Constructing Linguistically Motivated Structures from Statistical Grammars

Proceedings of Recent Advances in Natural Language Processing, pages 63–69, Hissar, Bulgaria, 12-14 September 2011. Constructing Linguistically Motivated Structures from Statistical Grammars Ali Basirat NLP Lab, School of ECE, College of Engineering, University of Tehran, Tehran, Iran a.basirat@srbiau.ac.ir Heshaam Faili NLP Lab, School of ECE, College of Engineering, University of Tehran, Tehran, Iran hfaili@ut.ac.ir Abstract This paper discusses two Hidden Markov Models (HMM) for linking linguistically motivated XTAG grammar and the auto- matically extracted LTAG used by MICA parser. The former grammar is a detailed LTAG enriched with feature structures. And the latter one is a huge size LTAG that due to its statistical nature is well suited to be used in statistical approaches. Lack of an efficient parser and sparseness in the supertags set are the main obstacles in us- ing XTAG and MICA grammars respec- tively. The models were trained by the standard HMM training algorithm, Baum- Welch. To converge the training algo- rithm to a better local optimum, the ini- tial state of the models also were estimated using two semi-supervised EM-based al- gorithms. The resulting accuracy of the model (about 91%) shows that the models can provide a satisfactory way for linking these grammars to share their capabilities together. 1 Introduction Tree Adjoining-Grammar (TAG) is a tree generat- ing system that forms the object language by the set of derived trees (Joshi and Schabez, 1991). This formalism as a Mildly Context Sensitive Grammar is supposed to be powerful enough to model the natural languages (Joshi, 1985). In the lexicalized case (LTAG), each lexical item of the object language is associated with at least one elementary structure of the grammar called elementary tree. Each elementary tree in LTAGs can be considered as a complex descrip- tion of its anchor that provides a domain of lo- cality over which the anchor can specify syntac- tic and semantic constraints(Bangalore and Joshi, 1999). Extended domain of locality and factoring of recursion from the domain of dependency are the main key properties of using these grammars (Bangalore and Joshi, 1999). There are two ways for creating the set of ele- mentary trees (Faili and Basirat, 2010). The first method is the manual crafting of the elementary trees as it was done in the XTAG project (XTAG- Group, 2001). And the alternate one is the au- tomatically extraction of them from some anno- tated treebanks as it was done in (Xia, 2001; Chen, 2001). The result of the former method is a de- tailed LTAG that is enriched with semantic rep- resentation but suffers from the lack of statisti- cal information. The output of the latter one on the other hand, is a huge size LTAG that suffers from the sparseness problem in the elementary trees set but contains enough statistical informa- tion that make it suitable to be used in statistical approaches. The relatively huge size of the auto- matically extracted elementary trees set is an ob- stacle in annotating these structures with semantic representation (Chen, 2001). One of the negative …

Content truncated to 500 words