Sharing syntactic structures

MT Summit VII Sept. 1999 Sharing Syntactic Structures Masahiro Ueki Takenobu Tokunaga Tokyo Institute of Technology Tokyo Institute of Technology Hozumi Tanaka Tokyo Institute of Technology Abstract Bracketed corpora are a very useful resource for natural language processing, but hard to build efficiently, leading to quantitative insufficiency for practical use. Disparities in morphological information, such as word segmentation and part-of-speech tag sets, are also troublesome. An application spe- cific to a particular corpus often cannot be applied to another corpus. In this paper, we sketch out a method to build a corpus that has a fixed syntactic structure but varying morphological anno- tation based on the different tag set schemes utilized. Our system uses a two layered grammar, one layer of which is made up of replaceable tag-set-dependent rules while the other has no such tag set dependency. The input sentences of our system are brack- eted corresponding to structural informa- tion of corpus. The parser can work using any tag set and grammar, and using the same input bracketing, we obtain corpus that shares partial syntactic structure. 1 Introduction The ready availability of large corpora, especially brack- eted corpora, facilitates corpus-based research such as probabilistic parsing. However each corpus has its own part-of-speech tag sets and notation schemes. Corpus- based markup schemes can become customized to a specific corpus, and incompatible with other corpora with different tag sets or notation schemes. A number of morphological information mapping methods have been proposed [5] [6] [1]. Mapping sys- tems have rewrite rules that are derived automatical- ly or manually, and map part-of-speech tags word by word. But mapping between part-of-speech tags, for example noun to pronoun, cannot be performed be- cause of the large numbers of words with multiple parts-of-speech. Conventional rewrite rules consider both the word itself and its original part-of-speech tag in selecting all possible alternative tags. But. in prac- tical use, this method has drawbacks. We consider that dependencies between phrases coincide between tagging schemes, and should help to solve the above problem. Thus, given a structured corpus, we can use its structural information to rela- bel it according to a second tag set. In this paper, we sketch out a method to build a corpus that partially shares syntactic structures and has alternative part- of-speech tags for each word as stipulated by different part-of-speech tag sets. Sentences in the source corpus are bracketed to describe the basic phrase dependen- cies, and analyzed by a parser. The parser can work using any tag set and grammar, using the same input bracketing. As a result, we obtain an efficient corpus representation. 2 Our Method 2.1 Two Layered Grammar First of all, let us look at examples of a parse tree. (Figs. 1,2) These two trees describe possible syntactic struc- tures of the Japanese sentence kare ga watashi ni atarashi- i jisho wo kure ta “He gave me a new dictionary”, where the auxiliary ta is the past tense marker. One difference between the two parse trees is the interpre- …

Content truncated to 500 words