Qingxiang Guo

—

Unclaimed · AI-structured profileORCID OpenAlex

Early-career full-duplex spoken dialogue researcher building turn-taking benchmarks and synthesized dialogue datasets

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

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—

Output is not scored yet: the snapshot records only works from 2026 (5 in the last three years) with no year-by-year publication history, so career-wide production volume and venue mix cannot be measured.

Impact—

Impact is not scored yet: citation indicators are unavailable or recorded as zero (h-index 0, i10-index 0), so citations, influence on later work, and downstream adoption of the datasets and benchmarks cannot yet be assessed.

Focus0/100

Focus scores 0 on the radar, driven by a field distribution split 50/50 between two overlapping labels ('speech processing / spoken dialogue systems' and 'speech and natural language processing (spoken dialogue systems)'). Beneath that labeling artifact, topic counts are tightly clustered: speech recognition and synthesis (5), speech and audio processing (4), speech and dialogue systems (4), versus single counts for discourse/communication strategies and emotion recognition. Substantively, the work sits in one narrow area: full-duplex dialogue.

Collaboration60/100

Collaboration scores 60. The snapshot counts 19 distinct coauthors across a very small set of works (5 in 2026), indicating a broad but shallow network so far. Recurring partners include Wenke Fan, Yingxin Shang, Hongwei Cai, Weixu Wang, Shuofeng Zhao, Dawei Yang, and Zhiyang Zhou. The two reviewed papers show role fluidity: first author on DuplexDrama, middle author on ECHO, with no corresponding-author record yet.

Activity52/100

Activity scores 52. All recorded output falls in 2026, with 5 works in the last three years — a burst-like recent start rather than a sustained multi-year cadence. Both catalog papers appeared on arXiv in 2026, and the record contains no earlier publications, so the score reflects a newly active profile whose consistency cannot yet be judged over time.

Reproducibility38/100

Reproducibility scores 38 (reproMean 38 across 2 scored papers). Evolve rates DuplexDrama 'low' readiness, noting a solid synthesis pipeline but no human perceptual validation; ECHO rates 'medium', with a matched-insertion design offset by label-dependent interruption acoustic processing and a missing gain-free control. Releasing datasets, code, and validation protocols would be the clearest lever to raise this score.

Research summary

Qingxiang Guo is an early-career researcher working on speech processing and spoken dialogue systems, with a focus on full-duplex dialogue — settings where spoken agents must listen and speak simultaneously. The bibliometric snapshot (OpenAlex identity, ORCID 0009-0009-5544-4067) records 5 works in 2026, all within the last three years, distributed across topics in speech recognition and synthesis (5 tagged works), speech and audio processing (4), and speech and dialogue systems (4), with single counts in language/discourse/communication strategies and emotion recognition. Two reviewed papers anchor the catalog: DuplexDrama (2026, first author), claimed as the first 800-hour TTS-synthesized full-duplex dialogue dataset covering personas, behaviors, expressive speech, and sound events, and ECHO (2026, middle author), a matched-contrast benchmark exposing YIELD bias in turn-taking. Evolve's reviews credit DuplexDrama's construction pipeline while flagging the absence of human perceptual validation, and credit ECHO's paired design while questioning its label-dependent interruption acoustic processing. Collaboration is broad relative to output: 19 distinct coauthors, with recurring partners including Wenke Fan, Yingxin Shang, Hongwei Cai, and Weixu Wang. Citation indicators (h-index, i10-index) are recorded as zero and the output and impact radar axes are treated as missing, consistent with a profile that has only just begun to publish.

Collaboration pattern

Guo's network is broad for an early-stage profile: 19 distinct coauthors across 5 recorded 2026 works. Frequent partners include Wenke Fan, Yingxin Shang, Hongwei Cai, Weixu Wang, Shuofeng Zhao, Dawei Yang, Zhiyang Zhou, Zhou Wang, Lin Yang, Shuran Zhou, and Yang Song, suggesting a stable core team behind the full-duplex dialogue line. Author roles alternate between first author (DuplexDrama) and middle author (ECHO), with no corresponding-author credit yet recorded.

Research keywords

full-duplex spoken dialogueturn-taking modelinginterruption detectionbackchannelbenchmark constructionminimal-pair evaluationspoken dialogue datasetsTTS synthesisexpressive speechsound eventsdialogue systems evaluationspeech recognition and synthesis

Highlights

First-author 800-hour synthesized full-duplex dialogue dataset
DuplexDrama (2026, first author) claims to be the first 800-hour TTS-synthesized full-duplex dialogue dataset spanning personas, full-duplex behaviors, expressive speech, and sound events; Evolve credits the pipeline but flags the lack of human perceptual validation (repro readiness: low).
Matched-contrast benchmark exposing YIELD bias in turn-taking
ECHO (2026, middle author) builds a matched-insertion paired benchmark for context-sensitive turn-taking in full-duplex dialogue; Evolve credits the design for exposing YIELD bias but notes that label-dependent interruption acoustic processing and the absence of a gain-free or waveform-reuse control weaken the central claim (repro readiness: medium).
A concentrated 2026 burst in full-duplex dialogue research
The OpenAlex snapshot records 5 works in 2026 (all within the last three years) with 19 distinct coauthors, concentrated in speech recognition and synthesis (5), speech and audio processing (4), and speech and dialogue systems (4) — an early, fast-moving profile built around dataset and benchmark artifacts.

Timeline

2026Released DuplexDrama on arXiv as first author: a claimed first 800-hour TTS-synthesized full-duplex dialogue dataset covering personas, full-duplex behaviors, expressive speech, and sound events.
2026Co-authored ECHO on arXiv as a middle author: a matched-contrast benchmark exposing YIELD bias in full-duplex turn-taking.
2026Recorded 5 works in the OpenAlex year count, spanning the topics speech recognition and synthesis, speech and audio processing, and speech and dialogue systems.

Papers in the Evolve catalog

Frequent co-authors

Wenke Fan×2 Yingxin Shang×2 Hongwei Cai×2 Weixu Wang×2 Shuofeng Zhao×1 Dawei Yang×1 Zhiyang Zhou×1 Zhou Wang×1 Lin Yang×1 Shuran Zhou×1 Yang Song×1 Shuofeng Zhao×1

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 indicators are computed from OpenAlex data only; institution and country are not recorded in the snapshot.
Output and impact radar axes are missing: h-index and i10-index are recorded as 0 and no citation history is available, so no impact claim is made in this profile.
Only 2 papers carry Evolve reproducibility verdicts; the other recorded 2026 works are unscored, so the reproducibility mean (38) rests on a small sample.
Identity is resolved from public bibliographic records linked to ORCID 0009-0009-5544-4067; the profile reflects only what those records capture.