چكيده به لاتين
Semantic role labeling (SRL) is a crucial task in natural language processing that involves identifying the semantic roles played by each constituent in a sentence. Two widely-used resources for semantic role labeling are Proposition Bank and FrameNet, which provide annotated data for predicate-argument structures and frames and their frame elements, respectively. However, while Proposition Bank focuses on the syntactic structure of predicates and their arguments, FrameNet emphasizes the semantic relations between frames and their elements. Mapping Proposition Bank arguments to FrameNet's frame elements can enhance the accuracy of semantic role labeling and improve the performance of NLP applications.
This thesis proposes a methodology for mapping Proposition Bank arguments to FrameNet frame elements and applies this methodology to a corpus of text in order to analyze the distribution and frequency of semantic roles in the text. The methodology involves using existing tools and algorithms to annotate the text with Proposition Bank and FrameNet annotations, mapping the Proposition Bank arguments to FrameNet frame elements based on their semantic similarity, and evaluating the accuracy of the mapping using standard metrics. The results of the analysis reveal the distribution and frequency of semantic roles in the corpus and provide insights into the meaning of language.
The thesis also discusses the limitations and challenges of the proposed methodology, including the need for additional annotated data and the difficulty of handling ambiguous and polysemous words. Overall, this research contributes to the development of more accurate and efficient NLP applications and provides new perspectives on the relationship between Proposition Bank and FrameNet for semantic role labeling.