Weixu Wang — spoken dialogue systems researcher: full-duplex turn-taking benchmarks and synthesized dialogue datasets
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
18
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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 scored: the radar lacks the underlying data for this axis. It would measure publication volume and productivity over time; the snapshot offers only a partial signal, with 27 works in the most recent three years and no full output metric.
Impact—
Not scored: the impact axis has insufficient underlying data in this radar. It would measure citation counts, citation velocity, and field-normalized influence of the work; no such metrics are currently available for scoring.
Focus0/100
Scores 0/100 despite an even 50/50 fieldDist split between two adjacent speech labels ('speech processing / spoken dialogue systems' and 'speech and natural language processing (spoken dialogue systems)'). The OpenAlex topic list mixes in plant molecular biology (counts 12, 12, 7) and traffic themes (counts 10, 9), indicating strong cross-domain dispersion or profile noise, so topical concentration cannot be established from this snapshot.
Collaboration60/100
Scores 60/100. The profile records 19 distinct coauthors, and the frequent-coauthor list includes Qingxiang Guo, Wenke Fan, Yingxin Shang, Hongwei Cai, Shuofeng Zhao, Dawei Yang, Zhiyang Zhou, Zhou Wang, Lin Yang, Shuran Zhou, and Yang Song. Both reviewed 2026 papers place Wang in a middle author position rather than corresponding author, consistent with contribution within multi-author teams.
Activity97/100
The strongest axis at 97/100. The countsByYear series spans 2002–2026 with a marked acceleration after 2020: 15 works in 2020, 9 in 2022, 13 in 2023, 10 in 2024, and 12 in 2026, alongside 27 works in the most recent three-year window — sustained, high-volume recent production.
Reproducibility38/100
Scores 38/100, a reproMean computed across 2 assessed papers. Evolve rates ECHO's reproducibility readiness as medium, while DuplexDrama is rated low — the latter's 800-hour synthesized dataset pipeline lacks human perceptual validation, and ECHO's context-sensitivity claim is weakened by label-dependent interruption acoustic processing and the absence of gain-free or waveform-reuse controls.
Research summary
Weixu Wang is affiliated with Hebei University of Engineering (China) and works in speech and natural language processing, with a concentration on spoken dialogue systems. This profile is name-resolved from public OpenAlex bibliographic data. The bibliometric snapshot reports an h-index of 18, an i10-index of 29, and 27 works over the most recent three years, with activity spanning 2002–2026 and a pronounced surge from 2020 onward (15 works in 2020, 13 in 2023, 12 in 2026) — the basis for an activity dimension score of 97/100. Evolve has reviewed two 2026 arXiv preprints on which Wang is a middle author: ECHO, a matched-contrast benchmark for context-sensitive turn-taking in full-duplex dialogue that exposes YIELD bias, and DuplexDrama, a claimed 800-hour TTS-synthesized full-duplex dialogue dataset covering personas, behaviors, expressive speech, and sound events. Recorded collaboration spans 19 distinct coauthors (collaboration score 60/100). The reproducibility dimension stands at 38/100 from two assessed papers, while the radar's focus axis scores 0 and its OpenAlex topic mixture spans speech alongside plant molecular biology and traffic-safety themes — a pattern consistent with name-resolution noise, so topical boundaries should be read cautiously. Output and impact axes lack sufficient data and are not scored.
Collaboration pattern
Wang's recorded network spans 19 distinct coauthors, with recurring partners including Qingxiang Guo, Wenke Fan, Yingxin Shang, Hongwei Cai, Shuofeng Zhao, Dawei Yang, Zhiyang Zhou, Zhou Wang, Lin Yang, Shuran Zhou, and Yang Song. In both Evolve-reviewed 2026 papers Wang appears as a middle author rather than corresponding author, suggesting contribution within multi-author teams on benchmark and dataset engineering rather than sole-lead projects. The collaboration dimension scores 60/100.
Research keywords
Highlights
ECHO: a matched-contrast benchmark exposing YIELD bias in full-duplex turn-taking
DuplexDrama: an 800-hour synthesized full-duplex dialogue dataset
Timeline
- 2002Earliest publication year in the OpenAlex countsByYear series (2 works), indicating a bibliographic record reaching back over two decades.
- 2020Publication activity surges to 15 recorded works, the start of a sustained high-output phase in the countsByYear series.
- 2023A second peak year with 13 recorded works, reinforcing the post-2020 acceleration in output.
- 2026Two arXiv preprints enter the Evolve catalog with Wang as a middle author — the ECHO turn-taking benchmark and the DuplexDrama dialogue dataset — in a year with 12 recorded works.
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.
- Bibliometric figures (h-index, i10-index, countsByYear, topics) come from an OpenAlex-derived snapshot; only the two listed 2026 preprints have been quality-reviewed by Evolve.
- Output and impact radar axes are unscored due to insufficient data; no citation-count totals are quoted beyond the provided h-index and i10-index.
- The OpenAlex topic mixture spans plant molecular biology and traffic safety alongside speech fields, so this may be a merged or name-colliding profile; identity is name-resolved from public bibliographic data.
- The reproducibility score of 38 is averaged over only 2 assessed papers, both 2026 arXiv preprints with middle authorship.