Fast structural query with application to chinese treebank sentence retrieval

Abstract

In natural language processing a huge amount of structured data is constantly used for the extraction and presentation of grammatical structures in sentences. For example the Chinese Treebank corpus developed at the Institute of Information Science Academia Sinica Taiwan is a semantically annotated corpus that has been used to help parse and study Chinese sentences. In this setting users usually use structured tree patterns instead of keywords to query the corpus.In this paper we present an online prototype system that provides exploratory search ability. The system implements two flexible and efficient structural query methods and employs a user-friendly web-based interface. Although the system adopts the XML format to present the corpora and search results it does not use conventional XML query languages. As searching the Chinese Treebank corpora is structural in nature and often deals with structural similarities conventional XML query languages such as XPath and XQuery are inflexible and inefficient. We propose and implement a query algorithm called Parent-Child Relationship Filter (PCRF) which provides flexible and efficient structural search. PCRF is sufficiently flexible to provide several similarity-matching options such as wildcard unordered sibling sub-trees ancestor-descendant matching and their combinations. In addition PCRF supports stream-based matching to help users query their XML documents online. We also present three accelerating rules that achieve a 1.5- to 8-fold performance improvement in query time. Our experiment results show that our method archive a 10- to 1000-fold performance improvement compared to the usual text-based XPath query method.