Neural network implementation for the semantic interpretation of sentences

• Main Author: Sandigan, Amor Celeste
• Format: text
• Language: English
• Published: Animo Repository 1996
• Subjects: Neural networks; Language and languages > Data processing; Semantics > Network analysis; Grammar; Comparative and general > Sentences; Algorithms; Translating and interpreting; Information theory; Engineering
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/1756
• Institution: De La Salle University

This thesis presents a fundamental step towards the development of a computer system capable of understanding natural language. It introduces a new concept of empowering a computer to learn and distinguish the appropriate meaning of each word used in an English sentence. This process, known in linguistics as the semantic interpretation of sentences, is the main theme of the thesis. The study tackled this concept of semantic interpretation using the theory of neural networks. It discussed how a sentence may be encoded in a computer's memory and how the corresponding codes may be processed to come up with the correct interpretation of the compromising words in a manner very similar to how, we, human beings interpret them. The study tackled this approach further by describing the details of the algorithms used in implementing the artificial neurons, as well as the underlying principles employed in training the network. Finally, the study provided an analysis of the results of the implementation, its viability and adaptability with the other sentence processing methods. With the full implementation of this semantic interpreter concept, and if successfully combined with other fundamental sentence processing schemes such as case role representation, grammatical analysis, verb tense transformations, etc., the day is not far when man could fully communicate with computers and other intelligent machines using man's own spoken language.