search

gunthercox / ChatterBot

ChatterBot is a machine learning, conversational dialog engine for creating chat bots
https://chatterbot.readthedocs.io
BSD 3-Clause "New" or "Revised" License
14.12k stars 4.45k forks source link

Few questions #2229

Open Zolooooo opened 2 years ago

Zolooooo commented 2 years ago

Hi!

We are a legal tech company planning to develop a legal assistant chatbot. I’d like to know more about the suitability of your framework with our needs. So, please answer the following questions:

  1. Is Mongolian supported?
  2. Is your framework capable of building a chatbot that answers all types of legal questions. Is there a similar use case? (maybe in different industries?)
  3. If your framework is capable of developing a legal assistance chatbot, approximately how many Q&As (intent & response) do we need to train so that our chatbot provides 10,000 different responses? We’re planning to collect 10,000 answers to 100,000 legal questions (10 questions for each response). Is this enough data to train this type of chatbot?
  4. What AI methods are used or what types of algorithms are used?

Best, Zolo

brightening-eyes commented 2 years ago

hi,

  1. yes, but you might right a custom comparison function (as spacy does not have word vectors for Mongolian (correct me if I'm wrong)).
  2. with logic adapters, you can do everything. the use cases are depend on your needs. (even you can do custom things like retrieving weather)
  3. as many as you can put in a database. just create 2 tables, one for questions and another for answers, and write a logic adapter that retrieves questions from user, preprocesss it, searches the database and returns an answer randomly. (or with whatever strategy you prefer).
  4. you can have a logic adapter that can use deep learning method. but, chatterbot already uses naive bayes. you also can generate your answers with deep learning in your logic adapters.