Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/81437
Title: Implementation of a spoken language system
Author: Perez Guijarro, Jessica
Tutor: Isern, David  
Others: Universitat Oberta de Catalunya
Ventura, Carles  
Abstract: This project consists of a implementation of a Spoken Language System that is part of the dialogue systems like Siri. The system is constituted by two independent blocks: Automatic Speech Recognition, in charge of identifying who is verbalizing the user and transforming a text, and the Spoken Language System, in charge of reading the text, identifying the significant parts of said text. Each of the compo- nents has been trained with different techniques and datasets since it does not share a common goal. In particular, for the development of the ASR module we have worked with a previously selected subset of the VoxForge English dataset, whose data has been trained using Hidden Markov Models for gener- ate the Acoustic Model. On the other hand, for the development of the SLU module we have worked with Recurrent Neural Networks and a variant of the ATIS dataset previously trained with the Word Embedding method. Although the precision obtained in all the components is more than acceptable, the performance of the integration of both components results unstable.
Keywords: automatic speech recognition
spoken language understanding
recurrent neural network
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: 5-Jun-2018
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 

TFG Presentacio¿n.mov

576,73 MBVideo QuicktimeView/Open
jperezguijaTFG0618memoria.pdfMemòria del TFG913,56 kBAdobe PDFThumbnail
View/Open
jperezguijaTFG0618presentación.pdfPresentación del TFG462,72 kBAdobe PDFThumbnail
View/Open