← Work

2025 · in revision · cross-modal knowledge distillation · Hindi, Bengali, Marathi, English

Audio-informed multilingual disfluency correction

A three-stage cross-modal knowledge distillation framework that corrects speech disfluencies in Hindi, Bengali, Marathi and English by giving a text model access to what the audio knew.

Measurements
Languages
4
Recordings collected
2,166
Base models
3B × 2
Adaptation
LoRA

समस्या

Problem

Spontaneous speech is full of repairs, restarts and filled pauses, and a transcript records the words while discarding the evidence that a repair happened at all. Working from text alone, a correction model has to guess whether a repeated phrase is emphasis or a false start.

Indian languages make the data problem acute: there was no multilingual Indian disfluency corpus to train or evaluate on, so the framework and the dataset had to be built together.

निर्णय

Decision

The framework distils acoustic information into a text model in three stages, so the corrector benefits from audio cues at training time without needing the audio encoder at inference.

Adaptation used parameter-efficient fine-tuning — LoRA adapters over Qwen2.5-3B and Llama-3.2-3B — rather than full fine-tuning of either base model. With four languages and a small corpus, adapter-level adaptation was the difference between one experiment and a usable matrix of them.

Alongside it we constructed a multilingual Indian speech disfluency dataset of 2,166 spontaneous recordings across Hindi, Bengali, Marathi and English.

परिणाम

Result

The distilled models improved BLEU and chrF2 over zero-shot baselines across the four languages. The paper is in revision, and the dataset is the reusable part: it is the evaluation set the next model gets measured on.

प्रणाली

Stack
Qwen2.5-3B · Llama-3.2-3B · PEFT (LoRA) · Hugging Face Transformers · BLEU / chrF2