Development of Accent-Based Automatic Speech Recognition (ASR) for the Kashmiri Language

Project Goal & Challenge

This project aims to develop an advanced **Accent-Based Automatic Speech Recognition (ASR)** system tailored for the **Kashmiri language**. The core challenge addressed is the significant phonetic and accentual diversity across different regions where Kashmiri is spoken. The research focuses on capturing and modelling these accent variations to build a robust, dialect-aware ASR system.

The system will be capable of accurately transcribing spontaneous Kashmiri speech irrespective of **regional accent differences**, moving beyond standard dialect recognition to full transcription capability.

Regional Dialect Focus

The research specifically models accent variations primarily from three major dialect groups, each with distinct pronunciation patterns, vocabulary influences, and phonological structures:

Methodology & Technology

The project involves several key technical components:

• Comprehensive Accented Speech Dataset Creation:Large-scale audio recordings from native speakers representing the regional dialects are collected.
• Data Annotation: Detailed phoneme-level and accent-specific annotations are performed on the collected dataset.
• Acoustic Modeling: Development of specialized acoustic models to handle phonetic variations.
• Language Model Development: Creation of robust language models using advanced deep neural architectures.
• Model Architectures:Utilizing modern techniques including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and transformer-based models for state-of-the-art ASR performance.

Societal Impact & Future Applications

The system will serve as a foundational resource for building key applications for the Kashmiri community, contributing significantly to digital accessibility and language preservation:

• Kashmiri voice assistants and intelligent agents.
• Accurate speech-to-text services for media and journalism.
• Educational tools for language learning.
• E-governance applications to improve public service access.
• Advancing AI technologies for low-resource and underrepresented languages in India.