Learning Knowledge from User Search

Learning Knowledge from User Search Yen-­‐‑Kuan Lee, Kun-­‐‑Ta Chuang Dept. of Computer Science and Information Engineering, National Cheng Kung University E-­‐‑mail:yklee@netdb.csie.ncku.edu.tw, ktchuang@mail.ncku.edu.tw Abstract In this paper, we introduce the concept of a novel application, called Knowledge Learning from User Search, aiming at identifying timely new knowledge triples from user search log. In the literature, the need of knowledge enrichment has been recognized as the key to the success of knowledge-based search. However, previous work of automatic knowledge extraction, such as Google Knowledge Vault, attempt to identify the unannotated knowledge triples from the full web-scale content in the offline execution. In our study, we show that most people demand a specific knowledge, such as the marriage between Brad Pitt and Angelina Jolie, soon after the information is announced. Moreover, the number of queries of such knowledge dramatically declines after a few days, meaning that the most people cannot obtain the precise knowledge from the execution of the offline knowledge enrichment. To remedy this, we propose the SCKE framework to extract new knowledge triples which can be executed in the online scenario. We model the ’Query-Click Page’ bipartite graph to extract the query correlation and to identify cohesive pairwise entities, finally statistically identifying the confident relation between entities. Our experimental studies show that new triples can also be identified in the very beginning after the event happens, enabling the capability to provide the up-to-date knowledge summary for most user queries. Introduction The technology of Knowledge Bases, abbreviated as KB, is recently highlighted by The 2015 Conference on Computational Linguistics and Speech Processing ROCLING 2015, pp. 248-262  The Association for Computational Linguistics and Chinese Language Processing 248 Internet giants such as Google, Yahoo and so on. For example, the Knowledge Graph project 1 , announced by Google at 2012, attempts to integrate the semantic information from KB into the search engine, enabling the capability of Question- Answering for specific queries. Currently, when we issue ’Avatar Director’, the exact answer ’James Cameron’ will be revealed as the conspicuous block in the search result. As the evolution of the interactive interface moving toward small screens (such as smart phones or wearable devices), the search content with lots of relevant URLs is no longer considered as an effective manner. It is believed that the QA- based search engine is the key ingredient of the next-generation information technology. However, the public large-scale knowledge bases, such as crowd-based Dbpedia [1], Freebase [3], NELL [4], and YAGO [26], have been reported to encounter the progressively slow growth of content [27]. The bottleneck implies that the human editing is no longer the effective manner for knowledge maintenance. Since the size of knowledge in current KBs still deviates far from completion, Google recently devotes to develop a systematic solution, called Knowledge Vault [9] (called KV for short), for the purpose of knowledge enrichment. As the full scan of Google-indexed pages, the KV framework is able to annotate new relation between two known entities such as people or movies. The design …

Content truncated to 500 words