| Project/Area Number |
19H04215
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 62020:Web informatics and service informatics-related
|
|---|
| Research Institution | University of Tsukuba |
|---|
Principal Investigator |
于 海涛 筑波大学, 図書館情報メディア系, 准教授 (30751052)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
吉川 正俊 大阪成蹊大学, データサイエンス学部, 教授 (30182736)
康 シン 徳島大学, 大学院社会産業理工学研究部(理工学域), 助教 (80777350)
Adam Jatowt 京都大学, 情報学研究科, 特定准教授 (00415861)
|
|---|
| Project Period (FY) |
2019-04-01 – 2024-03-31
|
| Project Status |
Completed (Fiscal Year 2023)
|
| Budget Amount *help |
¥15,860,000 (Direct Cost: ¥12,200,000、Indirect Cost: ¥3,660,000)
Fiscal Year 2023: ¥2,990,000 (Direct Cost: ¥2,300,000、Indirect Cost: ¥690,000)
Fiscal Year 2022: ¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
Fiscal Year 2021: ¥2,340,000 (Direct Cost: ¥1,800,000、Indirect Cost: ¥540,000)
Fiscal Year 2020: ¥3,640,000 (Direct Cost: ¥2,800,000、Indirect Cost: ¥840,000)
Fiscal Year 2019: ¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
|
|---|
| Keywords | Federated Learning / Neural Tree Ensembles / Large Language Model / Generative IR / privacy-preserving / federated learning / semantic matching / online learning-to-rank / interactive search / on-device learning / Metric Optimization / Policy Gradient / Differential Privacy / Privacy-preserving IR / Federated Optimization / On-device Learning / policy gradient / learning to rank / probabilistic regression / Semantic Matching / Learning-To-Rank / Metric Optimisation / Privacy-preserving / Semantic matching / Federated learning / On-device intelligence |
|---|
| Outline of Research at the Start |
This project aims to initiate research into privacy-preserving information retrieval (IR) and develop a scalable privacy-preserving IR system. The proposed IR system builds upon a federated paradigm which fuses on-device machine learning, federated optimization and semantic matching.
|
| Outline of Annual Research Achievements |
The main research objective of this fiscal year is to develop and evaluate the proposed federated information retrieval framework that accounts for both privacy preservation and personalization. To this end, we firstly conducted a comprehensive comparison of GBDT models and the newly proposed neural tree ensembles for ranking. Then we evaluated the effectiveness of performing ranking with GBDT-like models in a federated manner. Moreover, we investigated how to use LLMs to enhance ranking performance by ensembling multiple LLMs in a voting manner. Furthermore, we developed a conversational information seeking system that accounts for personalized retrieval and response generation, and the evaluation over the TREC iKAT test collection demonstrates the superiority of the proposed framework.
|
| Research Progress Status |
令和5年度が最終年度であるため、記入しない。
|
| Strategy for Future Research Activity |
令和5年度が最終年度であるため、記入しない。
|
