Author profiling in social media with multimodal information

Determine aspects of a person as gender, age, residency, occupation, among others, through his/her texts is a task that is part of the natural language processing and is known as author profiling. In this thesis work, we propose a solution for the task of profiling authors in social networks. Our so...

Descripción completa

Detalles Bibliográficos
Autor: Miguel Ángel Alvarez Carmona
Tipo de recurso: tesis doctoral
Estado:Versión aceptada para publicación
Fecha de publicación:2019
País:México
Institución:Instituto Nacional de Astrofísica, Óptica y Electrónica
Repositorio:Repositorio Institucional del INAOE
Idioma:inglés
OAI Identifier:oai:inaoe.repositorioinstitucional.mx:1009/1686
Acceso en línea:http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/1686
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/Inspec/Natural languaje processing
info:eu-repo/classification/Inspec/Textual classification
info:eu-repo/classification/Inspec/Author profiling
info:eu-repo/classification/Inspec/Multimodal information
info:eu-repo/classification/Inspec/Images representation
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/12
info:eu-repo/classification/cti/1203
info:eu-repo/classification/cti/120323
Descripción
Sumario:Determine aspects of a person as gender, age, residency, occupation, among others, through his/her texts is a task that is part of the natural language processing and is known as author profiling. In this thesis work, we propose a solution for the task of profiling authors in social networks. Our solution uses a multimodal approach to extracting information from written messages and images shared by users. Previous work has shown the existence of useful information for this task in these modalities; however, our proposal goes further demonstrating the complementarity of the modalities when merging these two sources of information. To do this, we propose to map images in a text, and with that, to have the same framework of representation through which to achieve the fusion of information. Our work explores different methods for extracting information either from the text or from the images. To represent the textual information, different distributional term representations approaches were explored in order to identify the topics addressed by the user. For this purpose, an evaluation framework was proposed in order to identify the most appropriate method for this task. To represent visual information, approaches were explored to convert an image into a set of descriptive terms. The results show that the textual descriptions of the images contain information for the author profiling task, and the fusion of textual information with information extracted from the images increases the accuracy of this task.