Dear search engine: what's your opinion about...?: sentiment analysis for semantic enrichment of web search results

Abstract

Search Engines have become the main entry point to Web content, and a large part of the "visible" Web consists in what is presented by them as top retrieved results. Therefore, it would be desirable if the first few results were a representative sample of the entire result set. This paper provides a preliminary study about opinions contained in search engine results for controversial queries such as "cloning" or "immigration". To this end, we extract sentiment metadata from web pages, and compare search engine results for several queries. Furthermore, we compare opinions expressed in the top results to those in other retrieved results to examine whether the top-ranked pages are a good sample of all results from an opinion perspective. In a preliminary empirical analysis, we compare up to 50 results from 3 commercial search engines on 14 controversial queries to study the relation between sentiments, topics, and rankings.