Accelerating and evaluation of syntactic parsing in natural language question answering systems
Zhe Chen, Dunwei Wen · 2007
cation systems. For example, NLP could be used in Question Answering (QA) systems to understand users ’ natural language inputs, and communicate with users in a natural way, such as LUNAR [1] and some service systems [2]. These applications have greatly improved the way users interact with computer systems and overcome the disadvantages of traditional QA systems which use pattern matching algorithms, for example ALICE [3]. However, with the development of NLP technology, a big problem has emerged. Most researchers spend a lot of time thinking of how to improve the precision of Part-of-Speech (POS) taggers and syntactic parsers, but there are few researches on how to save CPU time in tagging and parsing without precision decrease. Actually nowadays, NLP is applied more and more in real-time QA systems, such as dialogue, web search, cell phone and PDA etc. [4]. As a result, the processing time problem becomes more and more important for NLP applications, because users need the responses to their requests in an acceptable length of time. In fact, the speed problem is the very reason why most QA systems choose pattern matching algorithm but not NLP methods. Then, how to accelerate the parsing speed of syntactic parser? A NLP system always includes several parts, such as a stemmer module, a word tagging module, and a syntactic parsing module etc. Many algorithms have been proposed for these modules. As we know, the syntactic parsing takes most of the processing time. So, improving syntactic parsing is one of the most important methods, and the optimization of other modules is also necessary. Syntactic patterns are needed in syntactic parsing module. But it is possible for humans to construct a syntactic pattern. Firstly, it is hard to define a large amount of syntactic patterns. SecarXiv:0903.0174v1