Bibliographic citations
This is an automatically generated citacion. Modify it if you see fit
Utia, J., (2020). Representación vectorial de relación de hiponimia e hiperonimia en español [Tesis, Pontificia Universidad Católica del Perú]. http://hdl.handle.net/20.500.12404/16962
Utia, J., Representación vectorial de relación de hiponimia e hiperonimia en español [Tesis]. PE: Pontificia Universidad Católica del Perú; 2020. http://hdl.handle.net/20.500.12404/16962
@mastersthesis{renati/25537,
title = "Representación vectorial de relación de hiponimia e hiperonimia en español",
author = "Utia Deza, Jose Vicente",
publisher = "Pontificia Universidad Católica del Perú",
year = "2020"
}
Title: Representación vectorial de relación de hiponimia e hiperonimia en español
Authors(s): Utia Deza, Jose Vicente
Advisor(s): Oncevay Marcos, Felix Arturo
Keywords: Computación semántica; Procesamiento de lenguaje natural (Computación); Enseñanza con ayuda de computadoras
OCDE field: https://purl.org/pe-repo/ocde/ford#1.02.00
Issue Date: 3-Sep-2020
Institution: Pontificia Universidad Católica del Perú
Abstract: Actualmente, gracias a Internet y a la Web se dispone de información casi ilimitada, la cual está representada a nivel de textos en su mayoría. Así, dado que acceder a estos textos en su mayoría es de libre acceso, nace el interés por su manipulación de una manera automatizada para poder extraer información que se considere relevante. El presente trabajo de investigación se ubica dentro de la detección automática de relaciones léxicas entre palabras, que son relaciones que se establecen entre los significados de las palabras tal como se consigna en el diccionario. En particular, se centra en la detección de relaciones de hiponimia e hiperonimia, debido a que éstas son relaciones de palabras en las que una de ellas engloba el significado de otra o viceversa, lo cual podría considerarse como categorización de palabras. Básicamente, el método propuesto se basa en la manipulación de una representación vectorial de palabras denominado Word Embeddings, para resaltar especialmente áquellas que tengan relación jerárquica, proceso que se realiza a partir de textos no estructurados. Tradicionalmente, los Word Embeddings son utilizados para tareas de analogía, es decir, para detectar relaciones de sinonimia, por lo que se considera un poco más complejo utilizar estos vectores para la detección de relaciones jerárquicas (hiperonimia e hiponimia), por consecuencia se proponen métodos adicionales para que, en conjunto con los Word Embeddings, se puedan obtener resultados eficientes al momento de detectar las relaciones entre distintos pares de palabras.
Currently, thanks to the Internet and Web, almost unlimited information is available, which is mostly represented at text level. Thus, given that access to these texts is mostly freely available, interest in their manipulation is born in an automated way to extract information that is considered relevant. The present research work is located within the automatic detection of lexical relations between words, which are relations that are established between the meanings of words as it is stated in the dictionary. In particular, it focuses on the detection of hyponymy and hyperonymy relationships, because these are word relationships in which one of them encompasses the meaning of another or vice versa, which could be considered as categorization of words. Basically, the proposed method is based on the manipulation of Word Embeddings to highlight especially words that have a hierarchical relationship, a process that is carried out from unstructured texts. Traditionally, Word Embeddings are used for analogy tasks, that is, to detect synonymy relationships, so it is considered a bit more complex to use these vectors for the hierarchical relationships (hyperonimia and hyponymy) detection, therefore, additional methods are proposed, so in conjunction with the Word Embeddings, efficient results can be obtained when detecting the relationships between different pairs of words.
Currently, thanks to the Internet and Web, almost unlimited information is available, which is mostly represented at text level. Thus, given that access to these texts is mostly freely available, interest in their manipulation is born in an automated way to extract information that is considered relevant. The present research work is located within the automatic detection of lexical relations between words, which are relations that are established between the meanings of words as it is stated in the dictionary. In particular, it focuses on the detection of hyponymy and hyperonymy relationships, because these are word relationships in which one of them encompasses the meaning of another or vice versa, which could be considered as categorization of words. Basically, the proposed method is based on the manipulation of Word Embeddings to highlight especially words that have a hierarchical relationship, a process that is carried out from unstructured texts. Traditionally, Word Embeddings are used for analogy tasks, that is, to detect synonymy relationships, so it is considered a bit more complex to use these vectors for the hierarchical relationships (hyperonimia and hyponymy) detection, therefore, additional methods are proposed, so in conjunction with the Word Embeddings, efficient results can be obtained when detecting the relationships between different pairs of words.
Link to repository: http://hdl.handle.net/20.500.12404/16962
Discipline: Informática con mención en Ciencias de la Computación
Grade or title grantor: Pontificia Universidad Católica del Perú. Escuela de Posgrado
Grade or title: Maestro en Informática con mención en Ciencias de la Computación
Register date: 3-Sep-2020
This item is licensed under a Creative Commons License