Xyla_bot

snapshot of telegram bot

Try it out!

Search xyla_bot on telegram!

Inspiration

At times, we receive audio files but are not at a convenient place to listen to them. Our Telegram bot, Xyla-bot, helps to translate these audio files into Telegram messages. However, sometimes intent/ emotion is lost. Hence, we use machine learning methods to identify intent as well as add life to messages through the use of emojis! :D

What it does

Converts Telegram messages into emoji
Reads audio and voice files and translates them into Telegram messages
Detects emotion/ intent in Telegram messages

How I built it

Telebot to lay the groundwork
Wit.ai to process audio/ voice files to Telegram messages
Wit.ai to detect emotion/ intent in Telegram messages
Emoji to convert keywords in Telegram messages into emoji

Challenges I ran into

Decoding of voice files from Telegram using Opus - it is evil :(
Limited Training for Wit.ai
Large amount of math involved in edit audio files
Getting a server to host our bot

Accomplishments that I'm proud of

A working prototype, building a telegram bot with little experience in web app dev
That we overcame Opus codec challenges with a workaround
Fairly Accurate Sentiment and Emoji recognition

What I learned

How to use Python packages
How to navigate Github, Telegram and Heroku APIs
Wit.ai and other APIs
A little understanding into the internal libraries and init files of packages and modules

What's next for Xyla_bot

Manipulating Audio - Chiptune, Jingles
Error-proofing the code
Training Xyla’s message processing capabilities through machine learning: More specific and wider range of sentiments, Better emoji recognition
More nuanced responses from Xyla