Campus Event Calendar
Event Entry
What and Who
Analyzing and Creating Top-K Entity Rankings
Evica Ilieva
International Max Planck Research School for Computer Science - IMPRS
PhD Application Talk
IMPRS Masters Student
AG 1, AG 2, AG 3, AG 4, AG 5, SWS, RG1, MMCI
Public Audience
English
Note: We use this to send email in the morning.
Date, Time and Location
Monday, 7 October 2013
08:50
120 Minutes
E1 4
024
Saarbrücken
Abstract
Ranking information based on different criteria is one of the most natural and widely used techniques to condense a potentially large amount of information into a concise form. Cars are compared by gas per mile, websites by page rank, students based on GPA, scientists by number of publications, and celebrities by beauty or wealth. Some rankings are highly subjective and need to be gathered through crowd sourcing user votes in public websites. On the other hand, there exists a large amount of rankings that can be derived from publicly available data.
In both cases-computed or crowdsourced-the topic of the ranking needs to be carefully determined to be semantically meaningful. We propose the usage of knowledge bases to phrase semantically meaningful rankings. To gain some insights on the characteristics of entity rankings, we first conducted a study on crowdsourced entity rankings collected from a popular ranking portal. The frequency in which different types of rankings are generated gives us insights on the importance of the various ingredients of a ranking and is further complemented by including the popularity of rankings in terms of attracted user views. Our ranking generation algorithm is assembling the thematic focus and ranking criteria of rankings by inspecting the present Subject, Predicate, Object (SPO) triples in a knowledge base. Similarly to the a priori principle of frequent itemset mining, we iteratively generate entity rankings in a bottom up fashion. To generate entity rankings that we believe would be of higher interest to the users, we also introduce a minimum support threshold t, that puts focus on entity rankings where at least t entities qualify for the ranking. Making use of numerical attributes contained in the knowledge base we are also able to compute the actual ranking content, i.e., entities and their performances. We report on first results obtained using the YAGO knowledge base.
In both cases-computed or crowdsourced-the topic of the ranking needs to be carefully determined to be semantically meaningful. We propose the usage of knowledge bases to phrase semantically meaningful rankings. To gain some insights on the characteristics of entity rankings, we first conducted a study on crowdsourced entity rankings collected from a popular ranking portal. The frequency in which different types of rankings are generated gives us insights on the importance of the various ingredients of a ranking and is further complemented by including the popularity of rankings in terms of attracted user views. Our ranking generation algorithm is assembling the thematic focus and ranking criteria of rankings by inspecting the present Subject, Predicate, Object (SPO) triples in a knowledge base. Similarly to the a priori principle of frequent itemset mining, we iteratively generate entity rankings in a bottom up fashion. To generate entity rankings that we believe would be of higher interest to the users, we also introduce a minimum support threshold t, that puts focus on entity rankings where at least t entities qualify for the ranking. Making use of numerical attributes contained in the knowledge base we are also able to compute the actual ranking content, i.e., entities and their performances. We report on first results obtained using the YAGO knowledge base.
Contact
Aaron Alsancak
068193251800
--email hidden
passcode not visible
logged in users only
Aaron Alsancak, 10/04/2013 11:05 -- Created document.