Publication: El modelo cortical HTM y su aplicación al conocimiento lingüístico
Loading...
Official URL
Full text at PDC
Publication Date
2022-04-21
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Complutense de Madrid
Citations
Exportar
Abstract
El problema que aborda este trabajo de investigación es encontrar un modelo neurocomputacional de representación y comprensión del conocimiento léxico, utilizando para ello el algoritmo cortical HTM, que modela el mecanismo según el cual se procesa la información en el neocórtex humano. La comprensión automática del lenguaje natural implica que las máquinas tengan un conocimiento profundo del lenguaje natural, lo que, actualmente, está muy lejos de conseguirse. En general, los modelos computacionales para el Procesamiento del Lenguaje Natural (PLN), tanto en su vertiente de análisis y comprensión como en la de generación, utilizan algoritmos fundamentados en modelos matemáticos y lingüísticos que intentan emular la forma en la que tradicionalmente se ha procesado el lenguaje, por ejemplo, obteniendo la estructura jerárquica implícita de las frases o las desinencias de las palabras. Estos modelos son útiles porque sirven para construir aplicaciones concretas como la extracción de datos, la clasificación de textos o el análisis de opinión. Sin embargo, a pesar de su utilidad, las máquinas realmente no entienden lo que hacen con ninguno de estos modelos. Por tanto, la pregunta que se aborda en este trabajo es si, realmente, es posible modelar computacionalmente los procesos neocorticales humanos que regulan el tratamiento de la información de tipo semántico del léxico. Esta cuestión de investigación constituye el primer nivel para comprender el procesamiento del lenguaje natural a niveles lingüísticos superiores...
The problem addressed by this research work is that of finding a neurocomputational representation and comprehension model of lexical knowledge, using for that purpose the HTM cortical algorithm, which models the mechanism according to which information is processed in the human neocortex.The automatic understanding of natural language implies that machines have a deep knowledge of natural language, which is currently far from being achieved. In general, computational models for Natural Language Processing (NLP), both in their analysis and comprehension aspects as well as in the generation aspect, use algorithms based on mathematical and linguistic models that try to emulate the way in which language has traditionally been processed, for example, by obtaining the implicit hierarchical structure of sentences or word endings. These models are useful because they serve to build concrete applications such as data mining, text classification or sentiment analysis. However, despite their usefulness, machines do not really understand what they are doing with any of these models. The question, therefore, addressed in this thesis is whether it is really possible to computationally model the human neocortical processes that regulate the processing of lexical semantic information. This research question constitutes the first level of understanding the processing of natural language at higher linguistic levels...
The problem addressed by this research work is that of finding a neurocomputational representation and comprehension model of lexical knowledge, using for that purpose the HTM cortical algorithm, which models the mechanism according to which information is processed in the human neocortex.The automatic understanding of natural language implies that machines have a deep knowledge of natural language, which is currently far from being achieved. In general, computational models for Natural Language Processing (NLP), both in their analysis and comprehension aspects as well as in the generation aspect, use algorithms based on mathematical and linguistic models that try to emulate the way in which language has traditionally been processed, for example, by obtaining the implicit hierarchical structure of sentences or word endings. These models are useful because they serve to build concrete applications such as data mining, text classification or sentiment analysis. However, despite their usefulness, machines do not really understand what they are doing with any of these models. The question, therefore, addressed in this thesis is whether it is really possible to computationally model the human neocortical processes that regulate the processing of lexical semantic information. This research question constitutes the first level of understanding the processing of natural language at higher linguistic levels...
Description
Tesis inédita de la Universidad Complutense de Madrid, Facultad de Filología, leída el 07-03-2022