Improving Targeted Multimodal Sentiment Classification with Semantic Description of Images

Targeted multimodal sentiment classification(TMSC)aims to identify the sentiment polarity of a target mentioned in a multimodal post.The majority of current studies on this task focus on mapping the image and the text to a high-dimensional space in order to obtain and fuse implicit representations,ignoring the rich semantic information contained in the images and not taking into account the contribution of the visual modality in the multimodal fusion representation,which can potentially influence the results of TMSC tasks.This paper proposes a general model for Improving Targeted Multimodal Sentiment Classification with Semantic Description of Images(ITMSC)as a way to tackle these issues and improve the accu-racy of multimodal sentiment analysis.Specifically,the ITMSC model can automatically adjust the contribution of images in the fusion representation through the exploitation of semantic descriptions of images and text similarity relations.Further,we propose a target-based attention module to capture the target-text relevance,an image-based attention module to capture the image-text relevance,and a target-image matching module based on the former two modules to properly align the target with the image so that fine-grained semantic information can be extracted.Our experimental results demonstrate that our model achieves comparable performance with several state-of-the-art approaches on two multimodal sentiment datasets.Our findings indicate that incorporating semantic descriptions of images can enhance our understanding of multimodal content and lead to improved sentiment analysis performance.

Using Informative Score for Instance Selection Strategy in Semi-Supervised Sentiment Classification

Sentiment classification is a useful tool to classify reviews about sentiments and attitudes towards a product or service.Existing studies heavily rely on sentiment classification methods that require fully annotated inputs.However,there is limited labelled text available,making the acquirement process of the fully annotated input costly and labour-intensive.Lately,semi-supervised methods emerge as they require only partially labelled input but perform comparably to supervised methods.Nevertheless,some works reported that the performance of the semi-supervised model degraded after adding unlabelled instances into training.Literature also shows that not all unlabelled instances are equally useful;thus identifying the informative unlabelled instances is beneficial in training a semi-supervised model.To achieve this,an informative score is proposed and incorporated into semisupervised sentiment classification.The evaluation is performed on a semisupervised method without an informative score and with an informative score.By using the informative score in the instance selection strategy to identify informative unlabelled instances,semi-supervised models perform better compared to models that do not incorporate informative scores into their training.Although the performance of semi-supervised models incorporated with an informative score is not able to surpass the supervised models,the results are still found promising as the differences in performance are subtle with a small difference of 2%to 5%,but the number of labelled instances used is greatly reduced from100%to 40%.The best finding of the proposed instance selection strategy is achieved when incorporating an informative score with a baseline confidence score at a 0.5:0.5 ratio using only 40%labelled data.

Convolutional Multi-Head Self-Attention on Memory for Aspect Sentiment Classification

This paper presents a method for aspect based sentiment classification tasks, named convolutional multi-head self-attention memory network(CMA-Mem Net). This is an improved model based on memory networks, and makes it possible to extract more rich and complex semantic information from sequences and aspects. In order to fix the memory network’s inability to capture context-related information on a word-level,we propose utilizing convolution to capture n-gram grammatical information. We use multi-head self-attention to make up for the problem where the memory network ignores the semantic information of the sequence itself. Meanwhile, unlike most recurrent neural network(RNN) long short term memory(LSTM), gated recurrent unit(GRU) models, we retain the parallelism of the network. We experiment on the open datasets Sem Eval-2014 Task 4 and Sem Eval-2016 Task 6. Compared with some popular baseline methods, our model performs excellently.

Sentiment Classification Based on Piecewise Pooling Convolutional Neural Network

Recently,the effectiveness of neural networks,especially convolutional neural networks,has been validated in the field of natural language processing,in which,sentiment classification for online reviews is an important and challenging task.Existing convolutional neural networks extract important features of sentences without local features or the feature sequence.Thus,these models do not perform well,especially for transition sentences.To this end,we propose a Piecewise Pooling Convolutional Neural Network(PPCNN)for sentiment classification.Firstly,with a sentence presented by word vectors,convolution operation is introduced to obtain the convolution feature map vectors.Secondly,these vectors are segmented according to the positions of transition words in sentences.Thirdly,the most significant feature of each local segment is extracted using max pooling mechanism,and then the different aspects of features can be extracted.Specifically,the relative sequence of these features is preserved.Finally,after processed by the dropout algorithm,the softmax classifier is trained for sentiment classification.Experimental results show that the proposed method PPCNN is effective and superior to other baseline methods,especially for datasets with transition sentences.

Chinese Sentiment Classification Using Extended Word2Vec

Sentiment analysis is now more and more important in modern natural language processing,and the sentiment classification is the one of the most popular applications.The crucial part of sentiment classification is feature extraction.In this paper,two methods for feature extraction,feature selection and feature embedding,are compared.Then Word2Vec is used as an embedding method.In this experiment,Chinese document is used as the corpus,and tree methods are used to get the features of a document:average word vectors,Doc2Vec and weighted average word vectors.After that,these samples are fed to three machine learning algorithms to do the classification,and support vector machine(SVM) has the best result.Finally,the parameters of random forest are analyzed.

Automatic Sentiment Classification of News Using Machine Learning Methods

With the rapid development of social economy,the society has entered into a new stage of development,especially in new media under the background of rapid development,makes the importance of news and information to get the comprehensive promotion,and in order to further identify the positive and negative news,should be fully using machine learning methods,based on the emotion to realize the automatic classifying of news,in order to improve the efficiency of news classification.Therefore,the article first makes clear the basic outline of news sentiment classification.Secondly,the specific way of automatic classification of news emotion is deeply analyzed.On the basis of this,the paper puts forward the concrete measures of automatic classification of news emotion by using machine learning.

Emphasizing Essential Words for Sentiment Classification Based onRecurrent Neural Networks

With the explosion of online communication and publication, texts become obtainable via forums, chat messages, blogs, book reviews and movie reviews. Usually, these texts are much short and noisy without sufficient statistical signals and enough information for a good semantic analysis. Traditional natural language processing methods such as Bow-of-Word (BOW) based probabilistic latent semantic models fail to achieve high performance due to the short text environment. Recent researches have focused on the correlations between words, i.e., term dependencies, which could be helpful for mining latent semantics hidden in short texts and help people to understand them. Long short-term memory (LSTM) network can capture term dependencies and is able to remember the information for long periods of time. LSTM has been widely used and has obtained promising results in variants of problems of understanding latent semantics of texts. At the same time, by analyzing the texts, we find that a number of keywords contribute greatly to the semantics of the texts. In this paper, we establish a keyword vocabulary and propose an LSTM-based model that is sensitive to the words in the vocabulary; hence, the keywords leverage the semantics of the full document. The proposed model is evaluated in a short-text sentiment analysis task on two datasets: IMDB and SemEval-2016, respectively. Experimental results demonstrate that our model outperforms the baseline LSTM by 1%similar to 2% in terms of accuracy and is effective with significant performance enhancement over several non-recurrent neural network latent semantic models (especially in dealing with short texts). We also incorporate the idea into a variant of LSTM named the gated recurrent unit (GRU) model and achieve good performance, which proves that our method is general enough to improve different deep learning models.

Multi-Domain Sentiment Classification with Classifier Combination

State-of-the-arts studies on sentiment classification are typically domain-dependent and domain-restricted. In this paper, we aim to reduce domain dependency and improve overall performance simultaneously by proposing an efficient multi-domain sentiment classification algorithm. Our method employs the approach of multiple classifier combination. In this approach, we first train single domain classifiers separately with domain specific data, and then combine the classifiers for the final decision. Our experiments show that this approach performs much better than both single domain classification approach （using the training data individually） and mixed domain classification approach （simply combining all the training data）. In particular, classifier combination with weighted sum rule obtains an average error reduction of 27.6% over single domain classification.

An Empirical Study of Unsupervised Sentiment Classification of Chinese Reviews

This paper is an empirical study of unsupervised sentiment classification of Chinese reviews. The focus is on exploring the ways to improve the performance of the unsupervised sentiment classification based on limited existing sentiment resources in Chinese. On the one hand, all available Chinese sentiment lexicons - individual and combined - are evaluated under our proposed framework. On the other hand, the domain dependent sentiment noise words are identified and removed using unlabeled data, to improve the classification performance. To the best of our knowledge, this is the first such attempt. Experiments have been conducted on three open datasets in two domains, and the results show that the proposed algorithm for sentiment noise words removal can improve the classification performance significantly.

Study and analysis of various sentiment classification strategies: A challenging overview

In large-scale social media,sentiment classification is a significant one for connecting gaps among social media contents as well as real-world actions,including public emotional status monitoring,political election prediction,and so on.On the other hand,textual sentiment classification is well studied by various platforms,like Instagram,Twitter,etc.Sentiment classification has many advantages in various fields,like opinion polls,educa-tion,and e-commerce.Sentiment classification is an interesting and progressing research area due to its applications in several areas.The information is collected from vari-ous people about social,products,and social events by web in sentiment analysis.This review provides a detailed survey of 50 research papers presenting sentiment classifica-tion schemes such as active learning-based approach,aspect learning-based method,and machine learning-based approach.The analysis is presented based on the categorization of sentiment classification schemes,the dataset used,software tools utilized,published year,and the performance metrics.Finally,the issues of existing methods considering conventional sentiment classification strategies are elaborated to obtain improved contri-bution in devising significant sentiment classification strategies.Moreover,the probable future research directions in attaining efficient sentiment classification are provided.

Multi-Task Learning Model with Data Augmentation for Arabic Aspect-Based Sentiment Analysis

Aspect-based sentiment analysis(ABSA)is a fine-grained process.Its fundamental subtasks are aspect termextraction(ATE)and aspect polarity classification(APC),and these subtasks are dependent and closely related.However,most existing works on Arabic ABSA content separately address them,assume that aspect terms are preidentified,or use a pipeline model.Pipeline solutions design different models for each task,and the output from the ATE model is used as the input to the APC model,which may result in error propagation among different steps because APC is affected by ATE error.These methods are impractical for real-world scenarios where the ATE task is the base task for APC,and its result impacts the accuracy of APC.Thus,in this study,we focused on a multi-task learning model for Arabic ATE and APC in which the model is jointly trained on two subtasks simultaneously in a singlemodel.This paper integrates themulti-task model,namely Local Cotext Foucse-Aspect Term Extraction and Polarity classification(LCF-ATEPC)and Arabic Bidirectional Encoder Representation from Transformers(AraBERT)as a shred layer for Arabic contextual text representation.The LCF-ATEPC model is based on a multi-head selfattention and local context focus mechanism(LCF)to capture the interactive information between an aspect and its context.Moreover,data augmentation techniques are proposed based on state-of-the-art augmentation techniques(word embedding substitution with constraints and contextual embedding(AraBERT))to increase the diversity of the training dataset.This paper examined the effect of data augmentation on the multi-task model for Arabic ABSA.Extensive experiments were conducted on the original and combined datasets(merging the original and augmented datasets).Experimental results demonstrate that the proposed Multi-task model outperformed existing APC techniques.Superior results were obtained by AraBERT and LCF-ATEPC with fusion layer(AR-LCF-ATEPC-Fusion)and the proposed data augmentation word embedding-based method(FastText)on the combined dataset.

Sentiment Drift Detection and Analysis in Real Time Twitter Data Streams

Handling sentiment drifts in real time twitter data streams are a challen-ging task while performing sentiment classifications,because of the changes that occur in the sentiments of twitter users,with respect to time.The growing volume of tweets with sentiment drifts has led to the need for devising an adaptive approach to detect and handle this drift in real time.This work proposes an adap-tive learning algorithm-based framework,Twitter Sentiment Drift Analysis-Bidir-ectional Encoder Representations from Transformers(TSDA-BERT),which introduces a sentiment drift measure to detect drifts and a domain impact score to adaptively retrain the classification model with domain relevant data in real time.The framework also works on static data by converting them to data streams using the Kafka tool.The experiments conducted on real time and simulated tweets of sports,health care andfinancial topics show that the proposed system is able to detect sentiment drifts and maintain the performance of the classification model,with accuracies of 91%,87%and 90%,respectively.Though the results have been provided only for a few topics,as a proof of concept,this framework can be applied to detect sentiment drifts and perform sentiment classification on real time data streams of any topic.

Multi-label Emotion Classification of COVID–19 Tweets with Deep Learning and Topic Modelling

The COVID-19 pandemic has become one of the severe diseases in recent years.As it majorly affects the common livelihood of people across the universe,it is essential for administrators and healthcare professionals to be aware of the views of the community so as to monitor the severity of the spread of the outbreak.The public opinions are been shared enormously in microblogging med-ia like twitter and is considered as one of the popular sources to collect public opinions in any topic like politics,sports,entertainment etc.,This work presents a combination of Intensity Based Emotion Classification Convolution Neural Net-work(IBEC-CNN)model and Non-negative Matrix Factorization(NMF)for detecting and analyzing the different topics discussed in the COVID-19 tweets as well the intensity of the emotional content of those tweets.The topics were identified using NMF and the emotions are classified using pretrained IBEC-CNN,based on predefined intensity scores.The research aimed at identifying the emotions in the Indian tweets related to COVID-19 and producing a list of topics discussed by the users during the COVID-19 pandemic.Using the Twitter Application Programming Interface(Twitter API),huge numbers of COVID-19 tweets are retrieved during January and July 2020.The extracted tweets are ana-lyzed for emotions fear,joy,sadness and trust with proposed Intensity Based Emotion Classification Convolution Neural Network(IBEC-CNN)model which is pretrained.The classified tweets are given an intensity score varies from 1 to 3,with 1 being low intensity for the emotion,2 being the moderate and 3 being the high intensity.To identify the topics in the tweets and the themes of those topics,Non-negative Matrix Factorization(NMF)has been employed.Analysis of emotions of COVID-19 tweets has identified,that the count of positive tweets is more than that of count of negative tweets during the period considered and the negative tweets related to COVID-19 is less than 5%.Also,more than 75%nega-tive tweets expressed sadness,fear are of low intensity.A qualitative analysis has also been conducted and the topics detected are grouped into themes such as eco-nomic impacts,case reports,treatments,entertainment and vaccination.The results of analysis show that the issues related to the pandemic are expressed dif-ferent emotions in twitter which helps in interpreting the public insights during the pandemic and these results are beneficial for planning the dissemination of factual health statistics to build the trust of the people.The performance comparison shows that the proposed IBEC-CNN model outperforms the conventional models and achieved 83.71%accuracy.The%of COVID-19 tweets that discussed the different topics vary from 7.45%to 26.43%on topics economy,Statistics on cases,Government/Politics,Entertainment,Lockdown,Treatments and Virtual Events.The least number of tweets discussed on politics/government on the other hand the tweets discussed most about treatments.

Modified Sine Cosine Optimization with Adaptive Deep Belief Network for Movie Review Classification

Sentiment analysis(SA)is a growing field at the intersection of computer science and computational linguistics that endeavors to automati-cally identify the sentiment presented in text.Computational linguistics aims to describe the fundamental methods utilized in the formation of computer methods for understanding natural language.Sentiment is classified as a negative or positive assessment articulated through language.SA can be commonly used for the movie review classification that involves the automatic determination that a review posted online(of a movie)can be negative or positive toward the thing that has been reviewed.Deep learning(DL)is becoming a powerful machine learning(ML)method for dealing with the increasing demand for precise SA.With this motivation,this study designs a computational intelligence enabled modified sine cosine optimization with a adaptive deep belief network for movie review classification(MSCADBN-MVC)technique.The major intention of the MSCADBN-MVC technique is focused on the identification of sentiments that exist in the movie review data.Primarily,the MSCADBN-MVC model follows data pre-processing and the word2vec word embedding process.For the classification of sentiments that exist in the movie reviews,the ADBN model is utilized in this work.At last,the hyperparameter tuning of the ADBN model is carried out using the MSCA technique,which integrates the Levy flight concepts into the standard sine cosine algorithm(SCA).In order to demonstrate the significant performance of the MSCADBN-MVC model,a wide-ranging experimental analysis is performed on three different datasets.The comprehensive study highlighted the enhancements of the MSCADBN-MVC model in the movie review classification process with maximum accuracy of 88.93%.

Leveraging Vision-Language Pre-Trained Model and Contrastive Learning for Enhanced Multimodal Sentiment Analysis

Multimodal sentiment analysis is an essential area of research in artificial intelligence that combines multiple modes,such as text and image,to accurately assess sentiment.However,conventional approaches that rely on unimodal pre-trained models for feature extraction from each modality often overlook the intrinsic connections of semantic information between modalities.This limitation is attributed to their training on unimodal data,and necessitates the use of complex fusion mechanisms for sentiment analysis.In this study,we present a novel approach that combines a vision-language pre-trained model with a proposed multimodal contrastive learning method.Our approach harnesses the power of transfer learning by utilizing a vision-language pre-trained model to extract both visual and textual representations in a unified framework.We employ a Transformer architecture to integrate these representations,thereby enabling the capture of rich semantic infor-mation in image-text pairs.To further enhance the representation learning of these pairs,we introduce our proposed multimodal contrastive learning method,which leads to improved performance in sentiment analysis tasks.Our approach is evaluated through extensive experiments on two publicly accessible datasets,where we demonstrate its effectiveness.We achieve a significant improvement in sentiment analysis accuracy,indicating the supe-riority of our approach over existing techniques.These results highlight the potential of multimodal sentiment analysis and underscore the importance of considering the intrinsic semantic connections between modalities for accurate sentiment assessment.

Seeker Optimization with Deep Learning Enabled Sentiment Analysis on Social Media;

World Wide Web enables its users to connect among themselves through social networks,forums,review sites,and blogs and these interactions produce huge volumes of data in various forms such as emotions,sentiments,views,etc.Sentiment Analysis(SA)is a text organization approach that is applied to categorize the sentiments under distinct classes such as positive,negative,and neutral.However,Sentiment Analysis is challenging to perform due to inadequate volume of labeled data in the domain of Natural Language Processing(NLP).Social networks produce interconnected and huge data which brings complexity in terms of expanding SA to an extensive array of applications.So,there is a need exists to develop a proper technique for both identification and classification of sentiments in social media.To get rid of these problems,Deep Learning methods and sentiment analysis are consolidated since the former is highly efficient owing to its automatic learning capability.The current study introduces a Seeker Optimization Algorithm with Deep Learning enabled SA and Classification(SOADL-SAC)for social media.The presented SOADL-SAC model involves the proper identification and classification of sentiments in social media.In order to attain this,SOADL-SAC model carries out data preprocessing to clean the input data.In addition,Glove technique is applied to generate the feature vectors.Moreover,Self-Head Multi-Attention based Gated Recurrent Unit(SHMA-GRU)model is exploited to recognize and classify the sentiments.Finally,Seeker Optimization Algorithm(SOA)is applied to fine-tune the hyperparameters involved in SHMA-GRU model which in turn enhances the classifier results.In order to validate the enhanced outcomes of the proposed SOADL-SAC model,various experiments were conducted on benchmark datasets.The experimental results inferred the better performance of SOADLSAC model over recent state-of-the-art approaches.

Improved Hybrid Deep Collaborative Filtering Approach for True Recommendations

Recommendation services become an essential and hot research topic for researchers nowadays.Social data such asReviews play an important role in the recommendation of the products.Improvement was achieved by deep learning approaches for capturing user and product information from a short text.However,such previously used approaches do not fairly and efficiently incorporate users’preferences and product characteristics.The proposed novel Hybrid Deep Collaborative Filtering(HDCF)model combines deep learning capabilities and deep interaction modeling with high performance for True Recommendations.To overcome the cold start problem,the new overall rating is generated by aggregating the Deep Multivariate Rating DMR(Votes,Likes,Stars,and Sentiment scores of reviews)from different external data sources because different sites have different rating scores about the same product that make confusion for the user to make a decision,either product is truly popular or not.The proposed novel HDCF model consists of four major modules such as User Product Attention,Deep Collaborative Filtering,Neural Sentiment Classifier,and Deep Multivariate Rating(UPA-DCF+NSC+DMR)to solve the addressed problems.Experimental results demonstrate that our novel model is outperforming state-of-the-art IMDb,Yelp2013,and Yelp2014 datasets for the true top-n recommendation of products using HDCF to increase the accuracy,confidence,and trust of recommendation services.

Weighted Co-Training for Cross-Domain Image SentimentClassification

Image sentiment classification, which aims to predict the polarities of sentiments conveyed by the images, has gained a lot of attention. Most existing methods address this problem by training a general classifier with certain visual features, ignoring the discrepancies across domains. In this paper, we propose a novel weighted co-training method for cross-domain image sentiment classification, which iteratively enlarges the labeled set by introducing new high-confidence classified samples to reduce the gap between the two domains. We train two sentiment classifiers with both the images and the corresponding textual comments separately, and set the similarity between the source domain and the target domain as the weight of a classifier. We perform extensive experiments on a real Flickr dataset to evaluate the proposed method, and the empirical study reveals that the weighted co-training method significantly outperforms some baseline solutions.

OKO-SVM:Online kernel optimization-based support vector machine for the incremental learning and classification of the sentiments in the train reviews

Online incremental learning is one of the emerging research interests among the researchers in the recent years.The sentiment classification through the online incremental learning faces many challenges due to the limitations in the memory and the computing resources available for processing the online reviews.This work has introduced an online incremental learning algorithm for classifying the train reviews.The sentiments available in the reviews provided for the public services are necessary for improving the quality of the service.This work proposes the online kernel optimizationbased support vector machine(OKO-SVM)classifier for the sentiment classification of the train reviews.This paper is the extension of the previous work kernel optimizationbased support vector machine(KO-SVM).The OKO-SVM classifier uses the proposed fuzzy bound for modifying the weight for each incoming review database for the particular time duration.The simulation uses the standard train review and the movie review database for the classification.From the simulation results,it is evident that the proposed model has achieved a better performance with the values of 84.42%,93.86%,and 74.56%regarding the accuracy,sensitivity,and specificity while classifying the train review database.

Transfer Learning via Multi-View Principal Component Analysis

Transfer learning aims at leveraging the knowledge in labeled source domains to predict the unlabeled data in a target domain, where the distributions are diiTerent in domains. Among various methods for transfer learning, one kind of Mgorithms focus on the correspondence between bridge features and all the other specific features from different domains, and later conduct transfer learning via the single-view correspondence. However, the single-view correspondence may prevent these algorithms from further improvement due to the problem of incorrect correlation discovery. To tackle this problem, we propose a new method for transfer learning in a multi-view correspondence perspective, which is called MultiView Principal Component Analysis （MVPCA） approach. MVPCA discovers the correspondence between bridge features representative across all domains and specific features from different domains respectively, and conducts the transfer learning by dimensionality reduction in a multi-view way, which can better depict the knowledge transfer. Experiments show that MVPCA can significantly reduce the cross domain prediction error of a baseline non-transfer method. With multi-view correspondence information incorporated to the single-view transfer learning method, MVPCA can further improve the performance of one state-of-the-art single-view method.