Practical Approach to Python based Supervised Machine Learning: User Generated Text Classification Techniques
Submitted by Kausik Ghatak (@kausikg) on Monday, 15 June 2015
In e-Commerce, we handle large volume of user genearted content in the forms of Reviews, Ratings, Question/Answer, Chat etc. These user generated content has lot of values in terms of taking right organization-wide business decission. This large volume of user generated text also imposes problem of classificaiton and moderation because the data is mostly unstructured. Combination of various Machine Learning techniques can be a convenient tool to handle the problems in this space. In this session we want to explore how we can bring the Python NLTK and other Machine learning components together to solve this problem and provide the solution as service (Python Flask) to integrate the intelligence with other parts of e-Commerce platform.
We will be covering the following topics in the session:-
1) Introduction to Supervised ML text classification. How we can model the text classification problem to a supervised ML problem. What are the traditional ways to combine various ML models.
2) What are the challanges to handle the User generated content/text. Why we can not just shoe-horn the established ML models in solving moderation problem in user generated content.
3) How we can integrate Python NLTK and other ML Components for text classification.
4) What are the best practices of using Python based ML models: Training and testing models
5) How we can combine multiple ML models and RegEx logic.
6) Python Flask for exposing the ML logic as service.
Basic Computer science knowledge. Knowledge in Python will be helpful.
Kausik is Senior Manager(/Architect) in Snapdeal, managing the Machine Learning component for User Generated Content(Reviews, Ratings, Question/Answer, Chat). Kausik is MBA from IIM Bangalore and BE from Jadavpur University, Kolkata. Kausik has previously worked in Amazon(Product Aggregation Technologies) and Goldman Sachs(Hedge Fund Risk Management).