Web Search Clustering and Labeling with Hidden Topics

Abstract

Web search clustering is a solution to reorganize search results (also called "snippets") in a more convenient way for browsing. There are three key requirements for such post-retrieval clustering systems: (1) the clustering algorithm should group similar documents together; (2) clusters should be labeled with descriptive phrases; and (3) the clustering system should provide high-quality clustering without downloading the whole Web page.This article introduces a novel framework for clustering Web search results in Vietnamese which targets the three above issues. The main motivation is that by enriching short snippets with hidden topics from huge resources of documents on the Internet, it is able to cluster and label such snippets effectively in a topic-oriented manner without concerning whole Web pages. Our approach is based on recent successful topic analysis models, such as Probabilistic-Latent Semantic Analysis, or Latent Dirichlet Allocation. The underlying idea of the framework is that we collect a very large external data collection called "universal dataset," and then build a clustering system on both the original snippets and a rich set of hidden topics discovered from the universal data collection. This can be seen as a richer representation of snippets to be clustered. We carry out careful evaluation of our method and show that our method can yield impressive clustering quality.

Labeling clusters from both linguistic and statistical perspectives: A hybrid approach

“…Tseng's approach [19] labels clusters by mapping category terms to generic terms. Before this research, rules are used in Nguyen's [13] work to find readable phrases. However, the rules used in these papers are mostly lexical rules that cannot cover syntactical features.…”

“…Keywords are used in some existing research but a single term rarely gives users enough information. Existing research has reported that phrases [10,13] are more informative than keywords for understanding. However, the readability of phrases is rarely studied in existing research because it is very difficult to formalize the measurement of readability for phrases.…”

DF-Miner: Domain-specific facet mining by leveraging the hyperlink structure of Wikipedia

“…The basic idea is that if an n-word sequence tends to appear together, the sequence is more likely to be an n-word phrase. The phrases extracted from a text collection are considered as the label candidates of the text collection [37,38]. Mei et al [39] and Lau et al [40] reported that twoword phrases (bigrams) usually work better than other n-word phrases for label generation.…”

Clustering Web Search Results with Maximum Spanning Trees

“…Very little further work on this topic has been done: vector-based WSI was successfully shown to improve bag-of-words adhoc Information Retrieval [36] and experimental studies [10] have provided interesting, though preliminary, insights into the use of WSI for Web search result clustering. More recently the use of hidden topics has been proposed to identify query meanings [29]. However, topics -estimated from a universal dataset -are query-independent and thus their number needs to be found beforehand.…”