Sunday 14:15–15:00 in Auditorium

Smart search using Support vector machines

Dr. Shahzia Holtom

Audience level:
Experienced

Description

Search is a common, if not an essential, feature of an app. A good search has to retrieve the information that matches a user's query but a great search should also personalize the information to improve the overall relevance. This talk will share the lessons from a recent implementation of a smart search for a grocery delivery app for one of UK's largest supermarkets.

Abstract

Smart search using support vector machines

Search is a common, if not an essential, feature of any digital app. A simple search engine should retrieve all the information that matches a user's query. But a smart search engine is one which also ranks the retrieved information in an order of relevance that is personalised. Machine learning classifiers such a support vector machines are a handy tool for implementing smart search. Using agile, data science practices such as test-driven development and paired programming in a balanced team, we have built a grocery delivery app, with smart search, for one of UK's largest supermarkets. This talk will discuss the approaches and lessons learnt from putting data science in production.