Text similarity by using GloVe word vector representations

Word embeddings are word representations in the form of vectors that allow to maintain certain semantic information of the words. There exist different ways of taking profit of the semantic information the words have, as there exist different ways of generating the word vectors that represent those...

Descripción completa

Detalles Bibliográficos
Autor: Sánchez Rodríguez, Iván
Tipo de recurso: tesis de maestría
Fecha de publicación:2017
País:España
Institución:Universitat Politècnica de València (UPV)
Repositorio:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:inglés
OAI Identifier:oai:riunet.upv.es:10251/90045
Acceso en línea:https://riunet.upv.es/handle/10251/90045
Access Level:acceso abierto
Palabra clave:Similitud entre textos
Vectores de palabras en español
Similitud semántica
Diferencia semántica
Representación vectorial de frases
Word vector representations
Word embeddings
Text similarity
Spanish word embeddings
Semantic difference
Phrase embeddings
Phrase similarity
Global Vectors
GloVe
Similitud entre frases
LENGUAJES Y SISTEMAS INFORMATICOS
Máster Universitario en Inteligencia Artificial, Reconocimiento de Formas e Imagen Digital-Màster Universitari en Intel·ligència Artificial, Reconeixement de Formes i Imatge Digital
Descripción
Sumario:Word embeddings are word representations in the form of vectors that allow to maintain certain semantic information of the words. There exist different ways of taking profit of the semantic information the words have, as there exist different ways of generating the word vectors that represent those words (e.g. Word2Vec model vs. GloVe model). By using the semantic information the word embeddings capture, we can build approximations to compare semantic information between phrases or even documents instead of words. In this project, we propose the use of the GloVe tool, presented by Stanford University, to train Spanish word embeddings, use them to compare semantic differences between Spanish phrases and compare the accuracy of the system with prior results in which other models were used, for example, Word2Vec.