| 研究課題/領域番号 |
18K12427
|
|---|
| 研究機関 | 神戸大学 |
|---|
研究代表者 |
ALBIN Aaron 神戸大学, 国際文化学研究科, 講師 (60794526)
|
|---|
| 研究期間 (年度) |
2018-04-01 – 2020-03-31
|
| キーワード | corpus / textbook / accent / Japanese / pitch / database / L2 / SLA |
|---|
| 研究実績の概要 |
During the first year of the project, we have successfully created a comprehensive digital database of the contents of the GENKI series of elementary Japanese textbooks (Japan Times Publishing, 2nd edition, 2011). This includes Textbook 1, Textbook 2, Workbook 1, Workbook 2, the Answer Key, and a specific subset of contents from the Teacher's Manual.
In addition, approximately half of the database has been annotated for pitch accent. In the first stage of annotation, a trained phonetician marked accent locations for all database entries with official CD audio. All other entries were given a first-pass annotation using the online tool "Suzuki-kun". All annotations were then hand-checked by a team of two trained undergraduate research assistants, and any conflicts were discussed and resolved.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
The original plan for the first year of the project was:
Semester 1: Build textbook database
Semester 2: Add accent annotations
Semester 1 was completed more or less as originally planned. (See above description for the scope of the completed work.) Regarding semester 2, ultimately only about half of the accent annotations were completed. However, due to adjustments made to the project plans for the second year, this will not be a problem, and it will still be possible to answer the original research questions.(See answer below for "Planning for the Future Work".)
|
| 今後の研究の推進方策 |
The original plan for the project's second year was:
Semester 3: Build two components of an AI system that can produce the accent patterns covered in the database ("lexicon" and "rulebase")
Semester 4: Run learning simulations using this AI system
It was discovered that the tasks originally planned for Semester 3 are largely unnecessary. Instead of making a lexicon from scratch, information can be used from UniDic - a large digital dictionary for Japanese that includes accent information. Instead of building a new rulebase, a program called Chamame can be used to analyze the text in the database and output which accent rules are involved in each sentence. By using these preexisting tools, the work in Semester 3 has become lighter, thus making up for the slight delay in the annotation process.
|
| 次年度使用額が生じた理由 |
A conference presentation was accepted for the Spring 2019 Acoustical Society of Japan. However, the Principal Investigator fell ill with severe food poisoning, resulting in a cancellation of the presentation. This created a surplus in the budget from the first year. However, since the accent annotation process will continue for one semester longer than originally planned, this works out quite well, since the surplus money can be used to fund the additional annotation work during Semester 3 of the grant.
|
