Ensemble of Generative and Discriminative Techniques for Sentiment Analysis of Movie Reviews

Auteur-es

Gregoire Mesnil, Tomas Mikolov, Marc'Aurelio Ranzato, Yoshua Bengio

Nombre Auteurs

4

Titre

Ensemble of Generative and Discriminative Techniques for Sentiment Analysis of Movie Reviews

Année de publication

2015

Référence (APA)

Mesnil, G., Mikolov, T., Ranzato, M., & Bengio, Y. (2015). Ensemble of Generative and Discriminative Techniques for Sentiment Analysis of Movie Reviews (arXiv:1412.5335). arXiv. http://arxiv.org/abs/1412.5335

résumé

Sentiment analysis is a common task in natural language processing that aims to detect polarity of a text document (typically a consumer review). In the simplest settings, we discriminate only between positive and negative sentiment, turning the task into a standard binary classification problem. We compare several machine learning approaches to this problem, and combine them to achieve a new state of the art. We show how to use for this task the standard generative language models, which are slightly complementary to the state of the art techniques. We achieve strong results on a well-known dataset of IMDB movie reviews. Our results are easily reproducible, as we publish also the code needed to repeat the experiments. This should simplify further advance of the state of the art, as other researchers can combine their techniques with ours with little effort.

URL

https://research.facebook.com/file/167078822167214/ensemble-of-generative-and-discriminative-techniques-for-sentiment-analysis-of-movie-reviews.pdf

doi

https://doi.org/10.48550/arXiv.1412.5335

Accessibilité de l'article

Libre

Champ

Artificial Intelligence

Type contenu (théorique Applicative méthodologique)

Applicatif

Méthode

The method involves comparing several machine learning approaches to sentiment analysis, including generative language models, and combining them to achieve a new state of the art.

Cas d'usage

IMDB

Objectifs de l'article

The objectives of the article are to improve the accuracy of sentiment analysis by combining several machine learning approaches. It aims to provide a reproducible method for achieving state-of-the-art result

Question(s) de recherche/Hypothèses/conclusion

The research question is how to improve the accuracy of sentiment analysis using machine learning approaches.

The hypothesis is that combining different machine learning approaches, including generative language models, will lead to improved accuracy in sentiment analysis.

The conclusions are that the proposed ensemble method achieves state-of-the-art results on a well-known dataset of movie reviews and that the code provided by the authors allows for easy reproducibility of their results.

Cadre théorique/Auteur.es

The theoretical framework of the article includes machine learning and natural language processing, and the main authors cited include Pang and Lee, Pascanu, Mikolov, Bengio, Socher, Pennington, Huang, Ng, Manning, and Stolcke.

Concepts clés

Sentiment analysis, Machine learning, Generative models, Discriminative models, Ensemble methods, Natural language processing

Données collectées (type source)

The article uses the Stanford IMDB dataset of movie reviews.

Définition des émotions

Non

Ampleur expérimentation (volume de comptes)

25 000 IMDB reviews for training

25 000 reviews for testing

Technologies associées

Machine learning algorithms, Natural language processing techniques, Generative language models.

Mention de l'éthique

Non

Finalité communicationnelle

"combining both generative and discriminative models together for sentiment prediction [...] one based on a generative approach (language models), one based on continuous representations of sentences and one based on a clever reweighing of tf-idf bag-of-word representation of the document." give better results.