Time-Aware Entity Search in DBpedia
Searching for entities is a common user activity on the Web. There is an increasing effort in developing entity search techniques in the research community. Existing approaches are usually based on static measures that do not reflect the time-awareness, w
- PDF / 128,803 Bytes
- 5 Pages / 439.37 x 666.142 pts Page_size
- 2 Downloads / 142 Views
Institute AIFB, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany {l.zhang,rettinger}@kit.edu, [email protected] 2 San Jose State University, San Jose, USA [email protected]
Abstract. Searching for entities is a common user activity on the Web. There is an increasing effort in developing entity search techniques in the research community. Existing approaches are usually based on static measures that do not reflect the time-awareness, which is a factor that should be taken into account in entity search. In this paper, we propose a novel approach to time-aware entity search in DBpedia, which takes into account both popularity and temporality of entities. The experimental results show that our approach can significantly improve the performance of entity search with temporal focus compared with the baselines.
1
Introduction
The ever-increasing quantities of entities in large knowledge bases on the Web, such as Wikipedia, DBpedia and YAGO, pose new challenges but at the same time open up new opportunities of information access on the Web. In this regard, many research activities involving entities have emerged in recent years. Entity search has become a major area of interest because users often search for specific entities instead of documents, which helps users to directly find the intended information. On the other hand, time-awareness is a crucial factor in entity search due to the high dynamics of the underlying data. In this paper, we address the problem of searching entities for a given user query in a time-aware setting. We formulate the time-aware entity search task as follows: given an entity collection E = {e1 , e2 , · · · , eN }1 , the input is a user query q = s, t, which consists of an entity name s and a time range of contiguous days t = {d1 , d2 , · · · , dM }, where di represents a specific day, and the output is the intended entity matching s of particular interest within t. Our approach allows users to restrict their search interests to a time range. However, in a reallife search scenario, users usually do not specify the time range explicitly. In this case, our system can easily use the current day on which users issue the query and a certain period of time before it (e.g., one week) as the time range. Assuming that users search for the entity name Irving on 2014-02-21 and the intended entity is Kyrie Irving, who won the NBA All-Star Game MVP Award 1
We use DBpedia as the entity collection, which extracts various kinds of structured information from Wikipedia and each DBpedia entity is tied to a Wikipedia article.
c Springer International Publishing Switzerland 2015 F. Gandon et al. (Eds.): ESWC 2015, LNCS 9341, pp. 175–179, 2015. DOI: 10.1007/978-3-319-25639-9 34
176
L. Zhang et al.
on 2014-02-17, the time range can be specified by our system as, for example, one week from 2014-02-15 to 2014-02-21. The challenge of our entity search task is name ambiguity, i.e., a entity name could refer to different entities. For entity linking [1,2], where the goal is to link words o
Data Loading...
