Efficient Sentence Parsing with Language Specific Features: A case Study of Czech

This paper presents two optimalizations of standard parsing techniques applied to Czech as a representative of a free word order language with rich morphology. Our approach is based on features that take advantage of particular knowledge of language specific properties. The first method involves a special type of merging possible due to the agreement fulfillment. Another use of language specific features lies in the exploitation of verb valency frames for probabilistic ordering of parsing output.

• Type: Conference
• Year: 2001
• Where: IWPT
• Authors: Ales Horák, Pavel Smrz
• Keywords: IWPT 2001, Language Specific Properties, Standard Parsing Techniques, System Software, Word Order Language