Intent-based Anomaly Seeker

Project Overview

Find outlier/anomaly for multi-class intents using Snips NLU.

Project Purpose

This is intend for a code challenge! Still lots of space to improve.
Since it is more of a natural language understanding - e.g. chatbot, voiced-base conversation... problem, I decided to go with Snips NLU.

Data Summary

List of lists within a json file in the form of - [ Query , Class ]

• Total Number of Classes : 150
• Total Number of Queries per Class : 150

For Example, data would look like this:

[
  ["what expression would i use to say i love you if i were an italian", "translate"], 
  ["if i were mongolian, how would i say that i am a tourist", "translate"],
  ... 
]

Requirements

Install require packages:

pip install -r requirements.txt

Install SpaCy en_core_web_lg corpus:

python -m spacy download en_core_web_lg

*Windows user might need to go through command line w/ Admin to install under the virtual environment of this project

Execution

In the terminal, type in:

python p2.py data.json

Wiki

Documentation on overview, approaches, how Snips NLU fits into tackling this challenge.
See here for more details.