Sentiments analysis of fMRI using automatically generated stimuli labels under naturalistic paradigm DOI Creative Commons

Rimsha Mahrukh,

Sadia Shakil, Aamir Saeed Malik

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: May 4, 2023

Abstract Our emotions and sentiments are influenced by naturalistic stimuli such as the movies we watch songs listen to, accompanied changes in our brain activation. Comprehension of these brain-activation dynamics can assist identification any associated neurological condition stress depression, leading towards making informed decision about suitable stimuli. A large number open-access functional magnetic resonance imaging (fMRI) datasets collected under conditions be used for classification/prediction studies. However, do not provide emotion/sentiment labels, which limits their use supervised learning Manual labeling subjects generate however, this method is subjective biased. In study, proposing another approach generating automatic labels from stimulus itself. We using sentiment analyzers (VADER, TextBlob, Flair) natural language processing to movie subtitles. Subtitles generated class positive, negative, neutral classification fMRI images. Support vector machine, random forest, tree, deep neural network classifiers used. getting reasonably good accuracy (42–84%) imbalanced data, increased (55–99%) balanced data.

Language: Английский

Sentiments analysis of fMRI using automatically generated stimuli labels under naturalistic paradigm DOI Creative Commons

Rimsha Mahrukh,

Sadia Shakil, Aamir Saeed Malik

et al.

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: May 4, 2023

Abstract Our emotions and sentiments are influenced by naturalistic stimuli such as the movies we watch songs listen to, accompanied changes in our brain activation. Comprehension of these brain-activation dynamics can assist identification any associated neurological condition stress depression, leading towards making informed decision about suitable stimuli. A large number open-access functional magnetic resonance imaging (fMRI) datasets collected under conditions be used for classification/prediction studies. However, do not provide emotion/sentiment labels, which limits their use supervised learning Manual labeling subjects generate however, this method is subjective biased. In study, proposing another approach generating automatic labels from stimulus itself. We using sentiment analyzers (VADER, TextBlob, Flair) natural language processing to movie subtitles. Subtitles generated class positive, negative, neutral classification fMRI images. Support vector machine, random forest, tree, deep neural network classifiers used. getting reasonably good accuracy (42–84%) imbalanced data, increased (55–99%) balanced data.

Language: Английский

Citations

10