Apollo, your music recommender

Context:
    Final project of the course in Natural Language Interaction that I attended during my master studies at Universitat Pompeu Fabra

Technologies:
    Natural Language Interaction, Graphical User Interface, Speach-to-text, Text-to-speach, Python, QML

Description:
    Apollo is a conversational agent in the form of a Graphical User Interface which has the goal to inform about classical music and make recommendations. It uses information from the online music database Last.fm and it follows its execution in 3 phases:

• The Greeter, which greets the user and ask for its name
• The Suggester, which suggests musical periods, composers and give information about them
• The Closer, which farewells the conversation

    The GUI is implemented in Python and the Qt framework and it uses sound effects and animations.
     This agent receives the user's inputs from the microphone and preprocesses it with the Speech Recognition and Punctuator libraries. Then, the Dialogue Manager of the agent tokenizes the data, tags the sentences and vectorizes them with Bag of Words by using the Natural Language Tool Kit (NLTK) library. This, classifies the dialogue in types of sentences and acts according to the information previously given. In order to generate the response, the agent uses a template base generator, which uses sample senteces with gaps filled by the model and then reproduce them with the text-to-speach library Pyttsx3.
    The information and questions from the user are stored in order to adapt the conversation to the previous dialogue, for example, by not repeating suggestions.

Github repository: github.com/pnonell/apollo

GUI's capture: