Nourchene Ferchichi
Johns Hopkins University · United States
Johns Hopkins researcher building full-duplex speech-to-speech models with native tool calling and streaming ASR
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
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 volume and authorship-role metrics are not yet computable from the current OpenAlex snapshot; this axis would measure total works, per-year publication counts, and first/corresponding-author share across venues.
Impact—
Citation-based impact cannot be assessed yet: available indicators (h-index 1, i10-index 1) rest on very thin coverage. This axis would measure citation counts, h-index trajectory, and per-paper influence over time.
Focus100/100
Focus scores 100/100: the field distribution contains a single field — speech processing and conversational AI — and the topic profile reinforces it, with Speech Recognition and Synthesis (4 works), Speech and Audio Processing (4), and Speech and dialogue systems (3) dominating, while Natural Language Processing Techniques (2) and Topic Modeling (1) stay peripheral. Both Evolve-reviewed 2026 papers attack the same full-duplex speech-to-speech problem.
Collaboration84/100
Collaboration is the profile's strongest axis (84/100): the OpenAlex record links 67 distinct coauthors, and the reviewed papers sit in large multi-author teams where Ferchichi appears as a middle author rather than corresponding author. Recurring collaborators include Edresson Casanova, Jagadeesh Balam, Elena Rastorgueva, Travis Bartley, and Zhehuai Chen, indicating stable work within a concentrated engineering group rather than one-off partnerships.
Activity47/100
Activity scores 47/100, reflecting a sparse but recent record: counts by year show a single indexed work in 2022 followed by 4 works in 2026, with 4 works inside the last three years. The 2026 cluster — including both Evolve-reviewed arXiv papers — suggests a current intensive phase, but the multi-year gap leaves the long-term publication cadence unestablished.
Reproducibility55/100
Reproducibility averages 55/100 across the 2 reviewed papers, both rated medium readiness. Evolve's reviews flag concrete caveats: the VoiceChat model shows weak argument accuracy (42.2%, Pass@1 33.0%) with design claims left untested, and the streaming transcription paper reports 10.21% WER on thin statistics. Open release of a full-duplex S2S model is a positive signal for replication.
Research summary
Nourchene Ferchichi is a speech and conversational AI researcher at Johns Hopkins University whose profile is name-resolved from OpenAlex bibliographic records. The indexed record is early-stage and highly concentrated: topic mass sits in speech recognition and synthesis (4 works), speech and audio processing (4), and speech and dialogue systems (3), with smaller exposure to natural language processing techniques (2) and topic modeling (1). Activity clusters sharply in 2026, when 4 works appear against a single earlier record in 2022, and recent three-year output totals 4 works. The Evolve-reviewed output centers on full-duplex speech-to-speech modeling: an open full-duplex S2S model with native tool calling, and a parallel streaming ASR head that yields in-model user transcription at a reported 10.21% WER. Both papers carry medium reproducibility readiness, averaging 55 across the 2 reviewed papers. Authorship is consistently middle-position and non-corresponding, consistent with contribution inside a large team — 67 distinct coauthors appear in the record. Citation indicators (h-index 1, i10-index 1) are not yet meaningful, so output and impact axes remain unscored.
Collaboration pattern
The record shows large-team collaboration: 67 distinct coauthors overall, with both reviewed 2026 papers written in sizeable multi-author teams where Ferchichi holds middle, non-corresponding positions. Recurring names — Edresson Casanova, Elena Rastorgueva, Jagadeesh Balam, Travis Bartley, Sanjay Chauhan, Chen Chen, Zhehuai Chen, Zijia Chen — point to a stable core group, consistent with the high collaboration score (84/100).
Research keywords
Highlights
First open full-duplex speech-to-speech model with native tool calling
Streaming in-model user transcription at 10.21% WER
A sharply focused 2026 research push
The OpenAlex record shows 4 works in 2026 against 1 in 2022, all concentrated in speech processing and conversational AI, with the two Evolve-reviewed papers attacking the same full-duplex speech-to-speech problem from modeling and transcription angles.
Timeline
- 2022First indexed work appears in the OpenAlex record (1 publication in 2022).
- 2026NemotronLabs VoiceChat, an open full-duplex speech-to-speech model with native tool calling, is released on arXiv.
- 2026A parallel ASR head brings streaming user transcription to full-duplex S2S models at a reported 10.21% WER (arXiv).
- 2026Output clusters in a single field: 4 works in 2026 in speech processing and conversational AI, within a 67-coauthor network.
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 indicators are computed from OpenAlex only and may undercount; no ORCID was available for identity verification.
- Output and impact axes are unscored: available citation indicators (h-index 1, i10-index 1) are too thin to support a rating.
- Qualitative assessments rest on only 2 Evolve-reviewed 2026 arXiv preprints; the reproducibility mean covers just those 2 papers.
- "Nourchene Ferchichi" is name-resolved from public bibliographic data; the 2022 record and the 2026 cluster are assumed to belong to the same person.