Ontology based content recommendation

From Master Projects
Jump to: navigation, search


has title::Ontology based content recommendation
status: finished
Master: project within::Information Sciences
Student name: student name::Simon Klink
Dates
Start start date:=2008/04/07
End end date:=2010/02/26
Supervision
Supervisor: Laura Hollink
Second reader: has second reader::Guus Schreiber
Thesis: has thesis::Media:Thesis.pdf
Poster: has poster::Media:Media:Posternaam.pdf

Signature supervisor



..................................

Abstract

With the amount of content on websites like Youtube, Ebay and Amazon increasing daily it is difficult to keep visitors interested. One way of keeping visitors interested is by recommending content which they might find interesting. Recommending content of their interest can be done in several ways. The most commonly used ways are through the use of tags, categories or collaborative filtering.

Unfortunately all these types of recommendation have shortcomings. Tags are just strings of characters with no underlining relation to other tags. Categories don’t always fit the content, and creating unlimited categories decreases the overview. Collaborative filtering is a good way of recommending other content but requires a lot of data in order to work properly.

In this thesis we will use ontology concepts for the annotation and recommendation of other content. We will focus on how we can use the relations between different concepts for the recommendation of other content that has been annotated with ontology concepts.

First we will chose an ontology that suites our requirements. After an ontology is chosen we will test in which way we can use the relations between the concepts that are present in the ontology. Next we will define three different ways in which we can use the relations between ontology concepts.

In order to test our test our different ways of recommendation we have created an annotation prototype as well as an recommendation prototype. This allows us to annotate content with ontology concepts and recommend content based on those annotations.

To test which of our ontology based recommendation methods works best we have conducted a survey among a test group. The results from the survey were than processed and submitted to various tests to compare the results. A quantitative analysis as well as a qualitative analysis of the results has been done to give an insight into the recommendation process.

We conclude with our findings on our ontology based recommendation system. Our system can make proper recommendations based on the relations that exist between different ontology concepts. However our methods are not the answer to all recommendation problems.