SIGIR eCom'19 Digest

Last month, July 25 2019 to be precise, the third SIGIR Workshop On eCommerce took place in Paris, France. Several quite intersting ideas and topics were discussed. Below I’ll try to share links to resources shared during the workshop as well as give and outline of the key ideas that were discussed during the event:

Observations

Unlike the main SIGIR conference in which recommendation was the predominant topic, SIGIR eCOM was more inclined towards the search component of e-commerce websites. In particular, among the 24 accepted short and long papers there were 12 of them about search.

More than 58% of the papers are 6 pages or more long.

A wordcloud of the papers’ keywords, shows that e-commerce is at the center of most papers, which confirms a heathly correlation between the papers and the workshop. Beyond that, learning with its different flavors (deeplearning, reinforcement learning, machine learning) stands out. Search and search related keywords (e.g: retrieval, spell checking, query etc) are predominant as well.

Deeper dive into the papers

As search is the predominant topic, we will start by making sense of the various ideas suggested to improve it.

A typical e-commerce website, or any website for that matter, has a search bar that enables users to find what they are looking for. In the case of e-commerce however, this search box is more critical.

Good search user experience and search results’ relevancy can make the difference between converting a visitor or sending the potential customer to the competition. Futhermore, as suggested by this article, which was outlined in paper [1], more customers are chosing e-commerce search engines, i.e the search bar in e-commerce websites, over a generic one such as google for their shopping. This puts the pressure on e-commerce search engines to deliver more relevant results.

References:

1. Leverage Implicit Feedback for Context-aware Product Search

(to be continued)