Publication: Análisis emocional para la inclusión digital
Loading...
Files
Official URL
Full text at PDC
Publication Date
2018
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
En la actualidad, gracias a Internet, tenemos acceso fácil y rápido a una cantidad enorme de información. Lo único que necesitamos es interpretar esta información correctamente para saber si es útil para nosotros o no y de qué manera. En este punto surge un problema porque no todas las personas tienen la misma facilidad para interpretar un texto, por ejemplo, la falta de empatía emocional en las personas con Trastornos del Espectro Autista (TEA) les impide reconocer el contenido afectivo de un texto lo que puede llegar a alterar en gran medida su significado. Algo similar ocurre cuando estas personas quieren escribir algo en la red: la dificultad para expresar sus emociones puede provocar malentendidos. Es necesario desarrollar herramientas que permitan la integración digital de personas con este tipo de discapacidad.
La computación afectiva ha ayudado a conseguir grandes avances en este ámbito llegando a permitir identificar el estado emocional del sujeto a través de diferentes fuentes: voz, texto, expresiones faciales... En el caso de un texto se pueden reconocer las emociones contenidas en este mediante diversas técnicas basadas en analizar aspectos de su sintaxis y semántica: el tipo de frase que lo forman, las palabras que contiene, el contexto que se puede deducir... La mayoría de estos avances se han realizado para el inglés, todavía queda mucho por hacer en castellano.
Este trabajo busca desarrollar una aplicación web que sea capaz de detectar automáticamente el grado de carga afectiva que tiene un texto. Representaremos la carga afectiva de un texto mediante las categorías emocionales que representan las cinco emociones básicas: tristeza, miedo, alegría, enfado y asco. El resultado se mostrará en la interfaz de forma simple y gráfica, mediante emoticonos y colores.
Una vez finalizado, el trabajo fue sometido a una evaluación donde se realizaron pruebas internas con diferentes tipos de textos escogidos por nosotras y a una evaluación con usuarios finales para probar tanto la funcionalidad como la interfaz. El resultado de estas evaluaciones indica que aún queda trabajo por hacer pero hemos establecido una base muy sólida para el trabajo futuro.
Nowadays, thanks to the Internet, we have quick and easy access to a huge amount of information. All we need is to interpret this information correctly to know if it is useful for us. At this point a problem arises because not everyone has the same easy to interpret a text, for example, the lack of emotional empathy in people with Autism Spectrum Disorders (ASD) impedes them to recognize the affective content in a text which can greatly alter its meaning. Something similar happens when these people want to write something on the network: the difficulties they have when expressing their emotions can cause misunderstandings. It is necessary to develop tools which could allow the digital integration of people with this type of disability. Affective computing has helped to achieve great advances in this area, allowing us to identify the emotional state of a person through different sources: voice, text, facial expressions... In a text, the emotions it contains can by recognized through various techniques based on analyzing aspects of text's syntax and semantic: the type of phrases it is formed by, the words which it contains, the context that can be inferred... Most of these advances have been made for the English language, but there is still much to do in Spanish. This project aims to develop a web application that is capable of automatically detect the degree of the emotions of a text. We will represent the affective weight of a text through the emotional categories that represent the five basic emotions: sadness, fear, joy, anger and disgust. The result will be displayed in the interface in a simple and graphic way, using emoticons and colors. Once completed, the project was evaluated. Different types of texts were chosen by us to evaluate the accuracy of our emotional annotation and an evaluation with end-users was performed to test both the functionality and the interface. The result of these evaluations indicates that there is still work to be done but we have established a very solid foundation for future work.
Nowadays, thanks to the Internet, we have quick and easy access to a huge amount of information. All we need is to interpret this information correctly to know if it is useful for us. At this point a problem arises because not everyone has the same easy to interpret a text, for example, the lack of emotional empathy in people with Autism Spectrum Disorders (ASD) impedes them to recognize the affective content in a text which can greatly alter its meaning. Something similar happens when these people want to write something on the network: the difficulties they have when expressing their emotions can cause misunderstandings. It is necessary to develop tools which could allow the digital integration of people with this type of disability. Affective computing has helped to achieve great advances in this area, allowing us to identify the emotional state of a person through different sources: voice, text, facial expressions... In a text, the emotions it contains can by recognized through various techniques based on analyzing aspects of text's syntax and semantic: the type of phrases it is formed by, the words which it contains, the context that can be inferred... Most of these advances have been made for the English language, but there is still much to do in Spanish. This project aims to develop a web application that is capable of automatically detect the degree of the emotions of a text. We will represent the affective weight of a text through the emotional categories that represent the five basic emotions: sadness, fear, joy, anger and disgust. The result will be displayed in the interface in a simple and graphic way, using emoticons and colors. Once completed, the project was evaluated. Different types of texts were chosen by us to evaluate the accuracy of our emotional annotation and an evaluation with end-users was performed to test both the functionality and the interface. The result of these evaluations indicates that there is still work to be done but we have established a very solid foundation for future work.
Description
Trabajo de Fin de Grado en Ingeniería Informática (Universidad Complutense, Facultad de Informática, curso 2017/2018)