Resolving Ambiguity in Entity and Fact Extraction Through a Hybrid Approach

This thesis present two new techniques namely Unambiguous Entity Extraction (UEE) and Unambiguous Fact Extraction (UFE) to resolve ambiguiti in entity and fact extraction. Both techniques are obtained by hybriding 4 major theories and approaches namely, possibility theory, fuzzy sets, a knowledge-based approach, NLP techniques (syntactic and semantic processing). In this thesis, a word that is classified into to Noun part-of-speech is considered as an entity. An entity is ambiguous if it has more than one semantic. The UEE technique is designed and developed to assign the most possible semantic to the word. The technique was tested using 12 test cases with 111 sentences. The obtained results indicate that UEE technique is able to give precision rate of 85.7% and recall rate of 80.3%. On the other hands, UFE focuses on extracting an ambiguous fact from a sentence. A fact is a meaning that can
be formally represented as a statement and determined its truthfulness. A sentence contains an ambiguous fact if it can be interpreted into more than one meaning. The UFE technique is designed and developed to select the most possible fact by selecting the most possible meaning from a sentence. In evaluating UFE technique, test cases have been created and tested. Each test case consist of sentences in the range of 5 to 8. The obtained results in the form of predicate calculus are evaluated manually. The results suggest that UFE technique is successful.