Search

This research presents a way of feature selection problem for classification of sentiments that use ensemble-based classifier. This includes a hybrid approach of minimum redundancy and maximum relevance (mRMR) technique and Forest Optimization Algorithm (FOA) (i.e. mRMR-FOA) based feature selection. Before applying the FOA on sentiment analysis, it has been used as feature selection technique applied on 10 different classification datasets publically available on UCI machine learning repository. The classifiers for example k-Nearest Neighbor (k-NN), Support Vector Machine (SVM) and Naïve Bayes used the ensemble based algorithm for available datasets. The mRMR-FOA uses the Blitzer’s dataset (customer reviews on electronic products survey) to select the significant features. The classification of sentiments has noticed to improve by 12 to 18%. The evaluated results are further enhanced by the ensemble of k-NN, NB and SVM with an accuracy of 88.47% for the classification of sentiment analysis task.

This study aims to determine the effect of price and product reviews on online purchasing decisions through Shopee during the Covid-19 pandemic in Banjarnegara Regency, Central Java. This study uses a quantitative approach. The data collection technique was conducted with a questionnaire. The number of samples in this study was 120 respondents using the purposive sampling method or with certain criteria, namely, the respondents are Shopee consumers who live in Banjarnegara Regency, Central Java, and have experience doing online transactions at Shopee at least 2 times since the outbreak of the Covid-19 pandemic in Indonesia. The data analysis technique used validity test, reliability test, multiple linear regression analysis and t-test (partial), and F test (simultaneous), and the coefficient of determination (R2). The data processing uses SPSS version 26.0. The results of this study indicate that prices and product reviews partially and simultaneously influence online purchasing decisions through Shopee during the Covid-19 pandemic in Banjarnegara Regency, Central Java.

Sentiment analysis is defined as the process of mining of data, view, review or sentence to Predict the emotion of the sentence through natural language processing (NLP) or Machine Learning Techniques. The sentiment analysis involve classification of text into three phase “Positive”, “Negative” or “Neutral”. The process of finding user Opinion about the topic or Product or problem is called as opinion mining. Analyzing the emotions from the extracted Opinions are defined as Sentiment Analysis. The goal of opinion mining and Sentiment Analysis is to make computer able to recognize and express emotion. Using social media, E-commerce website, movies reviews such as Face book, twitter, Amazon, Flipkart etc. user share their views, feelings in a convenient way. Sentiment analysis in a machine learning approach in which machines classify and analyze the human’s sentiments, emotions, opinions etc. about the products. Out of the various classification models, Naïve Bayes, Support Vector Machine (SVM) and Decision Tree are used maximum times for the product analysis. The proposed approach will do better result as compare to other machine learning techniques.

Online reviews and ratings play an important role in shaping the purchase decisions of customers in e-commerce. Many researches have been done to make proper recommendations for users, by exploiting reviews, ratings, user profiles, or behaviors. However, the dynamic evolution of user preferences and item properties haven’t been fully exploited. Moreover, it lacks fine-grained studies at the aspect level. To address the above issues, we define two concepts of user maturity and item popularity, to better explore the dynamic changes for users and items. We strive to exploit fine-grained information at the aspect level and the evolution of users and items, for dynamic sentiment prediction. First, we analyze three real datasets from both the overall level and the aspect level, to discover the dynamic changes (i.e., gradual changes and sudden changes) in user aspect preferences and item aspect properties. Next, we propose a novel model of Aspect-based Sentiment Dynamic Prediction (ASDP), to dynamically capture and exploit the change patterns with uniform time intervals. We further propose the improved model ASDP+ with a bin segmentation algorithm to set the time intervals non-uniformly based on the sudden changes. Experimental results on three real-world datasets show that our work leads to significant improvements.

At this current digital era, business platforms have been drastically shifted toward online stores on internet. With the internet-based platform, customers can order goods easily using their smart phones and get delivery at their place without going to the shopping mall. However, the drawback of this business platform is that customers do not really know about the quality of the products they ordered. Therefore, such platform service often provides the review section to let previous customers leave a review about the received product. The reviews are a good source to analyze customer's satisfaction. Business owners can assess review trend as either positive or negative based on a feedback score that customers had given, but it takes too much time for human to analyze this data. In this research, we develop computational models using machine learning techniques to classify product reviews as positive or negative based on the sentiment analysis. In our experiments, we use the book review data from amazon.com to develop the models. For a machine learning based strategy, the data had been transformed with the bag of word technique before developing models using logistic regression, naïve bayes, support vector machine, and neural network algorithms. For a deep learning strategy, the word embedding is a technique that we used to transform data before applying the long short-term memory and gated recurrent unit techniques. On comparing performance of machine learning against deep learning models, we compare results from the two methods with both the preprocessed dataset and the non-preprocessed dataset. The result is that the bag of words with neural network outperforms other techniques on both non-preprocess and preprocess datasets.

Spam detection frequently categorizes product reviews as spam and non-spam. The spam reviews may contain texts of fake reviews and non-review statements describing unrelated things about products. Most of the publicly available spam reviews are labelled as fake reviews, while non-spam texts that are not fake reviews could contain non-review statements. It is crucial to notice those non-review statements since they convey misperception to consumers. Non-review statements are hardly found, and those statements of large and long texts often need to be manually labelled, which is time-consuming. Because of the rareness in finding non-review statements, there is an imbalanced condition between non-spam as a major class and spam that consists of the non-review statement as a minor class. Augmenting fake reviews to add spam texts is ineffective because they have similar content to non-spam such as some opinion words of product features. Thus, the text generation of non-review statements is preferable for adding spam texts. Some text generation issues are the frequent neural network-based methods require much learning data, and the existing pre-trained models produce texts with different contexts to non-review statements. The augmented texts should have similar content and context represented by the structure of the non-review statement. Therefore, we propose a text generation model with content and structure-based preprocessing to produce non-review statements, which is expected to overcome imbalanced data and give better spam detection results in product reviews. Structure-based preprocessing identifies the feature structures of non-opinion words from part-of-speech tags. Those features represent the context of spam reviews in unlabeled texts. Then, content-based preprocessing appoints selected topic modeling results of non-review statements from fake reviews. Our experiments resulted an improvement on the metric value of ± 0.04, called as BLEU (Bi-Lingual Evaluation Understudy) score, for the correspondence evaluation between generated and trained texts. The metric value indicates that the generated texts are not quite identical to the trained texts of non-review statements. However, those additional texts combined with the original spam texts gave better spam detection results with an increasing value of more than 40% on average recall score.

Nowadays, digital reviews and ratings of E-commerce platforms provide a better way for consumers to buy the products. E-commerce giants like Amazon, Flipkart, etc provide customers with a forum to share their experience and provide potential consumers with true evidence of the product's outcomes. To obtain useful insights from a broad collection of reviews, it is important to separate reviews into positive and negative feelings. In the proposed work, Sentiment Analysis is to be done on the consumer review data and categorize into positive and negative feelings. Naive Bayes Classification, LSTM and Support Vector Machine (SVM) were employed for the classification of reviews from the various classification models. Many of the current SA techniques for these customer online product review text data have low accuracy and often takes longer time in the course of training. In this research work, BERT Base Uncased model which is a powerful Deep Learning Model is presented to elucidate the issue of Sentiment Analysis. The BERT model gave an improved performance with good prediction and high accuracy compared to the other methods of Machine Learning in the experimental evaluation.

Voting mechanisms are widely adopted for evaluating the quality and credibility of user-generated content, such as online product reviews. For the reviews that do not receive sufficient votes, techniques and models are developed to automatically assess their helpfulness levels. Existing methods serving this purpose are mostly centered on feature analysis, ignoring the information conveyed in the frequencies and patterns of user votes. Consequently, the accuracy of helpfulness measurement is limited. Inspired by related findings from prediction theories and consumer behavior research, we propose a novel approach characterized by the technique of iterative Bayesian distribution estimation, aiming to more accurately measure the helpfulness levels of reviews used for training prediction models. Using synthetic data and a real-world data set involving 1.67 million reviews and 5.18 million votes from Amazon, a simulation experiment and a two-stage data experiment show that the proposed approach outperforms existing methods on accuracy measures. Moreover, an out-of-sample user study is conducted on Amazon Mechanical Turk. The results further illustrate the predictive power of the new approach. Practically, the research contributes to e-commerce by providing an enhanced method for exploiting the value of user-generated content. Academically, we contribute to the design science literature with a novel approach that may be adapted to a wide range of research topics, such as recommender systems and social media analytics.

References In recent years, online customer reviews have become the main source to determine public opinion about offered products and services. Therefore, manufacturers and sellers are extremely concerned with customer reviews, as these can have a direct impact on their businesses. Unfortunately, there is an increasing trend to write spam reviews to promote or demote targeted products or services. This practice, known as review spamming, has posed many questions regarding the authenticity and dependability of customers’ review-based business processes. Although the spam review detection (SRD) problem has gained much attention from researchers, existing studies on SRD have mostly worked on datasets of English, Chinese, Arabic, Persian, and Malay languages. Therefore, the objective of this research is to identify the spam in Roman Urdu reviews using different classification models based on linguistic features and behavioral features. The performance of each classifier is evaluated in a number of perspectives: (i) linguistic features are used to calculate accuracy (F1 score) of each classifier; (ii) behavioral features combined with distributional and non-distributional aspects are used to evaluate accuracy (F1 score) of each classifier; and (iii) the combination of both linguistic and behavioral features (distributional and non-distributional aspects) are used to evaluate the accuracy of each classifier. The experimental evaluations demonstrated an improved accuracy (F1 score: 0.96), which is the result of combinations of linguistic features and behavioral features with the distributional aspect of reviewers. Moreover, behavioral features using distributional characteristic achieve an accuracy (F1 score: 0.86) and linguistic features shows accuracy (F1 score: 0.69). The outcome of this research can be used to increase customers’ confidence in the South Asian region. It can also help to reduce spam reviews in the South Asian region, particularly in Pakistan.

Aloe vera soothing gel is one of the best-selling products and the most widely reviewed on the Althea Korea website. This product has been reviewed by 1,448 users on the Althea website. The result of the research can be used to minimize mistakes in product purchases. Besides, through a review of a product, the company can analyze the level of customer satisfaction and can be a suggestion for improvements in the future. Therefore, a system is needed to analyze the sentiment towards aloe vera soothing gel to determine the review as a positive or negative sentiment. The method used in this research is the Naïve Bayes method and uses the classification carried out by linguists as a reference for determining positive and negative sentiment. There are two tests carried out in this research, namely confusion matrix testing and black-box testing. The result of the confusion matrix test found an accuracy of 94.62% and the result of black-box testing showed that the output produced was by the application functionality.

Sentiment classification aims to detect polarity from a piece of text. The polarity is usually positive or negative, and the text genre is usually product review. The challenges of sentiment classification are that it is hard to capture semantic of reviews, and the labeled data is hard to annotate. Therefore, we propose neural co-training to learn the semantic representation of each review using the neural network model, and learn the information from unlabeled data using a co-training framework. In particular, we use the attention-based bi-directional Gated Recurrent Unit (Att-BiGRU) to model the semantic content of each review and regard different categories of the target product as different views. We then use a co-training framework to learn and predict the unlabeled reviews with different views. Experiment results with the Yelp dataset demonstrate the effectiveness of our approach.

An important step when designers use machine learning models is annotating user generated content. In this study we investigate inter-rater reliability measures of qualitative annotations for supervised learning. We work with previously annotated product reviews from Amazon where phrases related to sustainability are highlighted. We measure inter-rater reliability of the annotations using four variations of Krippendorff's U-alpha. Based on the results we propose suggestions to designers on measuring reliability of qualitative annotations for machine learning datasets.

Sentiment analysis can be a very useful aspect for the extraction of useful information from text documents. The main idea for sentiment analysis is how people think for a particular online review, i.e. product reviews, movie reviews, etc. Sentiment analysis is the process where these reviews are classified as positive or negative. The web is enriched with huge amount of reviews which can be analyzed to make it meaningful. This article presents the use of lexicon resources for sentiment analysis of different publicly available reviews. First, the polarity shift of reviews is handled by negations. Intensifiers, punctuation and acronyms are also taken into consideration during the processing phase. Second, words are extracted which have some opinion; these words are then used for computing score. Third, machine learning algorithms are applied and the experimental results show that the proposed model is effective in identifying the sentiments of reviews and opinions.

Purpose:This research aims to mine review data on one of the e-commerce sites which ultimately produces clusters using the K-Means Clustering algorithm that can help potential customers to make a decision before deciding to buy a product or service. Design/methodology/approach: By using Octoparse we mine opinion or comment data in the form of customer online reviews, after getting the data we group the data using the k-emans clustering methode to obtain cluster Findings: Cluster Analysys can can help potential customers to make a decision before deciding to buy a product or service Research limitations/implications: WWW.Lazada.Com Practical implications: State your implication here. Originality/value: Paper type: This paper can be categorized as case study paper

Sentiment analysis plays an important role in understanding individual opinions expressed in websites such as social media and product review sites. The common approaches to sentiment analysis use the sentiments carried by words that express opinions and are based on either supervised or unsupervised learning techniques. The unsupervised learning approach builds a word-sentiment dictionary, but it requires lengthy time periods and high costs to build a reliable dictionary. The supervised learning approach uses machine learning models to learn the sentiment scores of words; however, training a classifier model requires large amounts of labelled text data to achieve a good performance. In this article, we propose a semisupervised approach that performs well despite having only small amounts of labelled data available for training. The proposed method builds a base sentiment dictionary from a small training dataset using a lasso-based ensemble model with minimal human effort. The scores of words not in the training dataset are estimated using an adaptive instance-based learning model. In a pretrained word2vec model space, the sentiment values of the words in the dictionary are propagated to the words that did not exist in the training dataset. Through two experiments, we demonstrate that the performance of the proposed method is comparable to that of supervised learning models trained on large datasets.

As the digital information technology domain is growing unprecedently, the process of making the devices mimic human like actions has gained an increased research attention. One of the main features of human like system is the human emotion or sentiment detection. Online shopping has made it easy for everyone in the world to buy different varieties of products from a single place. It also gives the opportunity for them to try new products. While trying to buy new products, people check for the reviews and ratings of those products given by the other customers. Those reviews and ratings can be used to determine the sentimentality of users, who have bought those products by using text mining techniques and classify them either as positive or negative reviews. To analyze these data, the proposed research work has used both Naive Bayes and linear regression algorithms. In our analysis, the accuracy rate of both algorithms is checked.

The purchase of products or administrations through an electronic trade called web based shopping over internet using a web browser. Online Product surveys are significant for up and coming purchasers in helping them decide. To this end, distinctive sentiment mining systems have been proposed, where making a decision about a survey sentence’s direction (e.g., positive or negative) is one of their key difficulties. As of late, Machine learning has risen as powerful methods for taking care of assumption order issues. An AI model inherently learns a helpful portrayal consequently without human endeavors. In any case, we propose a regulated AI structure for item audit conclusion arrangement which utilizes pervasively accessible evaluations as powerless supervision signals. To assess the proposed system, we build a dataset containing 2, 00,000 pitifully named survey sentences and 15000 marked audit sentences from Amazon. Trial results show the more exactness contrasted with past one.

Vietnamese is a tonal and isolated language. Its highly ambiguity makes the designing of methods for sentiment analysis being difficult. For getting the most effectiveness, the designed method has to analyze sentiment of sentences based on combining the grammar and syllable structures of Vietnamese. In this paper, a method to build a Vietnamese dataset of product reviews with many sentiment levels, including very negative, negative, neutral, positive and very positive, is proposed. This method can be scaled to a large dataset using for analyzing sentiment of product reviews. Moreover, a solution to add more grammar rules of Vietnamese into the pre-processing of sentiment analysis is also constructed. Those rules simulate the sentiment recognition of humans and help to increase the accuracy of sentiment determination. The combination of grammar rules and some methods for sentiment analysis are experimented on the Vietnamese dataset of product reviews to classify them into sentiment-levels. The testing results show that their accuracy and F-measure are improved and suitable to apply in the practical business analyzing of customer behaviors.

The increasing use of online retail platforms has generated an enormous amount of textual data on the user experiences with these products in online reviews. These reviews provide a rich resource to elicit customer requirements for a category of products. The recent research has explored this possibility to some extent. The study reported here investigates the coding of publically available user reviews to understand customer sentiments on environmentally-friendly products. The manual review typically consists of a qualitative analysis of textual content, which is a resource-intensive process. An automated procedure based on Aspect-Based Sentiment Analysis (ABSA) is proposed and explored in this study. This procedure can be beneficial in analyzing reviews of products that belong to a specific category. As a case study, environmentally-friendly products are used. Manual content analysis and automated ABSA-based analysis are performed on the same review data to extract customer sentiments. The results show that we obtain over a 50% classification accuracy for a multiclass classification NLP task with a very elementary word vector-based model. The drop in accuracy (compared to human annotation) can be offset because an automated system is thousands of times faster than a human. Given enough data, it will perform better than its human counterpart in tasks on customer requirement modeling. We also discuss the future routes that can be taken to extend our system by leveraging more sophisticated paradigms and substantially improving our system’s performance.

Research into utilising text classification to analyse product reviews from e-commerce websites has increased tremendously in recent years. Machine Learning and Deep Learning classifiers have been utilised to organise, categorise and classify product reviews, enabling the identification of polarity and sentiment within product reviews. In this paper, we propose a methodology to classify product reviews using machine learning and deep learning with the intention to identify and predict the activity (use case) in which the consumer used the product they have reviewed.

A product review is needed by a customer before he buys a product. Currently, several platforms can be used to provide product reviews, one of which is the beauty product. Every customer can read beauty product reviews, not only from one aspect of the review but it can be from several aspects of the review. it is difficult for consumers to find all the reviews from various aspects quickly. Therefore, in this study, a combination of LDA modeling methods and Word Embedding Word2vec were used, to obtain sentiments from each of the predetermined aspects of the review. In this study, the accuracy of the combination of LDA will be compared with the Word2vec Skip-gram and Continuous-bag-of-word (CBOW) models. From the two combinations, it is found that the combination accuracy of LDA and Word2vec Skip gram is 80.36%, and for CBOW is only 74.37%. Meanwhile, the SVM and K-Fold Cross-Validation algorithms are used to find the accuracy of sentiment predictions on the aspects of price, packaging, and fragrances. Compared to the other two aspects, the packaging aspect has the highest accuracy at 89.71%.

In this paper, we propose the Brand-Topic Model (BTM) which aims to detect brand-associated polarity-bearing topics from product reviews. Different from existing models for sentiment-topic extraction which assume topics are grouped under discrete sentiment categories such as ‘positive’, ‘negative’ and ‘neural’, BTM is able to automatically infer real-valued brand-associated sentiment scores and generate fine-grained sentiment-topics in which we can observe continuous changes of words under a certain topic (e.g., ‘shaver’ or ‘cream’) while its associated sentiment gradually varies from negative to positive. BTM is built on the Poisson factorisation model with the incorporation of adversarial learning. It has been evaluated on a dataset constructed from Amazon reviews. Experimental results show that BTM outperforms a number of competitive baselines in brand ranking, achieving a better balance of topic coherence and unique-ness, and extracting better-separated polarity-bearing topics.

Visible Body has created a teaching and learning platform and mobile apps for anatomy and physiology instruction. For this review I will be evaluating the instructor courseware and its accompanying apps: Anatomy & Physiology, Human Anatomy Atlas, Muscle Premium, and Physiology Animations. I will be using an iPad with the apps but they function with both iOS and Android.

Everyday many online product sales websites and specialized reviewing forums publish a massive volume of human-generated product reviews. People use these reviews as valuable free source of knowledge when decide to buy products. Therefore, an accurate automated system for distinguishing useful reviews from non-useful ones is of great importance. This article presents a new model for specifying the usefulness of comments using the textual features extracted from the reviews. Various types of features including emotion-related, linguistic and text-related features, valence, arousal, and dominance (VAD) values, review-length and polarity of comments are exploited in this study. Moreover, two new algorithms are presented: an improved evidential algorithm for emotion recognition, and an algorithm for extracting VAD values for each review. Finally, the usefulness of reviews is predicted using the mentioned features and an improved Dempster-Shafer score fusion algorithm. The proposed method is applied to review datasets of Books and Video Games of Amazon. The results show that combining the features associated with emotions, features of VAD, and text-related features improves the accuracy of predicting the usefulness of reviews. Also, in comparison with the original Dempster-Shafer method, the precision of the improved Dempster-Shafer algorithm for both datasets is 15% and 11% higher, respectively.

Sentiment analysis is also known as opinion mining which shows the people’s opinions and emotions about certain products or services. The main problem in sentiment analysis is the sentiment polarity categorization that determines whether a review is positive, negative or neutral. Previous studies proposed different techniques, but still there are some research gaps, i) some studies include only 3 sentiment classes: positive, neutral and negative, but none of them considered more than 3 classes ii) sentiment polarity features were considered on individual basis but none of them considered on both individual and on combined basis iii) No previous technique considered five sentiment classes with 3 sentiment polarity features such as a verb, adverb, adjective and their combinations. In this study, we propose a sentiment polarity categorization technique for a large data set of online reviews of Instant Videos. A comprehensive data set of five hundred thousand online reviews is used in our research. There are five classes (Strongly Negative, Negative, Neutral, Positive and Strongly Positive). We also consider three polarity features Verb, Adverb, Adjective and their combinations with their different senses in review-level categorization. Our experiments for review-level categorization show promising outcomes as the accuracy of our results is 81 percent which is 3 percent better than many previous techniques whose average accuracy is 78 percent.

In recent years, with the development of e-commerce, the scale of comment data has shown an exponential growth trend. In this paper, a product review hot spot discovery algorithm based on MapReduce-PR-HD is proposed. The algorithm uses the Vector Space Model to vectorize the text data of the reviews, and utilize the TF-IDF algorithm to calculate the position weight of the feature words, then combines the Canopy algorithm and the K-Means algorithm to achieve the hot spot discovery of product reviews. At the same time, the algorithm obtain the ability to process massive data through the MapReduce framework. Experiments demonstrate that the PR-HD algorithm has high accuracy and parallel efficiency. This allows product developers to obtain more direct and effective suggestions and feedback, which allows product developers to obtain more direct and effective suggestions and feedback.

With the popularity of e-commerce, posting online product reviews expressing customer’s sentiment or opinion towards products has grown exponentially. Sentiment analysis is a computational method that plays an essential role in automating the extraction of subjective information i.e. customer’s sentiment or opinion from online product reviews. Two approaches commonly used in Sentiment analysis tasks are supervised approaches and lexicon-based approaches. In supervised approaches, Sentiment analysis is seen as a text classification task. The result depends not only on the robustness of the machine learning algorithm but also on the utilized features. Bag-of-word is a common utilized features. As a statistical feature, bag-of-word does not take into account semantic of words. Previous research has indicated the potential of semantic in supervised SA task. To augment the result of sentiment analysis, this paper proposes a method to extract text features named sentence level features (SLF) and domain sensitive features (DSF) which take into account semantic of words in both sentence level and domain level of product reviews. A word sense disambiguation based method was adapted to extract SLF. For every similarity employed in generating SLF, the SentiCircle-based method was enhanced to generate DSF. Results of the experiments indicated that our proposed semantic features i.e. SLF and SLF + DSF favorably increase the performance of supervised sentiment analysis on product reviews.

With the rapid development of the economy, product design has gradually shifted to emotional design that focuses on satisfying users’ emotional needs. Kansei engineering is the commonly used method in product emotional design, the first and vital stage of which needed to be addressed is the acquisition of Kansei knowledge. Considering the development of natural language processing technology and online shopping, a computerized method to extract Kansei knowledge from online product reviews is firstly proposed in this article, and a relational extraction method to establish the relationship between product features and user perceptions is further provided. This article analyzes and extracts the Kansei words of 10 mice respectively using the proposed computerized method, taking the mouse as the case study. Then three evaluation indicators including diversity, effectiveness, and concentration are defined to assess the method, which evaluates the superiority with the advantage of 19.03% in diversity, 6.91% in effectiveness, 22.18% in the concentration and 8.9 times higher in the total score compared with traditional method. Furthermore, taking the best-selling mouse for example, the relational extraction method is applied to extract the relationship between the user concern and the user attitude, establish the relational table, draw Kansei knowledge tree, and finally model connection between product features and user perceptions. By utilizing natural language processing technology and integrating Kansei engineering, linguistics and computer science, it could be considered that the results of this article can accelerate the traditional user survey process, clarify users’ emotional needs, guide the adjustment of product design, and assist the user-centered product emotional design.

Product reviews are valuable for upcoming buyers in helping them make decisions. To this end, different opinion mining techniques have been proposed, where judging a review sentence�s orientation (e.g. positive or negative) is one of their key challenges. Recently, deep learning has emerged as an effective means for solving sentiment classification problems. Deep learning is a class of machine learning algorithms that learn in supervised and unsupervised manners. A neural network intrinsically learns a useful representation automatically without human efforts. However, the success of deep learning highly relies on the large-scale training data. We propose a novel deep learning framework for product review sentiment classification which employs prevalently available ratings supervision signals. The framework consists of two steps: (1) learning a high-level representation (an embedding space) which captures the general sentiment distribution of sentences through rating information; (2) adding a category layer on top of the embedding layer and use labelled sentences for supervised fine-tuning. We explore two kinds of low-level network structure for modelling review sentences, namely, convolutional function extractors and long temporary memory. Convolutional layer is the core building block of a CNN and it consists of kernels. Applications are image and video recognition, natural language processing, image classification

Consumer reviews have emerged as one of the most influential factors in a person's purchase behavior. The existing open-source approaches for detecting expert reviewers and determining product ratings suffer from limitations and are susceptible to manipulation. In this work, we addressed these limitations by developing two algorithms and evaluated them on three datasets from amazon.com (the largest dataset contains nearly eight million reviews). In the first algorithm, we used a combination of the existing open-source approaches such as filtering by volume of contribution, helpfulness ratio, volume of helpfulness, and deviation from the estimated actual rating to detect the experts. The second algorithm is based on link analytic mutual iterative reinforcement of product ratings and reviewers' weights. In the second algorithm, both reviewers and products carry weights reflecting their relative importance. The reviewers influence the product rating according to their weight. Similarly, the reviewers' weights are impacted by their amount of deviation from the estimated actual product rating and the product's weight. Our evaluation using three datasets from amazon.com found the second algorithm superior to the other algorithms in detecting experts and deriving product ratings, significantly reducing the avg. error and avg. Mean Squared Error of the experts over the best of the other algorithms even after maintaining similar product coverage and quantity of reviews., Experiential product; Expert reviewer; Reviewer ranking; Product rating; Product ranking.

Beauty products are an important requirement for people, especially women. But, not all beauty products give the expected results. A review in the form of opinion can help the consumers to know the overview of the product. The reviews were analyzed using a multi-aspect-based approach to determine the aspects of the beauty category based on the reviews written on femaledaily.com. First, the review goes through the preprocessing stage to make it easier to be processed, and then it used the Support Vector Machine (SVM) method with the addition of Semantic Similarity and TF-IDF weighting. From the test result using semantic, get an accuracy of 93% on the price aspect, 92% on the packaging aspect, and 86% on the scent aspect.

Product reviews play a crucial role in online customer purchase decisions. In today’s marketplaces, such as Amazon, the quantity and interest of product reviews is growing, along with fierce competition between sellers. At the same time, the phenomenon of fake reviews is ever-growing. Many proposals capable of detecting fake reviews based on Machine Learning (ML) exist in literature. Nevertheless, bad practices implemented by the sellers make genuine reviews difficult to recognize. For instance, some Telegram Channels are known for giving products for free in exchange for 5-star reviews. This work focuses on the analysis of two review data streams of Amazon products. The first one is composed of the reviews corresponding to the products in the AmazonBasics category. The latter collects reviews of the products in the Telegram channels mentioned above. The analysis reveals a substantial dissimilarity between the two sources of reviews, that conducts to the construction of a ground truth dataset employed in the classification model training. The classification activity can assist during product rating interpretation, which could be invalidated by too many fake reviews. The experimental results reveal that 1&2-star reviews are good predictors of the review’s trustworthiness and the product itself.

Large number of product reviews may help the product looks more interesting, but it also may cause difficulties for customers in deciding to buy the product. The seller will also be facing difficulties such as to improve the quality of their product due to the huge number of reviews. The aim of this project is to develop a web-based system for customers and sellers to get an analysis of the online product reviews. The customer and seller will get the visualization for the positive, neutral, and negative distributions of the product reviews. The system uses MySQL as a database and Python as a programming language for development of the system. For the future work, the system could be able to compare the analysis of product reviews from other brands.

Users usually express their sentiment online which influences purchased products and services. The computational study of people's feelings and thoughts on entities is known as sentiment analysis. The Long Short-Term Memory (LSTM) model is one of the most common deep learning models for solving sentiment analysis problems. However, they possess some drawbacks such as longer training time, more memory for training, easily over fits, and sensitivity to randomly generated parameters. Hence, there is a need to optimize the LSTM parameters for enhanced sentiment analysis. This paper proposes an optimized LSTM approach using a newly developed novel Pastoralist Optimization Algorithm (POA) for enhanced sentiment analysis. The model was used to analyze sentiments of customers retrieved from Amazon product reviews. The performance of the developed POA-LSTM model shows an optimal accuracy, precision, recall and F1 measure of 77.36%, 85.06%, 76.29%, and 80.44% respectively, when compared with LSTM model with 71.62%, 78.26%, 74.23%, and 76.19% respectively. It was also observed that POA with 20 pastoralist population size performs better than other models with 10, 15, 25 and 30 population size.

Sentiment is central to many studies of communication science, from negativity and polarization in political communication to analyzing product reviews and social media comments in other sub-fields. This study provides an exhaustive comparison of sentiment analysis methods, using a validation set of Dutch economic headlines to compare the performance of manual annotation, crowd coding, numerous dictionaries and machine learning using both traditional and deep learning algorithms. The three main conclusions of this article are that: (1) The best performance is still attained with trained human or crowd coding; (2) None of the used dictionaries come close to acceptable levels of validity; and (3) machine learning, especially deep learning, substantially outperforms dictionary-based methods but falls short of human performance. From these findings, we stress the importance of always validating automatic text analysis methods before usage. Moreover, we provide a recommended step-by-step approach for (automated) text analysis projects to ensure both efficiency and validity.