Publication: Implementación y evaluación de rendimiento de una aplicación para procesamiento de lenguaje natural
Loading...
Official URL
Full text at PDC
Publication Date
2021
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
El Procesamiento del Lenguaje Natural (NLP) es el campo de conocimiento de la Inteligencia Artificial encargado de investigar la manera de comunicar las máquinas con las personas mediante el uso de lenguas naturales, como el español, el inglés o el chino. Una de las utilidades más conocidas del NLP pueden ser los asistentes virtuales o los chatbots, pero no son las únicas. Además, es importante entender que el NLP no dota de inteligencia a un chatbot, solo le da la capacidad de procesar y generar lenguaje humano. Sin embargo, existen otras muchas áreas de interés en las que se aplica de forma intensiva NLP, entre las que destacan, la comprensión de lenguaje natural, generación de lenguaje natural, recuperación de información, reconocimiento y síntesis del habla, traducción automática, resumen automático de textos o detección de sentimientos y emociones. Muchas de estas aplicaciones, además, requieren un tiempo de respuesta reducido que, habitualmente, no permite el procesamiento remoto de los datos en grandes centros de procesamiento de datos (en la nube), sino en un dispositivo cercano a la fuente de la información (por ejemplo, un dispositivo móvil o un dispositivo en el borde de la red –Edge Computing–). El trabajo propone la implementación de algoritmos para procesamiento de lenguaje natural tomando como base implementaciones ya existentes (por ejemplo, BERT), y sus versiones reducidas para dispositivos móviles (mobileBERT), así como la evaluación de su rendimiento en términos de tiempo de respuesta en diferentes entornos. El objetivo final es realizar un estudio sobre el rendimiento de diferentes modelos con diferentes tamaños de entrada en diferentes entornos de trabajo, para determinar la eficiencia de cada uno y compararla con el resto para obtener la mayor eficiencia posible.
Natural Languaje Proccesing (NLP) is the field of knowledge of Artificial Intelligence that deals with investigating the way to communicate with people through the use of natural languajes, such as Spanish, English or Chinese. Virtual assisteants or chatbots are one of the best known NLP utilities, but they are not the only one. In addition, it is important to understand that the NLP does not endow a chatbot with intelligence, it only gives it the ability to process and generate human languaje. However, there are many other areas of interest in where NLP is applied intensively, including natural languaje comprehension, natural languaje generation, information retrieval, speech recognition and synthesis, automatic translation, automaticsummarization of texts or detection of feelings and emotions. Many of these applications also require a reduced response time that usually does not allow remote data processing in large data processing centers (in the cloud), but in a device close to the source of the information, (for example, a mobile device or a device on the edge of the network –Edge Computing–). The project proposes the implementation of algorithms for natural languaje processing based on existing implementations (for example BERT) and their reduced versions for mobile devices (mobileBERT), as well as the evaluation of their perfomance in terms of response time over a set of mobile devices (for example Coral’s Edge TPU). The final objective is to ocarry out a study on the perfomance of different models with different input sizes in different working enviroments, to determine the efficiency of each one and compare it with the other ones to obtain the highest possible efficiency
Natural Languaje Proccesing (NLP) is the field of knowledge of Artificial Intelligence that deals with investigating the way to communicate with people through the use of natural languajes, such as Spanish, English or Chinese. Virtual assisteants or chatbots are one of the best known NLP utilities, but they are not the only one. In addition, it is important to understand that the NLP does not endow a chatbot with intelligence, it only gives it the ability to process and generate human languaje. However, there are many other areas of interest in where NLP is applied intensively, including natural languaje comprehension, natural languaje generation, information retrieval, speech recognition and synthesis, automatic translation, automaticsummarization of texts or detection of feelings and emotions. Many of these applications also require a reduced response time that usually does not allow remote data processing in large data processing centers (in the cloud), but in a device close to the source of the information, (for example, a mobile device or a device on the edge of the network –Edge Computing–). The project proposes the implementation of algorithms for natural languaje processing based on existing implementations (for example BERT) and their reduced versions for mobile devices (mobileBERT), as well as the evaluation of their perfomance in terms of response time over a set of mobile devices (for example Coral’s Edge TPU). The final objective is to ocarry out a study on the perfomance of different models with different input sizes in different working enviroments, to determine the efficiency of each one and compare it with the other ones to obtain the highest possible efficiency
Description
Trabajo de Fin de Grado en Ingeniería del Software, Facultad de Informática UCM, Departamento de Arquitectura de Computadores y Automática, Curso 2020/2021