Applications of Voting Theory to Information Mashups

Abstract

Blogs, discussion forums and social networking sites are an excellent source for people's opinions on a wide range of topics. We examine the application of voting theory to "Information Mashups" -the combining and summarizing of data from the multitude of often-conflicting sources. This paper presents an information mashup in the music domain: a Top 10 artist chart based on user comments and listening behavior from several Web communities.We consider different voting systems as algorithms to combine opinions from multiple sources and evaluate their effectiveness using social welfare functions. Different voting schemes are found to work better in some applications than others. We observe a tradeoff between broad popularity of established artists versus emerging superstars that may only be popular in one community. Overall, we find that voting theory provides a solid foundation for information mashups in this domain.

Context and Domain Knowledge Enhanced Entity Spotting in Informal Text

“…Closely related to our work, domain dictionaries have been widely used in NER, including Wikipedia [4] and Wiktionary [14], DBLP [10], KAON [3], and MusicBrainz [1]. They have also been used for the task of disambiguating entity senses, an important step in accurately extracting entities.…”

“…Understanding the semantic relationships between entities in these challenging domains is necessary for a variety of information-centric applications, including the BBC SoundIndex [1]. This application, developed by the authors and others, provides a realtime "top 40" chart of music popularity based on sources such as MySpace and YouTube.…”

MONGOOSE: MONitoring Global Online Opinions via Semantic Extraction

“…As a large proportion of valuable online data is in "Broken English" 2 , the system includes new techniques for spotting and making sense of "Broken English" use. 2 Current analytics systems assume structured English and that Natural Language Processing (NLP) can be used. This is not true online.…”

“…This might be as simple as counting the number of positive and negative comments on a discussion board, but may become more complicated when the underlying data comes from multiple sources of very different modalities (sales of songs verses listens of songs versus views of music videos). Some decisions need to be made how to do this combination that makes sense for the application (please see [2] for an extended discussion of the application of voting theory to this space).…”

“…The MONGOOSE base package contains acquisition templates for the automotive and music industries, preprocessing modules for comments and counts, the IBM Tunable Ranked List Producer algorithm for adjudication [2] and a Nagios-based [9] implementation of the MONGOOSE Control Platform, which performs workflow analysis and general system monitoring.…”