Shinichi FUJITA (Waseda University)
NOTICES