Friday 15:45–16:30 in Hall 1

Classifying Search Queries without User Click Data

Abhishek Thakur

Audience level:
Intermediate

Description

This talk discusses how machine learning/data mining techniques can be applied to classify search terms that people use in search engines like Google, Bing, Yahoo etc. The talk focuses on machine learning techniques such as LSTM (long short term memory) rather than traditional ways like analysing user-behaviour with the help of their search logs.

Abstract

Traditionally, search queries are classified into different categories by analysing user-behavior with the help of search logs. Instead of focussing on the search logs and by analysing queries with the help of machine learning models such as LSTM, it is possible to get a very decent model to classify the search queries.

This talk focuses majorly on LSTMs and their usefulness when it comes to search query classification. We also discuss how we can accurately classify search queries in hundreds of categories using open source data available online and how this can be combined with LSTMs to provide a much stable and better result.