Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/132207
Title: Refinament de l'anàlisi del sentiment per xarxes neuronals mitjançant tècniques de PLN
Author: Fraire Ferrer, Miquel Àngel
Tutor: Isern, David  
Others: Ventura, Carles  
Abstract: Both machine learning and basic natural language processing techniques come together in the sentiment analysis. The latter are usually applied mostly to adapt the input text corpus to the data types the model requires, as well as to reduce its dimensionality without losing relevant information, which allows the classifying algorithm to gain efficiency. However, a deeper knowledge of the linguistic particularities of the corpus would allow to carry out some actions in order to increase the degree of accuracy in the sentiment analysis results. Particularly, using the technique of Part-of-Speech tagging to assign more weight to those words with more semantic load during the algorithm training, should turn into improved ranking results. To that end, three different textual corpora have been selected and subjected to the usual operations in a sentiment analysis exercise, and then the results have been measured. Then, after applying the aforementioned technique to the original corpora, the model has been trained again. Subsequently, the accuracy of Part-of-Speech tagging has also been increased through techniques for recognizing and correcting informal language. Finally, the word frequency distribution has been calculated in order to know the differences in semantic weight amongst corpora, which can explain an incomprehensibly divergent or distant outcome, and occasionally correct these deviations. The conclusion is that the method of assigning more weight to semantically relevant words by means of Part-of-Speech tagging is especially effective in themed corpora not containing words of totally biased frequency towards a certain category.
Keywords: sentiment analysis
neural networks
natural language processing
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: 8-Jun-2021
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 

frairefmVideo2021.mp4

167,91 MBMP4View/Open
frairefmTFG0621memòria.pdfMemòria del TFG2,18 MBAdobe PDFThumbnail
View/Open
frairefmTFG0621presentació.pdfPresentació del TFG1,77 MBAdobe PDFThumbnail
View/Open