NLE technical design plan

[maqboolkhan]

Think about the technical design and how it will integrate into IDA.

We have our own branch from master name NLE. Nandeesh and I will be creating sub-branches from NLE and when our NLE would be ready for its prime time we will merge it into master

[Cortys]

This might be useful: https://github.com/percyliang/sempre

[maqboolkhan]

Initial thoughts,

• We can go with the intent classification so that every message by the user will be classified with its intent and then chatbot can decide action or handling of upcoming messages.
• As Python has more rich ML or SNLP frameworks or libraries then it's better to develop NLE on Python as another module.

[maqboolkhan]

Initial flow

[maqboolkhan]

Challenges with Intent Classification:

• Messages outside our scope should not classify with any intent.
• To make NLE more sophisticated with its reply, we have to handle messages with multiple intents liek (Hi, what can you do for me, the reply should be Hi, I can do blah blah..)
• How we gonna generate replies? are they going to be fixed with intents?
• Parsing of useful information from messages

[maqboolkhan]

Some helpful links

• https://medium.com/analytics-vidhya/building-a-simple-chatbot-in-python-using-nltk-7c8c8215ac6e (not so helpful but good to study)
• https://chatbotslife.com/how-i-developed-my-own-learning-chatbot-in-python-from-scratch-and-deployed-it-on-facebook-88bc828be0a8