Using Social Annotations to Smooth the Language Model for IR

Abstract

In the paper, we present an exploration of using social annotations provided by the Web 2.0 sites (such as Del.icio.us) in helping web search. More specifically, we consider using the social annotations as an additional resource to strengthen existing smoothing methods for the language model for IR. The social annotations can benefit the smoothing of language model in two aspects: 1) the annotations themselves can serve as the summaries of the web pages given by the users; 2) the annotations can be seen as the links of the web pages sharing the same annotations. We propose three smoothing methods, addressing the two aspects and their combination, respectively. We call the new language model of using the proposed smoothing methods 'Language Annotation Model (LAM). Preliminary experimental results show that LAM significantly outperforms the traditional language models.