Master Dissertation : Information Retrieval for Question Answering based on Distributed Representations
Commonly used methods for information retrieval such as TFIDF do not capture the semantics of the query or the document. This is a problem, especially in cases where the words used in the queries are not contained in the documents. Therefore more research needs to be done to investigate how text sem...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis de maestría |
| Fecha de publicación: | 2022 |
| País: | España |
| Institución: | Universidad Nacional de Educación a Distancia |
| Repositorio: | e-spacio. Repositorio Institucional de la UNED |
| Idioma: | inglés |
| OAI Identifier: | oai:e-spacio.uned.es:20.500.14468/14297 |
| Acceso en línea: | https://hdl.handle.net/20.500.14468/14297 |
| Access Level: | acceso abierto |
| Palabra clave: | 1203 Ciencia de los ordenadores |
| Sumario: | Commonly used methods for information retrieval such as TFIDF do not capture the semantics of the query or the document. This is a problem, especially in cases where the words used in the queries are not contained in the documents. Therefore more research needs to be done to investigate how text semantics can be applied to information retrieval, especially in cases where the corpus of documents is big and the queries and documents representations need to be compared fast and without the need of re-indexing. In this work, we conduct an exploratory study to investigate different embeddings and deep learning techniques and how this can be applied to the information retrieval task. We show that although existing methods based on word overlapping perform better in general, in particular cases where the word overlap between queries and documents is low, the use of semantic embedding outperforms other methods based on bag of words. |
|---|