Swecha Gontuka - 2023 Hackathon Problem Statements
Swecha Gontuka Project Modules
Let us pick the problem statement and work on it!
- Building Telugu Speech-to-text chatbots using Swecha Telugu Speech-To-Text Model (https://huggingface.co/swechafsmi/whisper-small-te-146h).
- Swecha Corpus Sentence Collector WebApp - Offline.
- To build the Swecha Gontka Corpus collection App for collecting the data in offline mode.
- Swecha Gontuka Leader Dashboard for representing the statistics of the Voice Data corpus collection, Validation and contributions.
- Benchmarking Swecha Gontuka against multiple datasets/tasks and metrics(can specify what datasets and metrics)
- Creating a dashboard to display these metrics
- Developing an API to convert input speech to text using Swecha Gontuka (this helps in processing large batches of input or integration with chatbots)
- Building text-to-speech web/mobile interface for Telugu.
- An existing ML model can be used in the backend.
- Develop an alternative to Google speech-to-text feature for Telugu on smartphones using Swecha Gontuka.
- Use live recording or pre-recording audio file as input.
- Develop an app to collect conversational speech data with record, translate, and submit functions.
- Takes a recording as input and converts it to text using the Swecha Telugu Speech-to-Text Model(https://huggingface.co/swechafsmi/whisper-small-te-146h)
- Submits both voice and text as a form to a database.
- Develop a web app like https://swecha.org/input to convert Telugu speech to text using Swecha Gontuka.
- Feature: We should be able to collect all evaluation data given by users to Swecha Gontuka (with permission).
- Develop a dashboard for querying in Telugu voice on search engines - using the Swecha Telugu Speech-to-text (https://huggingface.co/swechafsmi/whisper-small-te-146h).
- Should take Telugu voice as input and convert it to text using the Swecha Voice model
- It should display a list of relevant search engine results.
Have an idea? Propose through comments!