Speech AI researcher at Nanjing University: chain-of-thought TTS and non-verbal vocalization ASR.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
13
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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—
The output axis is not scored: this snapshot lacks verified publication-volume evidence such as a complete work count and venue-normalized totals. With fuller bibliographic coverage it would measure cumulative and per-year research output.
Impact—
The impact axis is not scored: citation evidence beyond the recorded h-index of 13 and i10-index of 14 is insufficient here. With complete citation data it would measure influence through citation counts and field-normalized impact.
Focus0/100
The focus axis scores 0, reflecting low topical concentration rather than narrow specialization. The field distribution is split evenly — speech synthesis (text-to-speech) 0.5, speech processing / computational paralinguistics 0.5 — and topic counts spread across five clusters: speech recognition and synthesis (62), speech and audio processing (57), music and audio processing (17), voice and speech disorders (15), and speech and dialogue systems (15).
Collaboration63/100
The collaboration score is 63. The recent three-year window records 42 works with 23 distinct co-authors — active but mid-sized. Recurring co-authors in the source data include Weizhen Bian, Sitong Cheng, Rongxiu Zhong, Jiahao Pan, Boyi Kang, Shilei Zhang, Jinglei Liu, Yue Wang, Junlan Feng, Bei Liu and Wei Xue. On the two reviewed papers this researcher appears as a middle and a last author, consistent with team-based contribution.
Activity100/100
The activity score is 100, the strongest axis. Recorded annual output rose from 2-5 works in 2018-2022 to 10 in 2023 and a peak of 21 in 2024, dipped to 6 in 2025, and reached 15 in 2026 to date — 42 works in the recent three-year window against a lifetime h-index of 13 and i10-index of 14. Output is heavily concentrated in 2023-2026.
Reproducibility55/100
The reproducibility score is 55, a mean over the 2 reviewed papers, both rated medium readiness. The NVVSpeech Track 1 system paper reports rank 4 (63.86) via sqrt and uniform category sampling, but leaves a local-versus-official score gap unexplained with thin statistics; COT-TTS ships a 9M-sample pipeline and an 800-sample benchmark while its chain-of-thought benefit remains untested. Artifacts exist; verification is partial.
Research summary
Liumeng Xue is a speech researcher at Nanjing University whose record is name-resolved from OpenAlex bibliographic data under ORCID 0000-0003-2815-8494. The portfolio is evenly split between speech synthesis (text-to-speech) and speech processing / computational paralinguistics, each 0.5 of the field distribution, with topic mass led by speech recognition and synthesis (62) and speech and audio processing (57). Bibliometric indicators are mid-career: an h-index of 13 and i10-index of 14, against 42 works in the recent three-year window and 23 distinct co-authors. Activity is the standout axis (100): annual output climbed from 2-5 works in 2018-2022 to 10 in 2023, a peak of 21 in 2024, and 15 recorded in 2026. Two 2026 arXiv papers reviewed by Evolve define the current agenda. COT-TTS introduces an audio context-aware text-to-speech task with chain-of-thought reasoning, backed by a 9M-sample pipeline and an 800-sample benchmark, where 0.6B and 1.7B models rival 34-39B cascades on MOS. The NVVSpeech Track 1 system takes rank 4 (63.86) on non-verbal vocalization-aware ASR through sqrt and uniform category sampling with label harmonization. In both, the role is collaborative — middle and last authorship rather than first or corresponding — and Evolve flags medium reproducibility readiness with unverified claims (the untested CoT benefit; an unexplained local-versus-official score gap).
Collaboration pattern
The collaboration score of 63 sits alongside 23 distinct co-authors across 42 recent three-year works — an active but mid-sized network. Recurring names in the source data include Weizhen Bian, Sitong Cheng, Rongxiu Zhong, Jiahao Pan, Boyi Kang, Shilei Zhang, Jinglei Liu, Yue Wang, Junlan Feng, Bei Liu and Wei Xue. Authorship positions on the two reviewed papers (middle and last) point to contributor roles within larger speech teams rather than solo-led projects.
Research keywords
Highlights
Chain-of-thought reasoning arrives in text-to-speech
Rank-4 system for non-verbal vocalization-aware ASR
Research output concentrates sharply after 2023
Annual works moved from 2-5 in 2018-2022 to 10 in 2023, a recorded peak of 21 in 2024, 6 in 2025 and 15 already in 2026 — 42 works in the recent three-year window.
Timeline
- 2018Earliest recorded works appear in the OpenAlex record, with 2 publications.
- 2023Output steps up to 10 works, roughly double the 2018-2022 baseline of 2-5 per year.
- 2024Recorded annual peak: 21 publications.
- 202615 works recorded so far, including the two Evolve-reviewed papers: the NVVSpeech Track 1 challenge system and COT-TTS.
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 come from OpenAlex only; identity is name-resolved from public bibliographic data under ORCID 0000-0003-2815-8494, and this name is not rare, so record-level certainty is limited.
- The record lists country 'SG' while the institution is Nanjing University (China) — an apparent metadata inconsistency retained as given.
- The frequent-coauthor list contains the researcher's own name, suggesting a duplicate author record in the source that may distort collaboration counts.
- Citation-level data beyond h-index 13 and i10-index 14 is unavailable, so output and impact axes are left unscored rather than estimated.