Análisis Semántico Latente para la Detección de Noticias Falsas sobre Covid-19 Utilizando Computación Heterogénea

Translated title of the contribution: Latent Semantic Analysis for the Detection of Fake News about Covid-19 Using Heterogeneous Computing

Bryam David Vega Moreno, Gabriel Alejandro Leon Paredes, David Andres Morales Rivera

Research output: Contribution to journalArticle

Abstract

The detection of fake news nowadays is a great challenge for prediction systems due to the large amount of information that is currently available, especially in information sources such as social networks, blogs or websites. In addition, the processing capacity required to analyze large amounts of data is very large so the execution time tends to be high. In this paper, a learning system using parallel processing paradigms at CPU and GPU level using the COVID-19 Open Research Dataset Challenge (CORD-19) dataset is proposed for a first approach to fake news detection on COVID-19. The prediction system is based on natural language processing techniques using latent semantic analysis (LSA) as a training model. Also, CPU-level multiprocessing techniques are used for text preprocessing, keyword retrieval, term-by-document matrix retrieval, value normalization using TF-IDF and cosine similarity retrieval, while for the dimensionality reduction part using singular value decomposition or SVD, the CUDA architecture has been used for GPU-level processing.
Translated title of the contributionLatent Semantic Analysis for the Detection of Fake News about Covid-19 Using Heterogeneous Computing
Original languageSpanish (Ecuador)
Pages (from-to)18-29
Number of pages12
JournalConvergence Tech
Volume5
Issue number5
DOIs
StatePublished - 26 Oct 2021

Keywords

  • Covid-19
  • Heterogeneous computing
  • Latent semantic analysis
  • Natural language processing

CACES Knowledge Areas

  • 116A Computer Science

Fingerprint

Dive into the research topics of 'Latent Semantic Analysis for the Detection of Fake News about Covid-19 Using Heterogeneous Computing'. Together they form a unique fingerprint.

Cite this