INESC TEC-Amazon collaboration optimises search engine results for special dates

The seasonality of search queries in search engines could be a factor for online businesses to consider if they seek to improve the ranking of their results. A new demo-paper featuring INESC TEC explored the creation of a database to present the Occasion-aware Recommender solution.

Although users may not be aware of it, the way they interact with search engines tends to change during festive periods, like Christmas or Valentine's Day. Search queries like "gift" or "romantic experience", without explicitly referencing a festive season, could be a strategy for online businesses to explore towards optimising the results presented to costumers.

In a new demo-paper, Hugo Sousa, a researcher at INESC TEC, joined Austin R. Ward and Omar Alonso, from Amazon, to assess the impact of including temporal information in search. Starting from a database with information on 2000 events, the process continued with the modelling of the daily importance of each event, based on historical data. With the events featuring validated time models, they were introduced into the system - which uses time models to determine which events are relevant to the date the search was made.

In "Don't Forget This: Augmenting Results with Event-Aware Search", published at the ACM International Conference on Web Search and Data Mining, conference A*, the researchers also stated that the "number of products related to the events presented in each research is proportional to the importance of the event according to the estimate of the temporal model." Previous research showed the importance of seasonality in research results, namely using Language Models; however, said models lack knowledge about world events proved to be a weakness.

The solution designed and presented in the paper is Occasion-aware Recommender.  Considering the insufficient research work that explores the influence of events on research results, the authors claimed that their work can contribute to broaden the line of research. The next goal is to extend the use of the database to other contexts, e.g., the integration of other types of events - like sports events and concerts, or personal events weddings or birthdays.

The researcher mentioned in this news piece is associated with INESC TEC and UP-FCUP.

This article was first published on 26 February by INESC.