Ranking of Keyword-Based Search Query Results in Knowledge Graphs
Title: |
Ranking of Keyword-Based Search Query Results in Knowledge Graphs |
Author(s): |
Antonio Noack |
Supervisor(s):
|
Sirko Schindler, Leila Feddoul, and Birgitta König-Ries |
Thesis Type:
|
Bachelor thesis |
Publication Place:
|
Friedrich Schiller University Jena |
Date:
|
2020-08-17 |
Abstract:
|
In this bachelor thesis, we propose a ranking based on learning to rank to order results of queries over knowledge graphs. Queries consist of basic graph patterns with variable entities and the results are bindings of actual entities to the variables. With these bindings, a subgraph of the knowledge graph can be reconstructed from the query. We discuss existing approaches from information retrieval (e.g. TF-IDF) and show how to adapt them to a graph environment. The main challenge is that information retrieval methods usually rely on rich text documents, which are often missing in knowledge graphs. The extensive text information existing in those graphs is mostly represented only
by rather short literal values. To address the issue, we propose a method to extract string data from knowledge graphs in order to make information retrieval methods applicable. In addition, we explore different graph metrics (e.g. PageRank) and their possible
application to a ranking scenario. Moreover, we suggest to use a neural network to create a score for ranking from a series of features. The proposed features are based on information retrieval techniques, our string extraction method and graph metrics. |
By Leila Feddoul in on October 22, 2020