This project is divided into two parts:

• an API is created to transcribe, translate, and analyze the sentiment values of audio files using multiple commercial APIs.
• Determining the accuracy of multiple commercial transcription models by transcribing 4070 audio files against each model

All of the following needs to be downloaded in your device

• Python 3.6
• virtualenv
• pip
  • Make sure that the folder where python is downloaded is in the enviroment variable
• Google Cloud Platform API key Optional:
• Deepspeech pre-trained model

• download the github repo

  git clone https://github.com/dighr/audio_analysis.git
  

• Navigate to the cloned repo
• Create a python virtualenv

virtualenv -p python3 env

• Activate the virtual environment

source env/bin/activate

 * In windows the command is 

source env/Scripts/activate

• Download all the dependencies

pip install -r requirements.txt

• Add the GOOGLE_APPLICATION_CREDENTIALS into the enviroment variables

  • If you don't have an account yet, create an account https://cloud.google.com/

  • Within the google cloud console, create a new project

  • Within the created project, enable the NLP API and the audio-to-text API

  • Create a service

  • Download the credentials of the created service as JSON

  • Add an enviroment variable named GOOGLE_APPLICATION_CREDENTIALS and point the downloaded JSON file into it

    • In linux, just execute the following in the terminal

      export  GOOGLE_APPLICATION_CREDENTIALS=path-to-the-downloaded-json-file
    

  • Add an environment variable to store the private API keys of both Watson and Azure

Assuming that the previous section is complete, you can start the server by executing the following in a terminal

python manage.py runserver

The API calls to transcribe audios require a language_code as a parameter to be passed in within the body. Find the correct language code from the following URL https://cloud.google.com/speech-to-text/docs/languages

Both 'text/analyze' and 'text/translate' API also requires 'language_code' as a parameter when making an API call. Use the correct ISO-639-1 Code value for 'language_code from the link below https://cloud.google.com/translate/docs/languages

To analyze a text, make a post request to the following URL PASS in both text=text_to_be_analyzed, method=google, and language_code into the body

http://localhost:PORT/text/analyze

To translate a text, make a post request to the following URL Provide both 'text' (text_to_be_translated) and 'source_language' (Explained above) as an input to make the translation

http://localhost:PORT/text/translate

To transcribe an audio file, make a post request to the following URL

http://localhost:PORT/audio/transcribe

Attach the audio file within the body of the request in the following format 'file=audio_file_path' pass in the 'language_code' parameter into the body

To analyze an audio file, make a post request to the following URL

http://localhost:PORT/audio/analyze

Attach the audio file within the body of the request in the following format 'file=audio_file_path' pass in the language_code parameter into the body

python manage.py test

This project is licensed under the MIT License.

• A lot of functionality was enabled thanks to the Pydub library https://github.com/jiaaro/pydub
• For long audio files, a method created by the following git repo https://github.com/akras14/speech-to-text.git was used. However, we made the process of audio encoding and the audio file splitting automatic