Esa\'u Villatoro-Tello

Idiap Research Institute · Switzerland

Unclaimed · AI-structured profileOpenAlex

Speech and NLP researcher at Idiap working on human-aligned ASR evaluation and speech emotion recognition

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 scoring is unavailable yet: this axis would measure total publication volume, venue quality, and authorship roles over a full career. The current record spans only 13 recent works — too narrow a window for a stable score.

Impact—

Impact scoring is unavailable yet: this axis would measure citations, h-index trajectory, and influence on later work. Existing indicators (h-index 1, i10-index 0) simply reflect how recent this publication burst is, not a measured impact level.

Focus0/100

The radar reports a focus score of 0. The underlying field distribution splits exactly 50/50 between speech recognition / NLP and speech processing / affective computing, and OpenAlex topic assignments spread across five topics, led by speech recognition and synthesis (8 works), speech and audio processing (6), and speech and dialogue systems (6). Topical breadth currently outweighs concentration on a single research line.

Collaboration53/100

Collaboration scores 53. The OpenAlex record lists 13 distinct co-authors over the recent window, and Evolve places this researcher in middle authorship on both reviewed 2026 papers, without corresponding-author roles. The recurring collaborator list — Petr Motlicek (appearing twice), Sergio Burdisso, Kadri Hacioglu, Manjunath K E, Hritika Sharma, and others — points to team-based work within and around Idiap's speech group.

Activity77/100

Activity scores 77, the strongest axis on this profile. OpenAlex counts 13 works in the last three years, and by-year counts show sharp acceleration: 1 work in 2025 rising to 12 in 2026. Both Evolve-reviewed papers appeared on arXiv in 2026, matching this surge and indicating an intense, ongoing publishing phase concentrated in speech technology preprints and venues.

Reproducibility55/100

Reproducibility scores 55, averaged from Evolve's assessments across 2 reviewed papers, both rated medium. The HATS-en evaluation study is judged credible but filtered, while the SpeechLLM emotion study reports consistent small gains on a single corpus (IEMOCAP) without significance tests. Together they suggest solid methodological practice whose main reproducibility risks are data access and narrow external validity.

Research summary

Esa'u Villatoro-Tello is a researcher at the Idiap Research Institute in Switzerland working at the intersection of speech technology and natural language processing. OpenAlex records attribute 13 works to this profile in the last three years — 1 in 2025 and 12 in 2026 — a sharply accelerating cadence, while citation indicators remain early-stage (h-index 1, i10-index 0). Topic assignments concentrate on speech recognition and synthesis (8 works), speech and audio processing (6), and speech and dialogue systems (6), with secondary activity in sentiment analysis and emotion recognition (3 each), matching an even 50/50 field split between speech recognition / NLP and speech processing / affective computing. Evolve has reviewed two 2026 arXiv papers from this profile: a human-aligned evaluation study introducing HATS-en, a resource of 1,000 human-preference triplets for ASR, which finds WER agrees least with human judgments, and a study reading emotions from frozen SpeechLLM token representations via lightweight discriminative heads. Both list the researcher in middle positions, consistent with a collaborative contributor within a team anchored by Petr Motlicek and a network of 13 distinct co-authors.

Collaboration pattern

The record shows a team-embedded pattern: 13 distinct co-authors in the recent OpenAlex window, with middle, non-corresponding authorship on both reviewed 2026 papers. The recurring collaborator list is anchored by Petr Motlicek (listed twice in Evolve's data) and includes Sergio Burdisso, Kadri Hacioglu, Manjunath K E, Hritika Sharma, Alessandra Pinto, Hyunggu Jung, Somang Nam, Hasindri Watawana, and Thibault Bañeras-Roux.

Research keywords

Automatic speech recognitionASR evaluation metricsHuman-aligned evaluationWord Error RateSemantic distance metricsSpeech emotion recognitionSpeechLLMsParameter-efficient adaptationSentiment analysisDialogue systems

Highlights

Human-aligned ASR evaluation: HATS-en and the limits of WER
The 2026 arXiv study introduces HATS-en, a resource of 1,000 human-preference triplets for automatic speech recognition output. Analysis shows WER agrees least with human judgments while CER nearly matches the best semantic metric (SemDist), prompting a rethink of standard ASR evaluation. Evolve rates it credible but filtered, with medium reproducibility.
Reading emotion from frozen SpeechLLM token representations
In a 2026 arXiv study, lightweight discriminative heads and LoRA-style adaptation applied to frozen SpeechLLM hidden states improve speech emotion recognition on IEMOCAP, with consistent small gains and added interpretability. Evolve notes the evidence rests on a single corpus and lacks significance tests; reproducibility is rated medium.
A 2026 publication surge across speech AI
OpenAlex attributes 12 works to this profile in 2026 versus 1 in 2025, with topics spanning speech recognition and synthesis, speech and audio processing, and speech and dialogue systems. This acceleration, alongside 13 recent works and 13 distinct co-authors, marks an intense early phase centered on evaluation and affective speech technology.

Timeline

2025First work in the current OpenAlex window appears: 1 publication recorded in 2025, the baseline before the following year's surge.
2026Contributes to 'Rethinking Human-Aligned Evaluation: An Analysis of Semantic Metrics Beyond WER' on arXiv, introducing HATS-en (1,000 human-preference triplets for ASR evaluation) as a middle author.
2026Contributes to 'Reading Emotions in the Token Space' on arXiv, applying discriminative adaptation of SpeechLLMs to emotion recognition on IEMOCAP, again as a middle author.
2026Output accelerates to 12 works recorded in 2026, with 13 distinct co-authors across the recent window — the most active year on record for this profile.

Papers in the Evolve catalog

Frequent co-authors

Petr Motlicek×2 Hritika Sharma×1 Thibault Bañeras-Roux×1 Alessandra Pinto×1 Petr Motlicek×1 Hyunggu Jung×1 Esaú Villatoro-Tello×1 Somang Nam×1 Hasindri Watawana×1 Sergio Burdisso×1 Manjunath K E×1 Kadri Hacioglu×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.

Bibliometrics (h-index 1, i10-index 0, counts by year, topics) come from an OpenAlex identity resolution; no ORCID was available to confirm the profile.
Citation-impact and output axes are unscored because the underlying data is insufficient; the h-index of 1 reflects a very recent publication window.
Only 2 papers from this profile have been reviewed by Evolve, both 2026 arXiv preprints with middle authorship; career facts beyond these records are unknown.
The name appears in variants ('Esa'u' vs 'Esaú'); identity is resolved from public bibliographic data only.