Publication Date

8-13-2021

Date of Final Oral Examination (Defense)

August 2021

Type of Culminating Activity

Thesis

Degree Title

Master of Science in Computer Science

Department

Computer Science

Major Advisor

Maria Soledad Pera, Ph.D.

Advisor

Michael Ekstrand, Ph.D.

Advisor

Casey Kennington, Ph.D.

Abstract

Online search engines for children are known to filter retrieved resources based on page complexity, and offer specialized functionality meant to address gaps in search literacy according to a user's age or grade. However, not every searcher grouped by these identifiers displays the same level of text comprehension, or requires the same aid with search. Furthermore, these search engines typically rely on direct feedback to ascertain these identifiers. This reliance on self identification may cause users to accidentally misrepresent themselves. We therefore seek to recognize users from skill based signals rather than utilizing age or grade identifiers, as skill dictates appropriate aid and resources. Therefore, in this thesis we propose a strategy that automatically recognizes users on the fly by analyzing search behavior found in search sessions. In particular, our efforts focus on recognizing the stereotypical 8 to 12 year old searcher, who we posit exhibits skills defined by developmental stages that have a strong impact on language development (Piaget's concrete operational stage) and search literacy (digital competency's first level). This strategy analyzes user-generated text extracted from queries and patterns of search interactions in order to infer features that are leveraged by a random forest classifier in order to determine whether or not a user is a part of this specific segment of searchers. The outcomes from this thesis lay the groundwork for enabling search engines to recognize users based on their search skills and provides further insight into the search behavior of youths.

DOI

https://10.18122/td.1858.boisestate

Share

COinS