Andreas Stolcke

—

Unclaimed · AI-structured profileOpenAlex

Speech and conversational AI researcher working on SpeechLLM emotion recognition and end-of-turn detection in dialogue

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-16) 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: this axis normally measures publication volume and productivity (total works, outputs per year). The current OpenAlex record for this name is sparse and fragmented across years, so no reliable output estimate can be reported yet.

Impact—

Not yet scored: this axis would capture citation-based influence such as citation counts, h-index trajectory, and field-weighted impact. Citation aggregates beyond h-index (5) and i10-index (4) are absent from the snapshot, so no impact estimate is reported.

Focus0/100

Focus scores 0 in the current radar, the lowest band, suggesting output spread thinly across topics. The topic mix tempers that reading: 23 works in speech recognition and synthesis, 18 in speech and audio processing, 16 in speech and dialogue systems, versus 5 in topic modeling and 4 in NLP techniques — speech-dominant, with only a small general-NLP fringe.

Collaboration44/100

Collaboration scores 44, a mid-range value. The recent record shows 19 works in the last three years with 8 distinct coauthors, and the profile lists eight recurring partners including Kadri Hacioglu, Petr Motlicek, and Esa'u Villatoro-Tello. Both cataloged 2026 papers are multi-author efforts on which this researcher appears as last author rather than corresponding, consistent with a senior-collaborator role.

Activity87/100

Activity scores 87, a high value. The yearly record shows a clear recent surge: 8 works in 2025 and 10 in 2026, against single works in scattered earlier years (1998, 2001, 2002, 2004, 2022, 2024), and 19 works within the last three years. This marks an intensive current publishing phase after a long quiet stretch in the indexed record.

Reproducibility38/100

Reproducibility scores 38, averaged over 2 reviewed papers (reproMean 38, reproPapers 2). The two cataloged 2026 arXiv papers sit at medium and low readiness respectively: the SpeechLLM emotion-recognition study uses the public IEMOCAP corpus but validates on a single corpus without significance tests, while the end-of-turn study relies on proprietary single-domain data that blocks independent replication.

Research summary

Andreas Stolcke is a researcher working in speech processing and conversational AI, with a profile name-resolved from OpenAlex bibliographic data. The indexed record spans scattered early works (1998–2004) and a concentrated recent phase: 19 works in the last three years, with 8 in 2025 and 10 in 2026. Topic modeling of the record places the core in speech recognition and synthesis (23 works), speech and audio processing (18), and speech and dialogue systems (16), with smaller footprints in topic modeling (5) and NLP techniques (4); the field split is even between conversational-AI speech processing and speech/affective computing within NLP. Evolve has reviewed two 2026 arXiv papers on which this researcher appears as last author: one adapting SpeechLLMs with a discriminative classification head for speech emotion recognition on IEMOCAP, and an ablation showing acoustic and prosodic cues outperform text-inclusive features for end-of-turn detection. Reviewers noted honest limitations in both — single-corpus evaluation without significance tests in the first, proprietary single-domain data in the second. Collaboration metrics show 8 distinct coauthors and eight recurring partners, with a mid-range collaboration score (44) and high activity (87); focus (0) and reproducibility (38) scores are low, and output and impact dimensions are not yet scored for lack of data.

Collaboration pattern

Collaboration scores 44 (mid-range). The recent record lists 8 distinct coauthors across 19 works in the last three years, with eight recurring partners: Kadri Hacioglu, Rini Sharon, Manickavela A, Hasindri Watawana, Sergio Burdisso, Esa'u Villatoro-Tello, Manjunath K E, and Petr Motlicek. Both Evolve-reviewed 2026 papers place this researcher as last (non-corresponding) author, a pattern consistent with a senior group member supervising or advising rather than day-to-day leading.

Research keywords

speech emotion recognitionSpeechLLMsend-of-turn detectionturn-takingprosodyconversational AIspeech recognition and synthesisspeech and audio processingspoken dialogue systemsmultimodal fusiontopic modelingparameter-efficient adaptation (LoRA)

Highlights

Adapting SpeechLLMs for emotion recognition in token space
A 2026 arXiv study (last-author position) applies a discriminative classification head on frozen SpeechLLM hidden states, with LoRA adaptation, for speech emotion recognition on IEMOCAP. Evolve's review finds consistent small gains and interpretability value, while noting the evaluation is limited to a single corpus without significance testing. Reproducibility readiness: medium.
What aspects of speech drive end-of-turn detection
A second 2026 arXiv paper (last-author position) runs a clean seven-subset ablation of the acoustic, prosodic, and textual cues behind end-of-turn detection, finding that acoustic and prosody-driven models beat text-inclusive variants. The verdict is convincing in-domain, but the proprietary, single-domain dataset limits generality; reproducibility readiness: low.
A concentrated recent publishing phase
The indexed record shows 19 works in the last three years, including 8 works in 2025 and 10 in 2026 — a sharp rise after isolated earlier entries (1998–2004, 2022, 2024). Activity scores 87, placing this researcher in the high-output band for the current period.

Timeline

1998Earliest indexed work in the record, with a single publication.
2004Last of a sparse early run of single-work years (1998, 2001, 2002, 2004) before a long gap.
2022The indexed record resumes with one publication after an 18-year gap in the yearly counts.
2025Activity accelerates sharply: 8 indexed works in a single year.
2026The most productive year so far, with 10 works, including the two Evolve-reviewed arXiv papers on SpeechLLM-based emotion recognition and end-of-turn detection, both with this researcher as last author.

Papers in the Evolve catalog

Frequent co-authors

Kadri Hacioglu×2 Rini Sharon×1 Manickavela A×1 Hasindri Watawana×1 Sergio Burdisso×1 Esa\'u Villatoro-Tello×1 Manjunath K E×1 Petr Motlicek×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.

All bibliometrics come from an OpenAlex name-resolved record; no institution, country, or ORCID was available to confirm identity, and the low h-index (5) and i10-index (4) alongside sparse pre-2022 counts suggest the record may not capture this name's full publication history.
Citation-based dimensions (output, impact) are unscored because citation aggregates are absent from the current snapshot.
Only two papers carry Evolve reviews; topic and field distributions beyond these come from OpenAlex metadata.
Yearly counts before 2022 are single works, so long-term trend statements are limited.