Subtask Mining from Search Query Logs for How-Knowledge Acceleration

Subtask Mining from Search Query Logs for How-Knowledge Acceleration Chung-Lun Kuo and Hsin-Hsi Chen Department of Computer Science and Information Engineering, National Taiwan University No. 1, Sec. 4, Roosevelt Road, Taipei, 10617 Taiwan E-mail: lounera@gmail.com; hhchen@ntu.edu.tw Abstract How-knowledge is indispensable in daily life, but has relatively less quantity and poorer quality than what-knowledge in publicly available knowledge bases. This paper first extracts task-subtask pairs from wikiHow, then mines linguistic patterns from search query logs, and finally applies the mined patterns to extract subtasks to complete given how-to tasks. To evaluate the proposed methodology, we group tasks and the corresponding recommended subtasks into pairs, and evaluate the results automatically and manually. The automatic evaluation shows the accuracy of 0.4494. We also classify the mined patterns based on prepositions and find that the prepositions like on, to, and with have the better performance. The results can be used to accelerate how-knowledge base construction. Keywords: Knowledge Acceleration; Query Log Mining; Subtask Mining. 1. Introduction Nowadays wikipedia and large scale knowledge bases such as YAGO (Suchanek et al., 2007), freebase (Bollacker et al., 2008) and DBpedia (Lehmann et al., 2015) are available for supporting what-knowledge. Users can find facts about entities and know their relationships from such a kind of knowledge bases. How-knowledge such as how to lose weight, how to rent a house, and how to fire lacks are different from the above. Given a how-to task, we would like to find several alternative ways to complete the task. How-to websites like wikiHow1 and eHow2 help how to do things in daily life. However, the quantity of wikiHow and the quality of eHow cannot compete with those of wikipedia. As we know, users often look for advices from the Web. They submit queries to search engines, click the results and browse the contents to meet their how-information needs. Users may have some idea about their tasks, e.g., to lose weight by eating only vegetables. Such a kind of queries is used to confirm the effectiveness of the subtask (e.g., eating only vegetables) and find its details. Users may further find more alternatives to complete the task by directly employing patterns such as “how to” + task description. Thus, potential how-knowledge is implicitly embedded in search query logs. It provides an opportunity to accelerate how-knowledge population by using search query logs. This paper aims at mining how-to patterns and applying them to extract subtasks to complete a given how-to task. This paper is organized as follows. Section 2 survey the related work. Section 3 specifies how to extract task-subtask pairs from wikiHow, proposes an algorithm to mine linguistic patterns from search query logs, and presents recommendation for given how-to tasks. Section 4 shows and discusses the experimental results. Section 5 concludes the remarks. 1 http://www.wikihow.com/Main-Page 2 http://www.ehow.com/ 2. Related Work Search query logs keep users’ search behaviors. Jones et al. (2008) define users’ search structures. Users may submit one or more related queries to meet their information needs. Lucchese et al. (2011) observe that …

Content truncated to 500 words