Presentation Title

Supporting Multilingual Users Through Personalized Multilingual Search

Faculty Mentor

Ben Steichen

Start Date

23-11-2019 8:45 AM

End Date

23-11-2019 9:30 AM

Location

164

Session

poster 2

Type of Presentation

Poster

Subject Area

engineering_computer_science

Abstract

With the global expansion of the Internet and the World Wide Web, online user diversity is continually growing. In particular, there is an increasing number of users who are proficient in multiple languages. Such multilingual users have, in theory, access to a greater set of information, because they have the ability to search for documents in multiple languages. In fact, prior research has found that multilingual users like to use each of their languages online, and that different topics elicit different search behaviors (e.g. for some topics, all languages are used, whereas specific languages are preferred for other topics).

While the number of multilingual users is rising, multilingual searching is typically not directly supported by search engines, as they tend to emphasize distinctions between languages, often requiring users to switch between different versions of a system in order to find information in different languages (or at least requiring the user to change the query language). In this project, we aim to build novel multilingual search engines that directly support multilingual information search, by retrieving and presenting search results in multiple languages. In particular, our current research uses machine learning models to accurately predict in what language a user would like to receive results following a query (i.e. all languages, their primary language, or a secondary language). Initial classification results using data from 530 multilingual users show that such an approach can attain accuracies of up to 68%, and that an individual user's language proficiencies, query topic, as well as prior language choices each contribute to these classifications.

This document is currently not available here.

Share

COinS
 
Nov 23rd, 8:45 AM Nov 23rd, 9:30 AM

Supporting Multilingual Users Through Personalized Multilingual Search

164

With the global expansion of the Internet and the World Wide Web, online user diversity is continually growing. In particular, there is an increasing number of users who are proficient in multiple languages. Such multilingual users have, in theory, access to a greater set of information, because they have the ability to search for documents in multiple languages. In fact, prior research has found that multilingual users like to use each of their languages online, and that different topics elicit different search behaviors (e.g. for some topics, all languages are used, whereas specific languages are preferred for other topics).

While the number of multilingual users is rising, multilingual searching is typically not directly supported by search engines, as they tend to emphasize distinctions between languages, often requiring users to switch between different versions of a system in order to find information in different languages (or at least requiring the user to change the query language). In this project, we aim to build novel multilingual search engines that directly support multilingual information search, by retrieving and presenting search results in multiple languages. In particular, our current research uses machine learning models to accurately predict in what language a user would like to receive results following a query (i.e. all languages, their primary language, or a secondary language). Initial classification results using data from 530 multilingual users show that such an approach can attain accuracies of up to 68%, and that an individual user's language proficiencies, query topic, as well as prior language choices each contribute to these classifications.