Mukher Kotha

Overview

Mukher Kotha is a platform which's goal is to revolutionize and transforming the way native speakers of Bengali language interact with their devices. It's mission is to create platforms and apps that offer delightful Bengali-first experiences for users. Mukher Kotha began it's journey with the development of a powerful Natural Language Processing (NLP) platform, designed to collect, process, and generate labeled audio and transcriptions for training machine learning models.

Since inception, Mukher Kotha has been used for collecting and processing vast volumes of labeled audio data, catering to both general and niche-specific use cases. As a result, our platform now empowers applications with Automatic Speech Recognition (ASR) and advanced voice search capabilities, revolutionizing the Bengali language technology landscape.

How the voice data is processed

Collection of Voice Data
We have gathered an extensive collection of local dialects primarily from rural regions in Bangladesh. The recordings were made by taking proper consent from the speakers.

Transcription
The segmented audio undergoes transcription and subsequent translation into standard Bengali

Segmentation of Audio
The collected audio has been segmented, with each sentence uttered by individual speakers accurately marked. Our system incorporates both manual and automatic segmentation capabilities

ML Model Training
The machine learning model is trained using all the processed voice and text data, enabling it to develop the capability of serving our user applications

Key Features

Bengali-First Approach
The platform's primary focus is to revolutionize and transform the interaction between native Bengali speakers and their devices, ensuring a Bengali-first experience for users.

Labeled Audio Collection
The platform has been actively used for collecting and processing extensive volumes of labeled audio data, catering to both general and niche-specific use cases.

Advanced Bengali Voice Search
The platform offers advanced voice search functionalities, revolutionizing the way Bengali speakers interact with technology and making information retrieval faster and more convenient.

NLP Platform
Mukher Kotha boasts a powerful NLP platform designed to efficiently collect, process, and generate labeled audio and transcriptions, forming the foundation for training machine learning models.

Automatic Speech Recognition
Mukher Kotha empowers applications with ASR capabilities, allowing seamless conversion of spoken Bengali language into text format, enabling voice-based interactions.