Sentiment Corpus Research Articles

Feature Extraction Network with Attention Mechanism for Data Enhancement and Recombination Fusion for Multimodal Sentiment Analysis

Multimodal sentiment analysis and emotion recognition represent a major research direction in natural language processing (NLP). With the rapid development of online media, people often express their emotions on a topic in the form of video, and the signals it transmits are multimodal, including language, visual, and audio. Therefore, the traditional unimodal sentiment analysis method is no longer applicable, which requires the establishment of a fusion model of multimodal information to obtain sentiment understanding. In previous studies, scholars used the feature vector cascade method when fusing multimodal data at each time step in the middle layer. This method puts each modal information in the same position and does not distinguish between strong modal information and weak modal information among multiple modalities. At the same time, this method does not pay attention to the embedding characteristics of multimodal signals across the time dimension. In response to the above problems, this paper proposes a new method and model for processing multimodal signals, which takes into account the delay and hysteresis characteristics of multimodal signals across the time dimension. The purpose is to obtain a multimodal fusion feature emotion analysis representation. We evaluate our method on the multimodal sentiment analysis benchmark dataset CMU Multimodal Opinion Sentiment and Emotion Intensity Corpus (CMU-MOSEI). We compare our proposed method with the state-of-the-art model and show excellent results.

Modeling Public Sentiments About JUUL Flavors on Twitter Through Machine Learning.

The availability of a variety of e-cigarettes flavors is one of the frequently cited reasons for their adoption. An active stream of discussion about flavoring can be observed online. Analyzing these real-time conversations offers nuanced insights into key factors related to the adoption of flavors, subsequently supporting public health interventions. Google's BERT, a state-of-the-art deep learning method was employed to model the first sentiment corpus on JUUL flavors. BERT, which is pre-trained with the complete English Wikipedia was fine-tuned by integrating a classification model, with human labeled Tweets, as training data. A collection of 30 075 Tweets about JUUL flavors was classified into positive and negative sentiments. Finally, using topic models, we identify and grouped thematic areas into positive and negative Tweets. With an average of 89% cross-validation precision for classifying Tweets, the fine-tuned BERT model classified 24 114 Tweets as positive and 5961 Tweets as negative. Through the topic modeling approach 10 thematic topics were identified from the predicted positive and negative sentiments expressed in the Tweets. JUUL flavors, notably mango, mint, and cucumber, provoke overwhelmingly positive sentiments indicating a strong likeness due to favorable taste and odor. Negative discourse about JUUL flavors revolve around addictiveness, high nicotine content, and youth targeted marketing. Limiting the content related to flavors and positive perceptions on social media is necessary to minimize exposure to youth. The novel methodology used in this study may be adopted to monitor e-cigarette discourse periodically, as well as other critical public health phenomena online.

Monotone submodular subset for sentiment analysis of online reviews

Along with online social media’s prosperity, the amount of user-generated reviews dramatically increases. The kinds of text-based user-generated content are conducive to estimating public sentiments. Many sentiment analysis works are based on the assumption that the sentiment expressed in online reviews can be retrieved from general text features. However, text redundancy and quantity can potentially impact the analysis performance, especially when strict corpus size constraints are applied. This paper proposes a sentiment subset selection framework to construct a small set of documents from the original corpus to convey a subjective representation. The framework can filter irrelevant sentiment information based on topic modeling and select subsets by submodular maximization with respect to a cardinality constraint. Our proposed score function can facilitate the framework to capture fine-grained sentiment features expressed in reviews compared with the conventional submodular-based one. An empirical analysis for the efficacy of the proposed sentiment subset selection framework (SentiSS) on different context domains is conducted. The comparative study of the subset’s metric impact on different sentiment levels, namely positive, neural, and negative, is also performed. Experimental results show that the SentiSS framework can compress the sentiment corpus and maintain the classifier’s performance on the metrics at the same time.

AraSenCorpus: A Semi-Supervised Approach for Sentiment Annotation of a Large Arabic Text Corpus

At a time when research in the field of sentiment analysis tends to study advanced topics in languages, such as English, other languages such as Arabic still suffer from basic problems and challenges, most notably the availability of large corpora. Furthermore, manual annotation is time-consuming and difficult when the corpus is too large. This paper presents a semi-supervised self-learning technique, to extend an Arabic sentiment annotated corpus with unlabeled data, named AraSenCorpus. We use a neural network to train a set of models on a manually labeled dataset containing 15,000 tweets. We used these models to extend the corpus to a large Arabic sentiment corpus called “AraSenCorpus”. AraSenCorpus contains 4.5 million tweets and covers both modern standard Arabic and some of the Arabic dialects. The long-short term memory (LSTM) deep learning classifier is used to train and test the final corpus. We evaluate our proposed framework on two external benchmark datasets to ensure the improvement of the Arabic sentiment classification. The experimental results show that our corpus outperforms the existing state-of-the-art systems.

A Semi-supervised Approach for Sentiment Analysis of Arab(ic+izi) Messages: Application to the Algerian Dialect

In this paper, we propose a semi-supervised approach for sentiment analysis of Arabic and its dialects. This approach is based on a sentiment corpus, constructed automatically and reviewed manually by Algerian dialect native speakers. This approach consists of constructing and applying a set of deep learning algorithms to classify the sentiment of Arabic messages as positive or negative. It was applied on Facebook messages written in Modern Standard Arabic (MSA) as well as in Algerian dialect (DALG, which is a low resourced-dialect, spoken by more than 40 million people) with both scripts Arabic and Arabizi. To handle Arabizi, we consider both options: transliteration (largely used in the research literature for handling Arabizi) and translation (never used in the research literature for handling Arabizi). For highlighting the effectiveness of a semi-supervised approach, we carried out different experiments using both corpora for the training (i.e. the corpus constructed automatically and the one that was reviewed manually). The experiments were done on many test corpora dedicated to MSA/DALG, which were proposed and evaluated in the research literature. Both classifiers are used, shallow and deep learning classifiers such as Random Forest (RF), Logistic Regression(LR) Convolutional Neural Network (CNN) and Long short-term memory (LSTM). These classifiers are combined with word embedding models such as Word2vec and fastText that were used for sentiment classification. Experimental results (F1 score up to 95% for intrinsic experiments and up to 89% for extrinsic experiments) showed that the proposed system outperforms the existing state-of-the-art methodologies (the best improvement is up to 25%).

Experiments in Text Classification: Analyzing the Sentiment of Electronic Product Reviews in Greek

ABSTRACT Sentiment analysis, which deals with people’s sentiments as they appear in the growing amount of online social data, has been on the rise in the past few years. In its simplest form, sentiment analysis deals with the polarity of a given text, i.e., whether the opinion expressed in it is positive or negative. Sentiment analysis, or opinion mining applications on websites and the social media range from product reviews and brand reception to political issues and the stock market. The vast majority of the research in sentiment analysis has mostly dealt with English data, where there’s an abundance of readily available and annotated for sentiment corpora. With a few notable exceptions, the research in other minor languages such as Greek is lacking. This paper deals with sentiment analysis of electronic product reviews written in Greek. To this end, a small dataset of 480 positive and negative reviews is compiled and used, taken from the popular Greek e-commerce website, www.skroutz.gr. Different computational models for training and testing the dataset are evaluated, ranging from simple Naive Bayes with n-gram features to state-of-the-art BERT. The results look very promising for such a small corpus.

Automatic Indonesian Sentiment Lexicon Curation with Sentiment Valence Tuning for Social Media Sentiment Analysis

A novel Indonesian sentiment lexicon (SentIL -- Sentiment Indonesian Lexicon) is created with an automatic pipeline; from creating sentiment seed words, adding new words with slang words, emoticons, and from the given dictionary and sentiment corpus, until tuning sentiment value with tagged sentiment corpus. It begins by taking seed words from WordNet Bahasa that mapped with sentiment value from English SentiWordNet . The seed words are enriched by combining the dictionary-based method with words’ synonyms and antonyms, and corpus-based methods with word embedding for word similarity that trained in positive and negative sentiment corpus from online marketplaces review and Twitter data. The valence score of each lexicon is recalculated based on its relative occurrence in the corpus. We also add some famous slang words and emoticons to enrich the lexicon. Our experiment shows that the proposed method can provide an increase of 3.5 times lexicon number as well as improve the accuracy of 80.9% for online review and 95.7% for Twitter data, and they are better than other published and available Indonesian sentiment lexicons.

Stacked DeBERT: All attention in incomplete data for text classification

In this paper, we propose Stacked DeBERT, short for StackedDenoising Bidirectional Encoder Representations from Transformers. This novel model improves robustness in incomplete data, when compared to existing systems, by designing a novel encoding scheme in BERT, a powerful language representation model solely based on attention mechanisms. Incomplete data in natural language processing refer to text with missing or incorrect words, and its presence can hinder the performance of current models that were not implemented to withstand such noises, but must still perform well even under duress. This is due to the fact that current approaches are built for and trained with clean and complete data, and thus are not able to extract features that can adequately represent incomplete data. Our proposed approach consists of obtaining intermediate input representations by applying an embedding layer to the input tokens followed by vanilla transformers. These intermediate features are given as input to novel denoising transformers which are responsible for obtaining richer input representations. The proposed approach takes advantage of stacks of multilayer perceptrons for the reconstruction of missing words’ embeddings by extracting more abstract and meaningful hidden feature vectors, and bidirectional transformers for improved embedding representation. We consider two datasets for training and evaluation: the Chatbot Natural Language Understanding Evaluation Corpus and Kaggle’s Twitter Sentiment Corpus. Our model shows improved F1-scores and better robustness in informal/incorrect texts present in tweets and in texts with Speech-to-Text error in the sentiment and intent classification tasks.11https://github.com/gcunhase/StackedDeBERT.

ArAutoSenti: automatic annotation and new tendencies for sentiment classification of Arabic messages

A corpus-based sentiment analysis approach for messages written in Arabic and its dialects is presented and implemented. The originality of this approach resides in the automation construction of the annotated sentiment corpus, which relies mainly on a sentiment lexicon that is also constructed automatically. For the classification step, shallow and deep classifiers are used with features being extracted applying word embedding models. For the validation of the constructed corpus, we proceed with a manual reviewing and it was found that 85.17% were correctly annotated. This approach is applied on the under-resourced Algerian dialect and the approach is tested on two external test corpora presented in the literature. The obtained results are very encouraging with an F1 score that is up to 88% (on the first test corpus) and up to 81% (on the second test corpus). These results, respectively, represent a 20% and a 6% improvement, respectively, when compared with existing work in the research literature.

Arabic dialect sentiment analysis with ZERO effort. \\ Case study: Algerian dialect

This paper presents an analytic study showing that it is entirely possible to analyze the sentiment of an Arabic dialect without constructing any resources. The idea of this work is to use the resources dedicated to a given dialect \textit{X} for analyzing the sentiment of another dialect \textit{Y}. The unique condition is to have \textit{X} and \textit{Y} in the same category of dialects. We apply this idea on Algerian dialect, which is a Maghrebi Arabic dialect that suffers from limited available tools and other handling resources required for automatic sentiment analysis. To do this analysis, we rely on Maghrebi dialect resources and two manually annotated sentiment corpus for respectively Tunisian and Moroccan dialect. We also use a large corpus for Maghrebi dialect. We use a state-of-the-art system and propose a new deep learning architecture for automatically classify the sentiment of Arabic dialect (Algerian dialect). Experimental results show that F1-score is up to 83% and it is achieved by Multilayer Perceptron (MLP) with Tunisian corpus and with Long short-term memory (LSTM) with the combination of Tunisian and Moroccan. An improvement of 15% compared to its closest competitor was observed through this study. Ongoing work is aimed at manually constructing an annotated sentiment corpus for Algerian dialect and comparing the results

Corpus Development for Malay Sentiment Analysis Using Semi Supervised Approach

Research on sentiment analysis have gained so much interest currently. However, research on Malay sentiment analysis and the availability of the resources is still lacking. The aim of this study is to develop a Malay sentiment corpus using a semi-supervised approach. Data from Twitter have been used in this corpus development. The corpus is developed using the combination of lexicon and machine learning approach. The sentiment lexicon will be used to build the seed of training data from unlabelled data resources. In addition, sentiment emoticons are used to compare the accuracy of the lexicon-based approach. After preparation of training data set, the process of adding new training data instances will be carried out using the seed set and machine learning classification method. The process of classification using machine learning approach consists of pre-processing, feature extraction and classification. Five types of classifiers are considered for the classification task. Based on the experimental results, the lexicon-based approach and Multinomial Naïve Bayes algorithm is the best classifier for Malay sentiment corpus development.

Cross-lingual deep neural transfer learning in sentiment analysis

Abstract In this article, we present a novel technique for the use of language-agnostic sentence representations to adapt the model trained on texts in Polish (as a low-resource language) to recognize polarity in texts in other (high-resource) languages. The first model focuses on the creation of a language-agnostic representation of each sentence. The second one aims to predict the sentiment of the text based on these sentence representations. Besides models evaluation on PolEmo 1.0 Sentiment Corpus, we also conduct a proof of concept for using a deep neural network model trained only on language-agnostic embeddings of texts in Polish to predict the sentiment of the texts in MultiEmo-Test 1.0 Sentiment Corpus, containing PolEmo 1.0 test datasets translated into eight different languages: Dutch, English, French, German, Italian, Portuguese, Russian and Spanish. Both corpora are publicly available under a Creative Commons copyright license.

A mixed approach of statistical weighting method and unsupervised method to improve Uyghur sentiment classification

Considering the scarcity of Uyghur sentiment resources, in this paper proposed a new combined unsupervised sentiment classification method for Uyghur text without any labeled corpora. In the first part, a Uyghur sentiment dictionary, UYSentiDict, was adopted to classify the sentences. For the sentiment vocabulary matching, both the matching of the original word and the stem were considered, and the influence of sentence patterns, negation words, and degree adverbs were further considered as well. Based on different thresholds, the sentences with higher sentiment values were selected from the lexicon-based classification results as a pseudo-labeled dataset. In the second part, different sentiment characteristics were learned from the pseudo-labeled dataset by the machine learning classifier, and the remaining categorical data were further classified. It can be concluded that the method proposed in this paper has good classification efficiency in Uyghur sentiment corpora in four different fields, and some results were performed better than the classification results of machine learning classifier. Moreover, this method is not restricted by the field of data and does not need to be marked in advance with good training corpus, and can solve the resource shortage problem in the field of Uyghur sentiment classification effectively.

Arabic sentiment analysis: studies, resources, and tools

To determine whether a document or a sentence expresses a positive or negative sentiment, three main approaches are commonly used: the lexicon-based approach, corpus-based approach, and a hybrid approach. The study of sentiment analysis in English has the highest number of sentiment analysis studies, while research is more limited for other languages, including Arabic and its dialects. Lexicon based approaches need annotated sentiment lexicons (containing the valence and intensity of its terms and expressions). Corpus-based sentiment analysis requires annotated sentences. One of the significant problems related to the treatment of Arabic and its dialects is the lack of these resources. We present in this survey the most recent resources and advances that have been done for Arabic sentiment analysis. This survey presents recent work (where the majority of these works are between 2015 and 2019). These works are classified by category (survey work or contribution work). For contribution work, we focus on the construction of sentiment lexicon and corpus. We also describe emergent trends related to Arabic sentiment analysis, principally associated with the use of deep learning techniques.

A Check on Annotation in Sentiment Research

The research literature on sentiment analysis methodologies has exponentially grown in recent years. In any research area, where new concepts and techniques are constantly introduced, it is, therefore, of interest to analyze the latest trends in this literature. In particular, we have chosen to primarily focus on the literature of the last five years, on annotation methodologies, including frequently used datasets and from which they were obtained. Based on the survey, it appears that researchers do more manual annotation in the formation of sentiment corpus. As for the dataset, there are still many uses of English language taken from social media such as Twitter. In this area of research, there are still many that need to be explored, such as the use of semi-automatic annotation method that is still very rarely used by researchers. Also, less popular languages, such as Malay, Korean, Japanese, and so on, still require corpus for sentiment analysis research.

Sentiment Classification of Reviews Based on BiGRU Neural Network and Fine-grained Attention

Text sentiment analysis is part and parcel of natural language processing. The task of sentiment classification is actually the process of feature extraction through models. The comment text of commodities is very different from the ordinary text. The comment text has no fixed grammar and writing format and the sentiment feature information is scattered in various places of text. Due to these factors, model learning of sentiment classification is becoming increasingly complex. The paper aims at establishing a fine-grained feature extraction model based on BiGRU and attention. Firstly, the vocabulary is vectorized by means of the skip-gram model. Then, according to the pre-trained word vector, the sentiment words list can be reached and noise filtering would be conducted by Naive Bayes algorithm. Finally, the model extracts features using BiGRU and fine-grained attentions. Based on the hypothesis that a long review may lead to feature differentiation, a fine-grained attention model is proposed. In this model, the attention layer is design to focus on the feature in different level such as word level, sentence level and paragraph level. This paper validate the proposed model on two sentiment corpus JD reviews and IMDB. Empirical results show that the FGAtten-BiGRU model achieves state of the art results on sentiment analysis tasks.

Attention-based long short-term memory network using sentiment lexicon embedding for aspect-level sentiment analysis in Korean

Although deep learning breakthroughs in NLP are based on learning distributed word representations by neural language models, these methods suffer from a classic drawback of unsupervised learning techniques. Furthermore, the performance of general-word embedding has been shown to be heavily task-dependent. To tackle this issue, recent researches have been proposed to learn the sentiment-enhanced word vectors for sentiment analysis. However, the common limitation of these approaches is that they require external sentiment lexicon sources and the construction and maintenance of these resources involve a set of complexing, time-consuming, and error-prone tasks. In this regard, this paper proposes a method of sentiment lexicon embedding that better represents sentiment word's semantic relationships than existing word embedding techniques without manually-annotated sentiment corpus. The major distinguishing factor of the proposed framework was that joint encoding morphemes and their POS tags, and training only important lexical morphemes in the embedding space. To verify the effectiveness of the proposed method, we conducted experiments comparing with two baseline models. As a result, the revised embedding approach mitigated the problem of conventional context-based word embedding method and, in turn, improved the performance of sentiment classification.

Sentiment analysis through recurrent variants latterly on convolutional neural network of Twitter

Sentiment analysis has been a hot area in the exploration field of language understanding, however, neural networks used in it are even lacking. Presently, the greater part of the work is proceeding on recognizing sentiments by concentrating on syntax and vocabulary. In addition, the task identified with natural language processing and for computing the exceptional and remarkable outcomes Recurrent neural networks (RNNs) and Convolutional neural networks (CNNs) have been utilized. Keeping in mind the end goal to capture the long-term dependencies CNNs, need to rely on assembling multiple layers. In this Paper for the improvement in understanding the sentiments, we constructed a joint architecture which places of RNN at first for capturing long-term dependencies with CNNs using global average pooling layer while on top a word embedding method using GloVe procured by unsupervised learning in the light of substantial twitter corpora to deal with this problem. Experimentations exhibit better execution when it is compared with the baseline model on the twitter’s corpora which tends to perform dependable results for the analysis of sentiment benchmarks by achieving 90.59% on Stanford Twitter Sentiment Corpus, 89.46% on Sentiment Strength Twitter Data and 88.72% on Health Care Reform Dataset respectively. Empirically, our work turned to be an efficient architecture with slight hyperparameter tuning which capable us to reduce the number of parameters with higher performance and not merely relying on convolutional multiple layers by constructing the RNN layer followed by convolutional layer to seizure long-term dependencies.

Mapping Consumer Sentiment Toward Wireless Services Using Geospatial Twitter Data

Hyper-dense wireless network deployment is one of the popular solutions to meeting high capacity requirement for 5G delivery. However, current operator understanding of consumer satisfaction comes from call centers and base station quality-of-service (QoS) reports with poor geographic accuracy. The dramatic increase in geo-tagged social media posts adds a new potential to understand consumer satisfaction towards target-specific quality-of-experience (QoE) topics. In our paper, we focus on evaluating users’ opinion on wireless service-related topics by applying natural language processing (NLP) to geo-tagged Twitter data. Current generalized sentiment detection methods with generalized NLP corpora are not topic specific. Here, we develop a novel wireless service topic-specific sentiment framework, yielding higher targeting accuracy than generalized NLP frameworks. To do so, we first annotate a new sentiment corpus called SignalSentiWord (SSW) and compare its performance with two other popular corpus libraries, AFINN and SentiWordNet. We then apply three established machine learning methods, namely: Naive Bayes (NB), Support Vector Machine (SVM), and Recurrent Neural Network (RNN) to build our topic-specific sentiment classifier. Furthermore, we discuss the capability of SSW to filter noisy and high-frequency irrelevant words to improve the performance of machine learning algorithms. Finally, the real-world testing results show that our proposed SSW improves the performance of NLP significantly.

Identification of fact-implied implicit sentiment based on multi-level semantic fused representation

Sentiment can be expressed in an explicit or implicit manner. Most of the current studies on sentiment analysis focus on the identification of explicit sentiment but ignore the implicit. According to our statistics during data labeling in previous work, nearly a third of subjective sentences contain implicit sentiment, and 72% of the implicit sentiment sentences are fact-implied ones. We analyze the characteristics of the sentences that express fact-implied implicit sentiment and consider that fact-implied implicit sentiment is usually affected by its sentiment target, context semantic background and its own sentence structure. This paper focuses on the recognition of fact-implied implicit sentiment at the sentence level. A multi-level semantic fusion method is proposed to learn the features for identification based on representation learning. Three features in different levels are learned from the corpus, namely, sentiment target representation at the word level, structure embedded representation at the sentence level and context semantic background representation at the document level. We manually construct a fact-implied implicit sentiment corpus in Chinese, and experiments on the datasets show that the proposed method can effectively recognize fact-implied implicit sentiment sentences.