🧑🏻‍🎓 📑 October'20 - April'21. Group uni project. The project topic is Speech-to-Text Assessment Tool. It is a research-type project, most of the documentation is in a private GitLab repository.

Created an ASR (Automatic Speech Recognition) system that takes in individual recordings. Each recording represents a sentence composed of 5-10 English language digits, separated by adequate pauses. The system involves segmenting the sentence using a classifier, differentiating between background and foreground sounds.

Interface for automated transcription and time alignment of conversational interview data

The project,being part of Kagglex BIPOC Mentorship Program final project, aims to train two separate Hindi ASR models using the Facebook Wav2Vec2 (300M parameters) and OpenAI Whisper-Small models, respectively. The goal is to compare their performance, with a target WER of less than 13%, across various Hindi accents and dialects.

Whisper realtime streaming for long speech-to-text transcription and translation.

Test Repo

French Child's Spontaneous Speech Data

Dialect-Pronunciation-Dictionary

This repo contains the code for "Privacy Preserving Data Selection for Bias Mitigation in Speech Models"

An ASR Gradio GUI based project that transcript the audion and provides NLP based analysis.

557-Hours-Kazakh-Spontaneous-Speech-Data

Functionality for speech data processing including time alignment, encoding with speech encoders (tokenizers) and data preprocessing of common datasets

Prototype of an intelligent research agent capable of literature retrieval, summarization, and contextual reasoning — a foundation for scientific automation tools.

The dataset of Korean conversational speech

Text and speech AI playground. The first version was vibe-coded with OpenAI Codex in the browser in ~10 days of 2 h/day sessions.

This project presents Hera, an Operating System level voice recognition package that understands voice commands to perform actions to simplify the user’s workflow. We propose a modernistic way of interacting with Linux systems, where the latency of conventional physical inputs are minimized through the use of natural language speech recognition.

ASR telegram assistant for voice/video messages transcribing