Papers by Kalvin Chang

9 papers
WikiHan: A New Comparative Dataset for Chinese Languages (2022.coling-1)

Copied to clipboard

Challenge: Currently, there are 1.3 billion speakers of Sinitic varieties, making the family one of the largest in terms of speaker count.
Approach: They have collected a single constituent and structured form of Chinese varieties for comparative linguistics and Chinese NLP.
Outcome: The proposed dataset contains 67,943 entries across 8 varieties and Middle Chinese . it achieves 54.11% accuracy and 17.69% error rate on a protoform reconstruction task .
POWSM: A Phonetic Open Whisper-Style Speech Foundation Model (2026.acl-long)

Copied to clipboard

Challenge: Phone-level modeling of speech is a common approach to speech recognition, but it relies on task-specific architectures and datasets.
Approach: They propose a phonetic framework capable of performing multiple phone-related tasks . they propose 'Phonetic Open Whisper-style Speech Model' that can perform these tasks together .
Outcome: The proposed model outperforms or matches specialized PR models of similar size while supporting G2P, P2G, and ASR.
Linear Script Representations in Speech Foundation Models Enable Zero-Shot Transliteration (2026.findings-acl)

Copied to clipboard

Challenge: We show that script information is linearly encoded in the activation space of multilingual speech models . modifying activations at inference time induces script change even in unconventional pairings .
Approach: They propose to add script vectors to activations at test time to induce script change . they also show that script information is linearly encoded in the activation space of multilingual speech models .
Outcome: The proposed approach can induce script change even in unconventional language-script pairings.
Programming by Example meets Historical Linguistics: A Large Language Model Based Approach to Sound Law Induction (2025.acl-long)

Copied to clipboard

Challenge: Historical linguists have written programs that convert reconstructed words into their attested descendants via ordered string rewrite functions.
Approach: They propose to use a model to generate a "similar distribution" for sound law induction . they propose four kinds of methods with varying amounts of inductive bias to investigate best performance .
Outcome: The proposed model shows that it can be fine tuned with training data and evaluation data.
Phonotactic Complexity across Dialects (2024.lrec-main)

Copied to clipboard

Challenge: Recent studies show a moderate negative correlation between phonotactic complexity and word length in 106 languages.
Approach: They propose to use a phone-level language model to measure phonotactic complexity . they find a tradeoff between word length and phonomactic complex .
Outcome: The proposed model shows that low phonotactic complexity dialects concentrate around capital regions.
PRiSM: Benchmarking Phone Realization in Speech Models (2026.acl-long)

Copied to clipboard

Challenge: Existing evaluations of phone recognition systems only measure surface-level transcription accuracy.
Approach: They propose to standardize transcription-based evaluation and assess downstream utility in clinical, educational, and multilingual settings with transcription and representation probes.
Outcome: The proposed system outperforms LALMs in clinical, educational, and multilingual settings.
Leveraging Allophony in Self-Supervised Speech Models for Atypical Pronunciation Assessment (2025.naacl-long)

Copied to clipboard

Challenge: Recent phoneme classifiers treat allophonic variation as a single phoneme . atypical pronunciation assessment requires distinguishing between a typical and asymmetric pronunciations .
Approach: They propose a new approach that leverages Gaussian mixture models to model phoneme distributions with multiple subclusters.
Outcome: The proposed approach achieves state-of-the-art across dysarthric and non-native speech datasets.
PWESuite: Phonetic Word Embeddings and Tasks They Facilitate (2024.lrec-main)

Copied to clipboard

Challenge: Existing word embedding methods overlook phonetic information that is crucial for many tasks.
Approach: They propose three methods that use articulatory features to build phonetically informed word embeddings.
Outcome: The proposed methods improve word retrieval and correlation with sound similarity and on rhyme and cognate detection tasks.
Transformed Protoform Reconstruction (2023.acl-short)

Copied to clipboard

Challenge: Historical linguists reconstruct proto-languages by identifying systematic sound changes that can be inferred from correspondences between attested daughter languages.
Approach: They propose to update their Latin protoform reconstruction model with the Transformer . romance data of 8,000 cognates spanning 5 languages and Chinese dataset are outperformed .
Outcome: The proposed model outperforms previous models on Romance and Chinese datasets.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations