SentiStory: multi-grained sentiment analysis and event summarization with crowdsourced social media data

Purpose This research employs innovative deep learning techniques to perform sentiment analysis on airline customer feedback, systematically investigating service quality dimensions and customer satisfaction determinants through the theoretical lens of the SERVQUAL framework. Design/methodology/approach We analyzed airline reviews from TripAdvisor and Skytrax from July 2014 to July 2023 using SERVQUAL theory. Coarse-grained clustering revealed themes in tangibles and reliability, while fine-grained sentiment analysis, using our FusionBERT model, assessed responsiveness, assurance and empathy for deeper service quality insights. Findings Our approach outperformed baselines, offering a nuanced analysis of airline customer satisfaction using the SERVQUAL theory. Coarse-grained sentiment analysis assesses tangibles and reliability (facilities, service consistency), while fine-grained analysis evaluates responsiveness, assurance, and empathy (staff promptness, professionalism and care). This dual-layered method enables comprehensive service quality evaluation, helping airlines identify strengths and improvement areas. Originality/value This research provides actionable management solutions and recommendations for airline managers.

Intent analysis and classification are performed to identify the expressions of intent in the given text. In this paper, the dataset is classified into emotion classifications by utilizing machine learning model SVM, Bipolar classification, Fine Grained Analysis, and Sarcasm detection, with Naïve Bayes and Random Forest techniques of deep learning, including Long Short-Term Memory to perform intention analysis on social media data. Then Fine-grained or Multi-Class Sentiment analysis is used for further classification of the five classes, viz. negative, strong negative, neutral, positive, and strong positive, which detects the sarcastic reviews in the movie dataset. The emotional intention behind the review comments is classified as happiness, rage, sadness, joy, anger, and disgust by using SVM. The reviews are analyzed and calculated based on their subjectivity and context level similarity using Related Relaxed Word Mover Distance (RRWMD) semantic similarity measure. With the advantage of the RRWMD algorithm, the reviews from the context containing deviated or irrelevant contents were removed before being applied to the classification algorithms, thereby reducing the execution time, which obtains a 3% improvement in accuracy. The disadvantage of the RRWMD algorithm is only one deep learning algorithm is compared. From the observed accuracy scores and classification reports, the LSTM has provided higher accuracy, despite the long execution time. The Naïve Bayes model has produced lower accuracy than the neural network model but was efficient, taking less time to fit and classify. The results from various experiments have proven that the semantic similarity measure provides more accurate results than the state-of-the-art model.

The assessment of faculty teaching effectiveness plays a pivotal role in shaping educational practices and ensuring academic excellence. While numerical evaluations provide quantifiable measures of teaching effectiveness, textual comments offer valuable qualitative insights into the faculty's instructional methods and interactions. However, discrepancies between these evaluation formats may hinder the accurate assessment of faculty performance. This study aims to determine the alignment between the numerical and textual evaluations of faculty teaching effectiveness using fine-grained sentiment analysis and determine the relationship between numerical ratings and textual comments to identify patterns of consistency or divergence. A large dataset of anonymized BSCS students’ feedback is analyzed through sentiment analysis to extract sentiments expressed in the textual comments. The study results reveal a misalignment between students’ subjective perceptions expressed in the textual comments and the numerical ratings. This misalignment arises due to several factors, such as student communication style variations, individual interpretations of rating scales, or subjective biases. This study advances faculty evaluation methodologies and offers useful recommendations for organizations looking to improve assessment procedures as well as educational institutions that want to improve the precision and efficacy of their faculty evaluation systems for better faculty development and decision-making.

The rapid growth of internet facilities has increased the comments, posts, blogs, feedback, etc., on a large scale on social networking sites. These social media data are available in an unstructured form, which includes images, text, and videos. The processing of these data is difficult, but some sentiment analysis, information retrieval, and recommender systems are used to process these unstructured data. To extract the opinion and sentiment of internet users from their written social media text, a sentiment analysis system is required to develop, which can work on both monolingual and bilingual phonetic text. Therefore, a sentiment analysis (SA) system is developed, which performs well on different domain datasets. The system performance is tested on four different datasets and achieved better accuracy of 3% on social media datasets, 1.5% on movie reviews, 1.35% on Amazon product reviews, and 4.56% on large Amazon product reviews than the state-of-art techniques. Also, the stemmer (StemVerb) for verbs of the English language is proposed, which improves the SA system's performance.

Aspect-based sentiment analysis as fine-grained sentiment analysis problem, mainly aims to clarify the sentiment of different aspect. Previous fine-grained sentiment analysis methods usually used Recurrent Neural Networks as feature extractors, and achieved good results in sentiment analysis. However, they were limited to the sequential processing of sequence by RNN. Its training and operation efficiency are relatively low and time consuming. Aiming at this problem, combining the multi-head attention mechanism in Transformer, fusion model of convolution neural network and hierarchical attention coding network is proposed, and convolution neural network is used as feature encoder to improve the efficiency of model training. At the same time, multi-head attention is used to calculate the degree of relevance between aspect words and context, thus capturing emotional elements corresponding to entity words. Experiments on the SemEval2014 dataset have improved the training and inference efficiency of aspect-based sentiment analysis.

Under the goal of carbon peak and carbon neutrality, developing battery electric vehicles (BEVs) is an important way to reduce carbon emissions in the transportation sector. To popularize BEVs as soon as possible, it is necessary to study selection strategies for BEVs from the perspective of consumers. Therefore, the Latent Dirichlet Allocation (LDA) model based on fine-grained sentiment analysis is combined with the multi-criteria decision-making (MCDM) model to assess ten types of BEV alternatives. Fine-grained sentiment analysis is applied to find the vehicle attributes that consumers care about the most based on the word-of-mouth data. The LDA model is suggested to divide topics and construct the indicator system. The MCDM model is used to rank vehicles and put forward the corresponding optimization path to increase consumer purchases of BEVs in China. The results show that (a) via the LDA model based on fine-grained sentiment analysis, attributes that consumers care most about are divided into five topics: dynamics, technology, safety, comfort, and cost; (b) based on the DEMATEL technique, the dimensions in the order of importance are as follows: safety, technology, dynamics, comfort, and cost; (c) the price is the most important criteria that affect customers’ satisfaction by the DANP model; and (d) based on the VIKOR model, the selection strategies present that Aion S is highlighted as the best choice, and the optimization path is discussed to promote the performance of BEVs to increase customers’ satisfaction. The findings can provide a reference for improving the sustainable development of the automobile industry in China. The proposed framework serves as the basis for further discussion of BEVs.

The emergence of various e-commerce sites has led to an increase in review sites for various services and products. People nowadays easily get information about products and services that will be used through reviews. Here sentiment analysis plays an important role in classifying the polarity of product reviews. However, with a large number of reviews, a sentiment analysis that only gives overall polarity is not sufficient. This will make it difficult to find the reviews of certain aspects (features) of the product. Aspect-based sentiment analysis as fine-grained sentiment analysis is able to provide specific polarity for each aspect contained in a sentence. Various kinds of development methods have been carried out to provide accurate results in aspect-based sentiment analysis. This paper will discuss the various deep learning methods that have been carried out and provide the possibility of research that can be carried out from Aspect-Based Sentiment Analysis.

In the present age of rapid digital and technological progress, individuals spend a significant portion of their day connected to the Internet, actively expressing their thoughts and emotions on various online platforms. The analysis of these comments is important for detecting fundamental sentiments that lead to informed decisions across numerous sectors. Sentiment analysis (SA) classifies textual or spoken expressions as positive, negative, or neutral. Although several machine learning and NLP models are available for SA, accurately extracting and categorizing opinions remains challenging. This study compared seven state-of-the-art models-LSTM, BiLSTM, CNN, random forest, logistic regression, support vector machine (SVM), and naïve Bayeswith the BERT model to achieve precise sentiment classification. TF-IDF vectorization was performed before the SA. Using 11,323 annotated sentences from three domains, the models were assessed based on the F1- score, recall, precision, and accuracy. Each sentence was labeled to reflect either a positive or negative sentiment. The experimental results show that the BERT model performs best, achieving an accuracy of 99 % for Games data, 94 % for the comprehensive data, 90 % for the social networking data, and 97% for the productivity data. For all the data types, BERT outperformed all the state-of-the-art models considered for comparison. The findings offer practical insights for the mobile app industry by enhancing the understanding of user opinions through SA. This study contributes to real-world applications by guiding the effective selection of mobile apps based on fine-grained sentiment analysis.

This paper describes the approach we used for SemEval-2017 Task 5: Fine-Grained Sentiment Analysis on Financial Microblogs. We use three types of word embeddings in our algorithm: word embeddings learned from 200 million tweets, sentiment-specific word embeddings learned from 10 million tweets using distance supervision, and word embeddings learned from 20 million StockTwits messages. In our approach, we also take the left and right context of the target company into consideration when generating polarity prediction features. All the features generated from different word embeddings and contexts are integrated together to train our algorithm

The 24-hour news cycle and barrage of online media is a constant drum beat. The flow of positive and negative news is always in flux, influencing our current perspective and reassessing our future outlook. Nowhere is this more true than in the capital markets where assets are priced and risk assessed based on future expectations. While many factors influence a trader's decision to buy or sell an asset it can be argued that the sentiment from the 24-hour news cycle greatly impacts their outlook on the future value of an asset. In this paper we propose new methods to predict the positive or negative sentiment of financial news. Our analysis has found that contemporary document level sentiment analysis methods break down at fine-grained levels. Fine-grained analysis methods are vitally important as the velocity and impact of small texts, such as tweets and news flashes, increase their influence over the decision process. Using Natural Language Processing methods we extract syntactic sentence patterns from financial news headlines. From these patterns we conduct experiments using both lexicon and machine learning sentiment analysis approaches to predict sentiment. We find that our sentiment prediction methods are able to consistently out perform lexicon methods. Our robust techniques give the financial practitioner a method to fold a fine-grained news sentiment factor into their pricing or risk prediction models.

Understanding human's complicated and capricious emotions remains a fundamental challenge. In this paper, we propose a fine-grained sentiment analysis system which classify emotions into 32 categories. For one direction, we cover more detailed emotions and for the other direction, we further measure each emotion with strength, such as describing angry by annoyance, anger and range. Taking Japanese as a test language, we describe our methods of building the training data, of constructing deep neural network classifiers, and of evaluating the models.

Aspect-based sentiment analysis is an important task in fine-grained sentiment analysis, which aims to infer the sentiment towards a given aspect. Previous studies have shown notable success when sufficient labeled training data is available. However, annotating adequate data is labor-intensive, which sets substantial barriers for generalizing the sentiment predictor to the new domain. Two main challenges exist in cross-domain aspect-based sentiment analysis. One challenge is acquiring the domain-invariant knowledge; the other challenge is mining the syntactic-related words towards the aspect-term. In this article, we propose a transformer-based semantic-primary knowledge transferring network (TSPKT) for cross-domain aspect-term sentiment analysis, which utilizes semantic-primary knowledge as a bridge to enable knowledge transfer across domains. Specifically, we first build an S-Graph from external semantic lexicons, and extract the semantic-primary knowledge from the S-Graph. Second, AoaGraphormer is proposed to learn the syntactically relevant words towards the aspect-term. Third, we extend the standard biLSTM classifier to fully integrate the semantic-primary knowledge by adding a novel knowledge-aware memory unit (KAMU) to the biLSTM cell. Extensive experiments on six cross-domain setups demonstrate the superiority of TSPKT against the state-of-the-art baseline methods.

Aspect-based sentiment analysis is a fine-grained sentiment analysis that focuses on the sentiment polarity of different aspects of text, and most current research methods use a combination of dependent syntactic analysis and graphical neural networks. In this paper, a graph attention network aspect-based sentiment analysis model based on the weighting of dependencies (WGAT) is designed to address the problem in that traditional models do not sufficiently analyse the types of syntactic dependencies; in the proposed model, graph attention networks can be weighted and averaged according to the importance of different nodes when aggregating information. The model first transforms the input text into a low-dimensional word vector through pretraining, while generating a dependency syntax graph by analysing the dependency syntax of the input text and constructing a dependency weighted adjacency matrix according to the importance of different dependencies in the graph. The word vector and the dependency weighted adjacency matrix are then fed into a graph attention network for feature extraction, and sentiment polarity is predicted through the classification layer. The model can focus on syntactic dependencies that are more important for sentiment classification during training, and the results of the comparison experiments on the Semeval-2014 laptop and restaurant datasets and the ACL-14 Twitter social comment dataset show that the WGAT model has significantly improved accuracy and F1 values compared to other baseline models, validating its effectiveness in aspect-level sentiment analysis tasks.

Background Fine-grained sentiment analysis is used to interpret consumers’ sentiments, from their written comments, towards specific entities on specific aspects. Previous researchers have introduced three main tasks in this field (ABSA, TABSA, MEABSA), covering all kinds of social media data (e.g., review specific, questions and answers, and community-based). In this paper, we identify and address two common challenges encountered in these three tasks, including the low-resource problem and the sentiment polarity bias. Methods We propose a unified model called PEA by integrating data augmentation methodology with the pre-trained language model, which is suitable for all the ABSA, TABSA and MEABSA tasks. Two data augmentation methods, which are entity replacement and dual noise injection, are introduced to solve both challenges at the same time. An ensemble method is also introduced to incorporate the results of the basic RNN-based and BERT-based models. Results PEA shows significant improvements on all three fine-grained sentiment analysis tasks when compared with state-of-the-art models. It also achieves comparable results with what the baseline models obtain while using only 20% of their training data, which demonstrates its extraordinary performance under extreme low-resource conditions.

Background Fine-grained sentiment analysis is used to interpret consumers’ sentiments, from their written comments, towards specific entities on specific aspects. Previous researchers have introduced three main tasks in this field (ABSA, TABSA, MEABSA), covering all kinds of social media data (e.g., review specific, questions and answers, and community-based). In this paper, we identify and address two common challenges encountered in these three tasks, including the low-resource problem and the sentiment polarity bias. Methods We propose a unified model called PEA by integrating data augmentation methodology with the pre-trained language model, which is suitable for all the ABSA, TABSA and MEABSA tasks. Two data augmentation methods, which are entity replacement and dual noise injection, are introduced to solve both challenges at the same time. An ensemble method is also introduced to incorporate the results of the basic RNN-based and BERT-based models. Results PEA shows significant improvements on all three fine-grained sentiment analysis tasks when compared with state-of-the-art models. It also achieves comparable results with what the baseline models obtain while using only 20% of their training data, which demonstrates its extraordinary performance under extreme low-resource conditions.