Contextual Autocomplete suggestions in Realtime
Autocomplete is a predominant feature in e-commerce search. By being relevant, Autocomplete should help users quickly find the query they intended to type with minimal keystrokes. This talk presents an approach on how this is achieved by considering the users context as a signal. This context is built in real-time using a series of models & fed into a ranking model which re-ranks suggestions accordingly.
This talk is about how we have improved Autocomplete suggestions at Target using session context. The sequence of ideas will be as follows:
How Autocomplete is implemented at Target. We will also see the underlying Architecture involved.
We will then introduce what context is and how it can enhance a user’s shopping experience.
Implementing context can be done in various ways. We will take an overview of various approaches we can use, with their merits and drawbacks.
We will then talk about the modeling approach we have taken. We also cover how our model and the underlying architecture support low latency, which is critical for a good Autocomplete experience.
We will also discuss the various challenges we encountered while doing this.
Basic understanding of Deep Learning
Dileep Kumar is a Lead Engineer at Target, working on various Data Science problems in e-commerce search. Dileep has a decade of experience in the field of Data Science and Data Analytics. He has worked on various problems.
Prior to his current role, he has 10 years of experience solving problems using data science across various industries - friend recommendations & content ranking in social networking, user segmentation & marketing campaign optimization in mobile games, and supply prediction in commodity markets.
His academic background is MS in Analytics from Georgia Tech, and BTech in Mechanical Engineering from IIT Madras.