Degree Type

Creative Component

Semester of Graduation

Spring 2020

Department

Computer Science

First Major Professor

Ying Cai

Degree(s)

Master of Science (MS)

Major(s)

Computer Science

Abstract

A popular search on Google Maps is top-k spatial keywords query. Given a value of k, a set of keywords, and a reference point, Google Maps returns the k spatial objects whose description and location are most relevant to the search keywords and the reference point. Here the relevance is computed using a scoring function. While Google Maps supports top k queries, it keeps the scoring functions in secret. Our research is interested in this secret. Specifically, we want to know the scoring functions used by Google Maps in processing top k spatial keywords queries. We believe that knowing the ranking behaviors of a search engine such as Google Maps will make it possible to leverage the engine to search for the information not available from the standard services it provides. To estimate the scoring functions, we develop a process that analyzes the top-k query results returned by Google Maps. Based on the monotonicity properties of geo-proximity and relevance of the textual description to the search keywords, we assume the scoring functions used in Google Maps are in the linear format and leverage the linear inequality systems to iteratively solve the coefficients and the constants in scoring functions. Together with different methods of computing the textual relevance and solving the linear inequality systems, we estimate six sets of scoring functions. To see the accuracy of the estimated functions, we use them to perform the top-k spatial keywords queries and compare the query results with those given returned by Google Maps in terms of the number of the common spatial objects and the ranking order. In this report, we will present the implementation and examine the performance results.

Copyright Owner

Lyu, Siying

File Format

Word

Share

COinS