Deliverable : Convert input voice files to corresponding text AND audio analytics (voice analytics/sentiment - this is NOT the sentiment of converted text, BUT the voice content itself)
Approach
1. Module to read input (audio files from various clouds, local storage) in an automated , incremental manner. Incremental meaning - incrementally based on create/modified date of the file , file name and any other signature - thus avoid getting the duplicate files
2. Module to process the audio files
2.a Analyze speech on dimensions of : Energy, Interruption, Empathy, Participation, Tone, Pace, Anger, Frustration
2.b Convert Audio files to text along with accuracy score
3. UI/UX
- User should be able to upload >=1 audio files thru UI that calls input module
- user should be able to mark which audio files to process (all or none or >=1)
- the output should produce visualization for each selected audio file in that batch , the score for each of the above dimensions
- the output should produce text version of the audio file and display on screen
Error handling and logging : The process should log every activity/notification, errors, warnings in order to enable troubleshooting and correction.
The code must have appropriate comments & modularity to make the maintenance easy
Consultant will procure the initial data set to work with.
Consultant will provide the initial deliverables with quality tested well. An accuracy of at least 80-85% is expected. Consultant agrees to have at least 2 iterations on top of the initial delivery as we test the model with more data after the initial delivery
After the initial delivery, consultant will provide the support for at least 2 agreed upon iterations in this scope.
Technology Stack
1. Open to suggestion. Goal is to utilize open source technologies for the above with highest accuracy.
About the recuiterMember since Mar 14, 2020 Norisa Rorah
from Calabria, Italy