A bot that serves your taste of music (hopefully).
Say something like 'Play Someone Like Me by Adele' and enjoy! 'Louder' if you don't hear it well, or 'Silent!' if you're tired for a moment. 'What about some hard rock?'* or 'Do you know Let it be by The Beatles?'.
*Playing songs by genre currently has not been implemented.
MIT, Copyright 2020 heckej
- To play music (in Linux), you should install cmus and add music to your library.
- Speech recognition (in Linux) needs
python 3with the SpeechRecognition library installed. - Text to speech (in Linux & Windows with modifications) needs
python 3with the Google-Text-to-Speech library gTTS. For Windows, it is sufficient to have Powershell installed (you still might need to update the path in the code to where Powershell is located).
To play music the Music Master Bot calls the cmus command line music player using the cmus-remote commands. No plans for support of a Windows music player.
The speech client allows a microphone to be used as input device for the chatbot. Communication with the chatbot is established by a custom adapter, more specifically a customised version of the Bot Builder Community Alexa Adapter (MIT License).
Speech recognition (in Linux, could work in Windows) is done in Python using the SpeechRecognition library. When the language of the spoken language is set to something different from English, English input will still be understood.
Requests sent to the custom adapter should contain a Language property. If the value is set to nl (Dutch), the chatbot will partially 'translate' the request to English and respond by saying the response out loud in Dutch to the user (currently not in the written text, because written and spoken text often contain two different values).
The bot can respond by saying some things out loud using text to speech. It therefore uses the Google-Text-to-Speech library in Linux (could work in Windows too) and built in tools in Powershell in Windows. Before sending the text to the TTS service, it is given to fastText (MIT License) to determine the language in which it's written. Based on the result, the bot can send varying language arguments to the TTS service, so the text (e.g. the title of a French song) is pronounced as ideally as possible.
In the code of the bot there is a class that allows to execute commands in a separate process, with a function specifically for python3 scripts, however it should be easy enough to change this to be used in Windows too.
To detect intents and entities a custom LUIS model is used.