Non-Linear Multiple Field Interactions Neural Document Ranking

TL;DR

This paper investigates how non-linear interactions between multiple document fields and query information can improve neural ranking models, providing new insights into document structure utilization for ranking tasks.

Contribution

It offers an in-depth analysis of field interactions in neural document ranking, highlighting the effects of query-field and non-linear interactions on performance.

Findings

Query-field interactions significantly impact ranking quality.

Non-linear field interactions improve model effectiveness.

Architecture choices influence the benefits of field interactions.

Abstract

Ranking tasks are usually based on the text of the main body of the page and the actions (clicks) of users on the page. There are other elements that could be leveraged to better contextualise the ranking experience (e.g. text in other fields, query made by the user, images, etc). We present one of the first in-depth analyses of field interaction for multiple field ranking in two separate datasets. While some works have taken advantage of full document structure, some aspects remain unexplored. In this work we build on previous analyses to show how query-field interactions, non-linear field interactions, and the architecture of the underlying neural model affect performance.