Tailored text augmentation for sentiment analysis

Feng, Zijian; Zhou, Hanzhang; Zhu, Zixiao; Mao, Kezhi

doi:10.1016/j.eswa.2022.117605

Cited by 16 publications

(3 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The adversarial word dilution approach was introduced by [24], which learns the dilution weights through a constrained min-max optimization process with the guidance of the labels. Moreover, [25] proposed tailored text augmentation by combining probabilistic synonym replacement and irrelevant zero masking. This method improves the generalization capability of the model to enhance its accuracy.…”

Section: Unsupervised Methodsmentioning

confidence: 99%

TAWC: Text Augmentation with Word Contributions for Imbalance Aspect-based Sentiment Classification

Santoso,

Mendonça,

Aritsugi

2023

Preprint

View full text Add to dashboard Cite

Text augmentation plays an important role in enhancing the generalization performance of language models. However, traditional methods often overlook the unique roles that individual words play in conveying meaning in text and imbalance class distribution, thereby risking suboptimal performance and compromising the model's generalization ability. This limitation motivated us to create a novel technique, Text Augmentation with Word Contributions (TAWC). Our approach tackles this problem in two core steps: Firstly, it employs analytical correlation and semantic similarity metrics to discern the relationships between words and their associated aspect polarities; and secondly, it tailors distinct augmentation strategies to individual words, based on their identified functional contributions within the text. Extensive experiments on two aspect-based sentiment analysis datasets reveal that TAWC significantly improves the classification performance of popular language models, achieving gains of up to 4%, thereby setting a new standard in the field of text augmentation.

show abstract

Section: Unsupervised Methodsmentioning

confidence: 99%

TAWC: Text Augmentation with Word Contributions for Imbalance Aspect-based Sentiment Classification

Santoso,

Mendonça,

Aritsugi

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The advancement of NLP led to a paradigm shift in text augmentation, with the introduction of various techniques such as synonym replacement, random operations, and back-translation ( Wei & Zou, 2019 ; Abonizio, Paraiso & Barbon, 2022 ; Feng et al, 2022 ; Karimi, Rossi & Prati, 2021 ). While these methods have shown effectiveness, their application is often constrained in low-resource languages, where tools akin to WordNet Miller (1995) for tasks like synonym or hyponym replacement are scarce or non-existent.…”

Section: Related Workmentioning

confidence: 99%

Text data augmentation and pre-trained Language Model for enhancing text classification of low-resource languages

Ziyaden,

Yelenov,

Hajiyev

et al. 2024

PeerJ Computer Science

View full text Add to dashboard Cite

Background In the domain of natural language processing (NLP), the development and success of advanced language models are predominantly anchored in the richness of available linguistic resources. Languages such as Azerbaijani, which is classified as a low-resource, often face challenges arising from limited labeled datasets, consequently hindering effective model training. Methodology The primary objective of this study was to enhance the effectiveness and generalization capabilities of news text classification models using text augmentation techniques. In this study, we solve the problem of working with low-resource languages using translations using the Facebook mBart50 model, as well as the Google Translate API and a combination of mBart50 and Google Translate thus expanding the capabilities when working with text. Results The experimental outcomes reveal a promising uptick in classification performance when models are trained on the augmented dataset compared with their counterparts using the original data. This investigation underscores the immense potential of combined data augmentation strategies to bolster the NLP capabilities of underrepresented languages. As a result of our research, we have published our labeled text classification dataset and pre-trained RoBERTa model for the Azerbaijani language.

show abstract

“…[12] Given such complexity, Natural Language Processing (NLP) approaches are quite useful for developing various solutions [13]. One of the most prominent NLP research fields nowadays is sentiment analysis [14], which is used in a variety of fields, including recommender systems, data-driven systems, healthcare research, and others. Due to the popularity of social media, massive volumes of data are created on the Internet every day.…”

Section: Introductionmentioning

confidence: 99%

The Impact of COVID-19 on E-Learning: Context-Based Sentiment Analysis Discourse Using Text Mining

Rehman,

Aslam,

Abid

et al. 2023

VAWKUM trans. comput. sci.

View full text Add to dashboard Cite

Finding the most dominant and pertinent user opinions on a certain topic is crucial to the sentiment analysis success factor. During the pandemic lockdowns around the world, the suspension of academic institutions leads to an exceptional increase in distance education. Academic institutions closed their campuses immediately to mitigate the effects of COVID-19 and prevent its pervasive spread, and educational activities were shifted to online platforms. The effectiveness of online education is a significant topic of interest for both students and their parents, especially in terms of how students and teachers perceive it and how technologically viable it is in a range of social circumstances. Before such a wide adoption of e-learning is possible, these issues must be analyzed from multiple perspectives. The present research aims to evaluate the efficacy of e-learning by examining individuals' perceptions of it. Opinions can be found on websites such as Instagram, Facebook, Twitter, etc. As social media has recently emerged as a significant means of communication. This study addresses factors connected to a significant change in the educational system. 200,000 tweets were gathered from Twitter to evaluate the opinions of Twitter users who were taking part in online learning. This study adopts VADER to analyze the subjectivity and polarity score of tweets, a topic model was also created using the LDA algorithm to determine the themes that were talked about on Twitter the most. The models have been constructed and evaluated using Word2Vec to capture the semantic relationships between words and LSTM and RNN sequential model for sentiment analysis. This study measured the efficiency of a sentiment analysis model using the accuracy metric, the conducted experiments reveal that the proposed hybrid model achieves an overall accuracy of 96.3%. The results also indicate a significant negative impact of the Covid-19 pandemic on individuals' emotions, with 64.4% of the analyzed tweets displaying negative sentiments. These findings provide valuable insights into the relationship between global events and individual emotions on social media platforms.

show abstract

Tailored text augmentation for sentiment analysis

Cited by 16 publications

References 32 publications

TAWC: Text Augmentation with Word Contributions for Imbalance Aspect-based Sentiment Classification

TAWC: Text Augmentation with Word Contributions for Imbalance Aspect-based Sentiment Classification

Text data augmentation and pre-trained Language Model for enhancing text classification of low-resource languages

The Impact of COVID-19 on E-Learning: Context-Based Sentiment Analysis Discourse Using Text Mining

Contact Info

Product

Resources

About