| In natural language, different sentences can express the same meaning, or a sentence can be modified without altering its meaning. This is called semantic equivalence". Semantic equivalence can be resolved humans, but it is still an unresolved problem for computerized systems. Shallow semantics is used in this research to recognize semantic equivalence in sentences in English, which makes the approach domain-independent. A case-based system is developed, which uses the Stanford Natural Language parser to obtain grammatical information and looks for patterns identified in each one of the cases. A modified version of the Microsoft Research Paraphrase corpus (MSRP) with 451 sentence pairs was utilized to test the system. An average rate of 89.80% of successful equivalence detection was obtained, which compares favorably with the success rate between 63.94% to 74.00% reported in the literature. The main contributions of this thesis are the development of a system to determine semantic equivalence among sentences using shallow semantics, the identification of the particular structures of the dependencies generated by the parser for each case and the development of a corpus of semantic equivalence sentences. |
