Kyoto University speech researcher spanning four decades of speech recognition, dialogue systems, and LLM-based ASR.
3
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
39
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored: the underlying publication-volume fields for this axis are insufficient in the current snapshot. Once populated, it would measure total works, per-year productivity, and the shape of the output trajectory across the 1988-2026 record.
Impact—
Not yet scored: citation-side fields are incomplete in this snapshot, although an h-index of 39 and an i10-index of 182 are recorded. The axis would measure citation counts, citing works, and influence on subsequent speech and dialogue research.
Focus0/100
Scores 0 because the field distribution splits evenly (0.5/0.5) between two adjacent labels: speech processing / natural language processing and computational linguistics (speech and multimodal interaction). The topic counts show the substance is in fact concentrated — Speech Recognition and Synthesis (433), Speech and Dialogue Systems (347), Speech and Audio Processing (307), with NLP Techniques (122) and Phonetics and Phonology (72) as smaller strands — so the zero reflects label-level breadth rather than genuine topic drift.
Collaboration73/100
Scores 73. The recent window records 37 distinct coauthors and a twelve-person frequent-coauthor list that combines a long-standing Japanese core — Hayato Futami, Taiga Mori, Koji Inoue, Mikey Elmers, Divesh Lala — with newer international collaborators such as Bryan Chen Zhengyu Tan, Weihua Zheng, Thong T. Doan, Xiaoyuan Yi, and Jing Yao. This sustained mid-to-large collaboration footprint supports the above-average score.
Activity100/100
Scores 100, the top of the scale. The record shows 94 works in the last three years and continuous annual output from 1988 through 2026, including 28 works in 2024, 30 in 2025, and 36 already in 2026 — the highest yearly count in a record now approaching four decades.
Reproducibility38/100
Scores 38, computed from two assessed papers (reproMean 38, reproPapers 2). In the reviewed catalog, the LLM-merging ASR paper is rated medium reproducibility readiness, while the TEIDAN dialogue corpus is rated low, reflecting its dependence on a large multimodal data release that is not yet confirmed.
Research summary
Tatsuya Kawahara is a researcher at Kyoto University; this profile is name-resolved from public bibliographic data (OpenAlex identity; ORCID 0000-0002-2686-2296). The bibliometric record spans 1988 to 2026, with an h-index of 39 and an i10-index of 182 in the current snapshot. Topic assignment concentrates overwhelmingly on speech: Speech Recognition and Synthesis (433 works), Speech and Dialogue Systems (347), and Speech and Audio Processing (307), with smaller strands in Natural Language Processing Techniques (122) and Phonetics and Phonology Research (72). Activity is exceptional: 94 works in the last three years, with 28 works in 2024, 30 in 2025, and 36 already in 2026 — the highest annual count in the record. Evolve has reviewed three 2026 arXiv papers. Two carry verdicts: a task-arithmetic study merging domain LLM LoRAs into LLM-based ASR (consistent modest gains at zero inference cost, with significance untested), and TEIDAN, a 69-session Japanese/English triadic multiparty dialogue corpus judged well documented and honestly caveated, though its data release is unconfirmed. A third, CultureConverse on culturally grounded multilingual assistance, is not yet reviewed. Across all three, Kawahara appears in senior co-author positions: last on two papers and middle on one. Reproducibility assessment covers two papers with a mean score of 38, while the output and impact axes lack sufficient data to score.
Collaboration pattern
The record lists 37 distinct coauthors in the recent window and a stable recurring core of frequent collaborators — Hayato Futami, Taiga Mori, Koji Inoue, Mikey Elmers, and Divesh Lala — alongside newer international collaborators including Bryan Chen Zhengyu Tan, Weihua Zheng, Thong T. Doan, Xiaoyuan Yi, and Jing Yao. In the three cataloged 2026 papers, Kawahara takes the last author position on two and a middle position on one, a pattern consistent with a senior supervising role within multi-author teams.
Research keywords
Highlights
Merging LLM knowledge into ASR at zero inference cost
TEIDAN: a multilingual multiparty dialogue corpus
CultureConverse: culturally grounded multilingual assistance
A 2026 arXiv preprint (middle author) on a multilingual multi-turn simulation harness for culturally grounded assistance in East and Southeast Asia. It sits at the intersection of the recorded dialogue-systems and NLP topic strands and has not yet been reviewed by Evolve.
A near-four-decade record with 2026 output at its peak
The bibliometric record runs from 1988 to 2026 and counts 36 works already in 2026, the highest annual figure recorded, within 94 works over the last three years and alongside an h-index of 39 and an i10-index of 182.
Timeline
- 2026arXiv preprint on merging domain LLM LoRAs into LLM-based ASR via task arithmetic; last author, medium reproducibility readiness.
- 2026TEIDAN multilingual multiparty dialogue corpus (69 sessions, Japanese/English, triadic setting) appears as an arXiv preprint; last author, low reproducibility readiness pending data release.
- 2026CultureConverse, a multilingual multi-turn simulation harness for culturally grounded assistance in East and Southeast Asia, appears on arXiv; middle author, not yet reviewed.
- 2026Bibliometric snapshot logs 36 works in 2026, the highest annual count since record-keeping began in 1988, within 94 works over the last three years.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometrics are derived from an OpenAlex snapshot only; other databases are not included in this profile.
- The output and impact radar axes are unscored because the underlying per-axis data is insufficient in this snapshot.
- The reproducibility score rests on only two assessed papers (mean 38), too few to generalize across the body of work.
- Affiliation metadata lists Kyoto University with country code SG, an inconsistency left unverified; only three papers have been reviewed in the Evolve catalog so far.