neulab / idiomata-bot

A Telegram bot for supporting conversations

IdiomataBot

This is a Telegram bot for supporting conversations.

Instructions for Running the Bot

Install requirementes:

pip install -r requirements.txt

Create a bot according to the Telegram Bot Instructions Add the token to a file named "token.txt".

Download data for some common languages:

bash download_data.sh

Run the bot:

python bot.py

Instructions for Using the Bot

Message IdiomataBot (or whatever other bot name you're using the software with). Alternatively, add IdiomataBot to a conversation between two people, click on its name, and then select "promote to administrator" so it can follow along with the conversation.

Note that this means that the bot will be able to read your conversation. It does not save any logs, and only keeps track of how many words you spoke in each language for now.

idiomata-bot's Issues

Spelling check

It would be nice to have spelling check for the translation capability.

Add database support

Right now all the information in the bot is saved in memory and disappears when the bot is reset. It would be nice to have this information in a database.

Conversation prompts

It would be nice if the bot could provide conversation prompts, that allow people to get something to talk about when they're stuck. (Idea from Pat Littell)

Ability to add dictionary entries

It would be nice to be able to add dictionary entries.

Better language ID

Currently our language ID is just based on word lookup, it would be better if we had a more sophisticated method.

Test the bot with listening in on a conversation

Right now the bot can be used in a one-on-one setting with human+bot. However, the intended usage of the bot is that it listens to a human-human conversation and records the number of words spoken. It would be good to test it in this setting.

Scores should reflect the language that people are trying to learn

Right now scores are with respect to any language that the app can recognize. However, in most cases the user will be interested in one or two languages that they speak, and scores should reflect the choice of language appropriately.

Better startup documentation

Right now it's hard to know how to use the bot unless someone tells you how to do so directly. It'd be good if there was a way to make it better documented so people can use it without instruction.

Better scores

It would be nice if we could create scores that are more engaging: amount of vocabulary covered, etc.

Make the bot respond to commands

Currently the bot responds to all messages. It would be nice if the bot could respond to commands and provide information on request.

Broader coverage language ID

It would be nice if we could do language ID in many languages, maybe all of the languages in Wikipedia?

Separate data from github

Right now we've uploaded data to github, but it'd be better to have the data accessible through download.