A unified relevance model for opinion retrieval

Abstract

Representing the information need is the greatest challenge for opinion retrieval. Typical queries for opinion retrieval are composed of either just content words, or content words with a small number of cue "opinion" words. Both are inadequate for retrieving opinionated documents. In this paper, we develop a general formal framework-the opinion relevance model-to represent an information need for opinion retrieval. We explore a series of methods to automatically identify the most appropriate opinion words for query expansion, including using query independent sentiment resources. We also propose a relevance feedback-based approach to extract opinion words. Both query-independent and query-dependent methods can also be integrated into a more effective mixture relevance model. Finally, opinion retrieval experiments are presented for the Blog06 and COAE08 text collections. The results show that, significant improvements can always be obtained by this opinion relevance model whether sentiment resources are available or not.

Biography-Dependent Collaborative Entity Archiving for Slot Filling

“…Generally, Relevance Model (RM) (Huang and Croft, 2009) refers to the probability distribution over all words conditioned on their occurrences in a set of previously-known relevant documents (or high-quality pseudo-relevant documents), i.e.,…”

Personalized Web Search Using Emotional Features

“…Eguchi et al [22] proposed several sentiment retrieval models based on probabilistic language models, assuming that users both input query topics and specify sentiment polarity. Similar methods proposed in [23] and [24] unified topic relevance and opinion relevance respectively based on a quadratic combination and a linear combination. A different method proposed by Li et al [25] combined topic-sentiment word pairs in a bipartite graph to effectively rank the documents.…”

“…The role of emotions in information retrieval is investigated in some researches [19][20][21]. Specially, researches on sentiment retrieval or opinion retrieval [22][23][24][25] aim to provide a general opinion search service, similar to traditional Web search in the way that both of them find pages relevant to the query, but different from the latter in the way that sentiment retrieval need further determine whether the pages express opinions on the query topic and whether their polarities are positive or negative.…”

Construction of a sentimental word dictionary

“…The problem of determining the polarities of words has been studied by many researchers (e.g., [5,8,9,10,15]). The general theme is that there are sets of seed words that have known polarities; e.g., the seed word "good" is positive and the seed word "bad" is negative.…”

“…Opinion mining (e.g., [15,6,10,8] and opinion retrieval (e.g., [16,17,9]) have significant practical interest. We aim to construct a sentimental word dictionary, which is of high quality and has a large coverage of sentimental words.…”