Conclussion |
---|
Natural Language Processing has taken a prominent position in the area of Artificial Intelligence. The study of how computer models can better understand language allows for the design and implementation of systems that can interact with humans in language-like dialog.
Researchers have managed to use Neural Networks for different types of Natural Language Processing. The topologies and learning strategies used in these experiments have been varied, and finding an optimal configuration is still an open question. In some cases, initial weight values and local minima have been reported to affect the system's performance. In my research, I will be using Genetic Algorithms to search for better sets of configurations for a particular language learning task. By studying the configurations chosen by the Genetic Algorithm as it progresses to Neural Networks with higher fitness values, I hope to be able to identify those parameters that become critical for the task at hand and what values of these parameters allow the Network to perform better. This information should provide a better understanding of how it is that the networks are achieving their goals and allow for better future network configurations. Such systems can lead to better man-machine interfaces and allow for a better understanding of the abilities and limitations of different Neural Network models in their relationship to the task of natural language processing.
|