From Questions to Needs: Extracting Actionable Need Knowledge from the Web

The Web has become a source of information and knowledge accessed through search engines and various mining tools. In addition, it has become a platform on which not only software but also human users participate in and interact for various tasks. In an effort to mine actionable knowledge from the Web associated with various objects, we devised a text mining method specifically for extracting the actions often applied to various interest-triggering objects. Since the actions are extracted from how questions, they are considered needs or actions to be taken for the objects. We show that relatively simple yet powerful pattern matching techniques applied to the Web, in conjunction with the ODP taxonomy, can result in actionable need knowledge of high quality for a set of interest-triggering objects. To evaluate the quality of the extracted knowledge, we conducted two different types of testing: a controlled experiment with a small number of judges for desirability and an open experiment with the general public for comprehensibility.