Berlin Chen

National Taiwan Normal University · Taiwan

Unclaimed · AI-structured profileOpenAlex

Low-resource ASR researcher at National Taiwan Normal University focused on Taiwanese Hokkien and Whisper adaptation

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-17) 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—

Publication-output data is insufficient in this snapshot: the OpenAlex identity lists only 7 works, all dated 2026, with no verified career-wide total. This axis would measure cumulative publication volume and long-term productivity across the researcher's full career.

Impact—

Citation-based impact cannot be assessed: this identity shows an h-index and i10-index of 0 across a thin, recent-only slice, with no reliable citation totals. This axis would measure citation counts, h-index, and scholarly influence over time.

Focus100/100

The profile is exceptionally concentrated: OpenAlex assigns a single field — speech processing / automatic speech recognition (computational linguistics) — and all 7 recent works fall in speech topics (Speech Recognition and Synthesis: 7; Speech and Audio Processing: 6; Music and Audio Processing: 3), yielding a focus score of 100.

Collaboration50/100

Collaboration scores mid-range (50) against 11 distinct coauthors in the snapshot. The frequent-coauthor list — Hung-Yang Sung, Chien-Chun Wang, Tien-Hong Lo, Yu-Sheng Tsao, Yung-Chang Hsu, Kuan-Tang Huang, Cheng-Yeh Yang, Hung-Shin Lee, Hsin-Min Wang — shows a stable recurring team, with Chien-Chun Wang appearing twice, consistent with dense in-group co-authorship around a last-author role.

Activity61/100

Activity registers 61, driven by countsByYear showing all 7 works dated 2026 and recent3yWorks equal to 7 — every indexed output sits in the most recent year. Both catalog papers are 2026 preprints, so this reads as an active, ongoing research push rather than a long publication record.

Reproducibility55/100

Reproducibility is scored 55 from a single assessed paper (reproPapers: 1, mean 55). T-SANDHI carries a medium readiness rating: Evolve's verdict calls its results plausible and well ablated, with consistent CER gains up to 12.27% relative, while flagging significance testing and baseline fairness as open questions.

Research summary

Berlin Chen is a researcher at National Taiwan Normal University whose profile is name-resolved from OpenAlex bibliographic data. The indexed snapshot is narrow but remarkably coherent: seven works, all dated 2026, concentrated in speech topics — Speech Recognition and Synthesis (7), Speech and Audio Processing (6) — with smaller footprints in music and audio processing, speech and dialogue systems, and phonetics and phonology. Both papers in the Evolve catalog place Chen in the senior, last-author position on low-resource ASR work. T-SANDHI (2026) proposes a tone sandhi-aware adaptive network with decoupled hybrid injection for Taiwanese Hokkien speech recognition: dual-stream gated adaptation is applied to a frozen Whisper model via parameter-efficient fine-tuning, delivering consistent relative character-error-rate gains of up to 12.27%, while diagnosing that foundation models confuse sandhi tones with rare citation tones. Evolve's verdict finds the results plausible and well ablated, flagging significance testing, baseline fairness, and novelty claims for scrutiny. The companion 2026 paper, Anchoring Speech with Semantics, develops a multimodal adapter mechanism for ASR in low-resource languages. Collaboration is mid-range (score 50; 11 distinct coauthors) around a recurring Taiwan-based team; focus is exceptional (100); activity is 61; reproducibility is 55 from a single assessed paper. Citation-based output and impact cannot be scored because this OpenAlex identity appears to capture only recent works.

Collaboration pattern

Collaboration sits at 50 on 11 distinct coauthors. The recurring circle — Hung-Yang Sung, Chien-Chun Wang, Tien-Hong Lo, Yu-Sheng Tsao, Yung-Chang Hsu, Kuan-Tang Huang, Cheng-Yeh Yang, Hung-Shin Lee, and Hsin-Min Wang — is tightly knit, with Chien-Chun Wang listed twice, pointing to a stable Taiwan-based speech team. Both catalog papers place Chen as last author (not corresponding), a senior supervisory role rather than hands-on first-author work.

Research keywords

automatic speech recognitionlow-resource languagesTaiwanese Hokkientone sandhiWhisper adaptationparameter-efficient fine-tuningspeech recognition and synthesisspeech and audio processingmusic and audio processingspeech and dialogue systemsmultimodal adapter mechanismphonetics and phonology

Highlights

T-SANDHI: tone sandhi-aware adaptation for Taiwanese Hokkien ASR
Last-author 2026 work applying a tone sandhi-aware adaptive network with decoupled hybrid injection to a frozen Whisper model, yielding consistent relative CER gains up to 12.27% on low-resource Taiwanese Hokkien; Evolve finds it plausible and well ablated but flags significance testing, baseline fairness, and novelty claims for scrutiny.
Semantics-anchored multimodal adapters for low-resource ASR
A 2026 companion preprint, again with Chen as last author, proposing a multimodal adapter mechanism that anchors acoustic recognition with semantic signals for automatic speech recognition in low-resource languages; it is not yet reviewed in the Evolve catalog, so no readiness rating exists.
A sharply single-field program in low-resource speech technology
The OpenAlex record assigns exactly one field — speech processing / automatic speech recognition (computational linguistics) — and 7 of 7 recent works cluster in speech topics (Speech Recognition and Synthesis: 7; Speech and Audio Processing: 6), producing a perfect focus score of 100.

Timeline

2026T-SANDHI preprint appears: a tone sandhi-aware adaptive network with decoupled hybrid injection for low-resource Taiwanese Hokkien ASR, with Chen in the last-author position.
2026A second low-resource ASR preprint, Anchoring Speech with Semantics, is released, again listing Chen as last author.
2026All seven OpenAlex-indexed works under this identity are dated 2026, spanning speech recognition and synthesis, speech and audio processing, music, dialogue systems, and phonetics.

Papers in the Evolve catalog

2026-09-17
T-SANDHI: Tone Sandhi-aware Adaptive Network with Decoupled Hybrid Injection for Low-resource Taiwanese Hokkien Speech Recognition

Frequent co-authors

Hung-Yang Sung×1 Chien-Chun Wang×1 Tien-Hong Lo×1 Yu-Sheng Tsao×1 Yung-Chang Hsu×1 Berlin Chen×1 Kuan-Tang Huang×1 Cheng-Yeh Yang×1 Chien-Chun Wang×1 Hung-Shin Lee×1 Hsin-Min Wang×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 a single OpenAlex identity (no ORCID provided); its coverage is thin — h-index and i10-index are 0 and all 7 works are dated 2026 — so productivity and impact are likely undercounted.
'Berlin Chen' is name-resolved from public bibliographic data; other researchers may share this name, and identity consolidation should be verified.
Qualitative assessment rests on only two catalog papers; one carries an Evolve verdict, and the reproducibility score is computed from a single assessed paper.
Output and impact dimensions are unscored (insufficient data), and fields beyond the single indexed OpenAlex field are inferred only from topic labels.