2019 Fiscal Year Research-status Report
Using artificial intelligence to simulate the process of learning Japanese accent rules: Toward the integration of accent into the grammar curriculum
| Project/Area Number |
18K12427
|
|---|
| Research Institution | Kobe University |
|---|
Principal Investigator |
ALBIN Aaron 神戸大学, 国際文化学研究科, 講師 (60794526)
|
|---|
| Project Period (FY) |
2018-04-01 – 2021-03-31
|
| Keywords | corpus / textbook / Japanese / accent / pitch / prosody / L2 / SLA |
|---|
| Outline of Annual Research Achievements |
During the Reiwa 1 fiscal year, we annotated the second half of the database for pitch accent in the same manner as the first half. Moreover, we organized all accent annotations into a database, and with the help of a phonetically-trained Tokyo Japanese native speaker, we have begin processing the two research assistants' memos and resolving any conflicting annotations.
When the full database was input to Chamame, we noticed a non-trivial number of parsing errors. Given its size, systematically detecting and fixing these in the entire database would be an enormous task, and so we turned attention to the Dialogues portion of the dataset, which contains a representative sample of the grammar points covered in Genki. This dataset is now fully cleaned up, and learning simulations can be run.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
The original goal for the second year was to get to the point where learning simulations could be run, and indeed this goal has been met.
|
| Strategy for Future Research Activity |
Since this grant has been extended into a third year, it has become possible to refine the analysis further. In particular, it will require a bit of experimentation to fully grasp the four-types of accent rules (C, F, M, and P) output by Chamame and how they should be implemented in the learning simulations.
Accordingly, in the first half of the Reiwa 2 year, we will experiment with various types of input data in order to systematically document how the various kinds of accent rule work. Toward this end, additional data will be used from the appendix of the Shinmeikai Japanese Accent Dictionary (2nd edition), which documents numerous accent rules in great detail. Equipped with this understanding, we will be able to run more refined simulations in the second half of the year.
|
| Causes of Carryover |
At the end of the second year, two conference presentations about this research were accepted - one at the American Association of Teachers of Japanese (as indicated in the original grant application) and the other at a phonology conference in Japan. However, due to COVID-19, both conferences were cancelled, although the former offered a portal for sharing presentation materials.
Thus, the remaining funds will be spent during the Reiwa 2 year to disseminate the results of the research at conferences, as originally planned.
|
