This project is based in 4 main modules:
SPR
: Speach Recognition Module --> Two tasks 1.Convert client's voice in text, 2.ConvertNLU
text in the agent's voiceNLU
: Natural Language Understanding Module --> Analyze and make predictions on the previous text using Machine Learning techniquesDM
: Dialogue Manager Module --> Flow process control. The main moduleNLG
: Natural Language Generator --> Prediction text formatter
Additionally the DM
has two Conversational trackers
:
ct_interest
: Conversational tracker in charge of flow conversations related to places of interestct_restaurant
: Conversational tracker in charge of flow conversations related to restaurants and food in general
The complete flow process can be represented with the following graph:
- Docker
- Pulseaudio
Install pulseaudio
. Then execute the run.sh
script (needs Docker).
After installation you can also do a quick test, just launch test.py
Natalia Pattarone
Sayan Goswami
Iñaki Lorente