Latent Dirichlet Allocation (LDA) Approximation Analysis of Financial-Related Text Messages

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26510%2F23%3APU147991" target="_blank" >RIV/00216305:26510/23:PU147991 - isvavai.cz</a>

Výsledek na webu

<a href="https://ecocyb.ase.ro/nr2023_1/2023_1_17_ZuzanaJANKOVA_online.pdf" target="_blank" >https://ecocyb.ase.ro/nr2023_1/2023_1_17_ZuzanaJANKOVA_online.pdf</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.24818/18423264/57.1.23.17" target="_blank" >10.24818/18423264/57.1.23.17</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Latent Dirichlet Allocation (LDA) Approximation Analysis of Financial-Related Text Messages

Popis výsledku v původním jazyce

Topic modeling is one of the most widely used NLP techniques for discovering hidden patterns and latent relationships in text documents. Latent Dirichlet Allocation (LDA) is one of the most popular methods in this field. There are different approaches to tuning the parameters of LDA models such as Gibbs sampling, variational method, or expected propagation. This paper aims to compare individual LDA parameter tuning approaches with respect to their performance and accuracy on textual data from the financial domain. From our results, it can be concluded that for text datasets obtained from financial social platforms, stochastic solvers are the most suitable and especially less time consuming than approximation methods.

Název v anglickém jazyce

Latent Dirichlet Allocation (LDA) Approximation Analysis of Financial-Related Text Messages

Popis výsledku anglicky

Topic modeling is one of the most widely used NLP techniques for discovering hidden patterns and latent relationships in text documents. Latent Dirichlet Allocation (LDA) is one of the most popular methods in this field. There are different approaches to tuning the parameters of LDA models such as Gibbs sampling, variational method, or expected propagation. This paper aims to compare individual LDA parameter tuning approaches with respect to their performance and accuracy on textual data from the financial domain. From our results, it can be concluded that for text datasets obtained from financial social platforms, stochastic solvers are the most suitable and especially less time consuming than approximation methods.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

50206 - Finance

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Economic Computation and Economic Cybernetics Studies and Research

ISSN

0424-267X

e-ISSN

1842-3264

Svazek periodika

57

Číslo periodika v rámci svazku

1

Stát vydavatele periodika

RO - Rumunsko

Počet stran výsledku

16

Strana od-do

267-282

Kód UT WoS článku

000960039800017

EID výsledku v databázi Scopus

2-s2.0-85151554545