Search in Elsevier Research Products: Current State and Lessons to be Learned

Abstract

In this work we report on an analysis of the current user experience for search in a number of Elsevier products from a Natural Language Processing (NLP) point of view. This is an exploration of search behaviour and identifying where established NLP insights could alleviate the most pressing issues. In our analysis we identify a number of clusters of search patterns and explore the possibilities for improving the response given the user's search intent. We are also interested in consistency in the search experience across Elsevier platforms and relate to the higher expectations that have been set in the outside world in recent years. Finally, we hypothesise how a more sophisticated way of ingesting our content can help to look beyond indexing on the level of isolated tokens, in order to have more context available to link a search query to the most suitable content.

Keywords: NLP for Search, Search Intent, Contextual Search