Vitrià i Marca, JordiJulià Carrillo, Oriol2017-04-062017-04-062016-06-26https://hdl.handle.net/2445/109443Treballs Finals de Grau de Matemàtiques, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2016, Director: Jordi Vitrià i MarcaLatent Dirichlet Allocation (LDA) are a suite of algorithms that are often used for topic modeling. We study the statistical model behind LDA and review how tensor methods can be used for learning LDA, as well as implement a variation of an already existing method. Next, we present an innovative algorithm for temporal topic modeling and provide a new dataset for learning topic models over time. Last, we create a visualization for the word-topic probabilities.59 p.application/pdfengcc-by-nc-nd (c) Oriol Julià Carrillo, 2016http://creativecommons.org/licenses/by-nc-nd/3.0/esTractament del llenguatge natural (Informàtica)Treballs de fi de grauAprenentatge automàticProbabilitatsAlgorismes computacionalsNatural language processing (Computer science)Bachelor's thesesMachine learningProbabilitiesComputer algorithmsinfo:eu-repo/semantics/bachelorThesisinfo:eu-repo/semantics/openAccess