hawzie197 / llama-api Goto Github PK

We need to clean up all the code and move things to the correct place.

Things to consider:

• Make a Parse class and nest methods that just deal with parsing
• Add docstrings to all active functions/methods
• Are we bridging the classifier and fuzzy together? or separate?

Bad urls:

• http://localhost:8000/api/analyze/?url=https://www.soundcloud.com
• http://localhost:8000/api/analyze/?url=https://www.pandora.com/

Here in the image above, the STORE action should be '0' as it is in the corpus data

We should have the get_fuzzy_result function return the object we will be adding to the return data.

The current implementation is to complex with indexing. Reference below:

action_map = dict(
                type=action.title(),
                quote=result[0][0],
                confidence=result[0][1],
                classification=result[0][2],
            )

Just like we have the policy URL parsing, we need to have a way to parse out the support email so that we can display a way for the users to easily get into contact with the company if need be.

Currently, the algorithm takes 14 seconds. We need to speed this up to less than 5 seconds.

Multiprocessing?
Reduce the number of keywords?