Intent-Based Categorization of Search Results Using Questions from Web Q&A Corpus

Intent feature discovery using Q&A corpus and web data

“…Experimental results of our previous work [18] demonstrated the efficiency of intent-based categorization using manual evaluation of search results included. We use this method as the baseline.…”

“…Automatic intent feature discovery has been tried using its matches with search result information [18]. However, in that case the relevancy had to be judged manually and semantic portions of intent features have not been sufficiently considered.…”

“…For extracting features of intent concerned with query, we need to analyze the Web which can be treated as a reflection of Web search intents. Yoon et al [18] showed that questions in Web Q&A corpus are good expressions to reveal questioners' intents, so we use these questions coordinated with query to collect candidate intent features. By sending a query to Web Q&A corpus, we can receive useful hints and topics connected to questions that are basis of our intent prediction process.…”

“…We thus decide to use semantic analysis methods to extract intent expressions using questions. Head-noun set extraction method [18] is used in order to detect the core meanings of questions, and regard them as candidate intent features.…”

“…In addition, the past data is useful enough to find statistically dominant intent of a query, but not sufficient enough to cover the variety of intents. 1 For information providers, as well as the users themselves, it is difficult to express intent precisely because they may not possess sufficient knowledge to 'imagine' and 'conceptualize' their needs and, moreover, the intent may change following various conditions such as querying time, search environment and acquired knowledge in the previous search sessions [18]. When users lack the knowledge or contextual awareness to formulate queries or navigate complex information spaces, the search task requires browsing and exploration [16].…”