Search results

Corporate social responsibility (CSR) is significant in the financial market. Despite plenty of existing research on CSR, few studies have quantified the fine-grained aspects of CSR and examined how diverse CSR aspects are associated with firms' trade credit. Based on the released CSR reports, this paper strives to measure the CSR fulfillment of firms and examine the relationships between CSR and trade credit in terms of textual features presented in these reports.

This research proposes a natural language processing-based framework to extract the overall readability and the sentiment of fine-grained aspects from CSR reports, which can signal the performance of firms' CSR in diverse aspects. Furthermore, this paper explores how the textual features are associated with trade credit through partial dependence plots (PDPs), and PDPs can generate both linear and nonlinear relationships.

The study’s results reveal that the overall readability of the reports is positively associated with trade credit, while the performance of the fine-grained CSR aspects mentioned in the CSR reports matters differently. The performance of the environment has a positive impact on trade credit; the performance of creditors, suppliers and information disclosure, shows a U-shaped influence on trade credit; while the performance of the government and customers is negatively associated with trade credit.

This study expands the scope of research on CSR and trade credit by investigating fine-grained aspects covered in CSR reports. It also offers some managerial implications in the allocation of CSR resources and the presentation of CSR reports.

To understand the meaning of a sentence, humans can focus on important words in the sentence, which reflects our eyes staying on each word in different gaze time or times. Thus, some studies utilize eye-tracking values to optimize the attention mechanism in deep learning models. But these studies lack to explain the rationality of this approach. Whether the attention mechanism possesses this feature of human reading needs to be explored.

The authors conducted experiments on a sentiment classification task. Firstly, they obtained eye-tracking values from two open-source eye-tracking corpora to describe the feature of human reading. Then, the machine attention values of each sentence were learned from a sentiment classification model. Finally, a comparison was conducted to analyze machine attention values and eye-tracking values.

Through experiments, the authors found the attention mechanism can focus on important words, such as adjectives, adverbs and sentiment words, which are valuable for judging the sentiment of sentences on the sentiment classification task. It possesses the feature of human reading, focusing on important words in sentences when reading. Due to the insufficient learning of the attention mechanism, some words are wrongly focused. The eye-tracking values can help the attention mechanism correct this error and improve the model performance.

Our research not only provides a reasonable explanation for the study of using eye-tracking values to optimize the attention mechanism but also provides new inspiration for the interpretability of attention mechanism.

Reorganising unstructured academic abstracts according to a certain logical structure can help scholars not only extract valid information quickly but also facilitate the faceted search of academic literature. This study aims to build a high-performance model for identifying of the functional structures of unstructured abstracts in the social sciences.

This study first investigated the structuring of abstracts in academic articles in the field of social sciences, using large-scale statistical analyses. Then, the functional structures of sentences in the abstract in a corpus of more than 3.5 million abstracts were identified from sentence classification and sequence tagging by using several models based on either machine learning or a deep learning approach, and the results were compared.

The results demonstrate that the functional structures of sentences in abstracts in social science manuscripts include the background, purpose, methods, results and conclusions. The experimental results show that the bidirectional encoder representation from transformers exhibited the best performance, the overall F1 score of which was 86.23%.

The data set of annotated social science abstract is generated and corresponding models are trained on the basis of the data set, both of which are available on Github (https://github.com/Academic-Abstract-Knowledge-Mining/SSCI_Abstract_Structures_Identification). Based on the optimised model, a Web application for the identification of the functional structures of abstracts and their faceted search in social sciences was constructed to enable rapid and convenient reading, organisation and fine-grained retrieval of academic abstracts.

Questionnaire studies of knowledge, attitude and practice (KAP) are effective research in the field of health, which have many shortcomings. The purpose of this research is to propose an automatic questionnaire-free method based on deep learning techniques to address the shortcomings of common methods. Next, the aim of this research is to use the proposed method with public comments on Twitter to get the gaps in KAP of people regarding COVID-19.

In this paper, two models are proposed to achieve the mentioned purposes, the first one for attitude and the other for people’s knowledge and practice. First, the authors collect some tweets from Twitter and label them. After that, the authors preprocess the collected textual data. Then, the text representation vector for each tweet is extracted using BERT-BiGRU or XLNet-GRU. Finally, for the knowledge and practice problem, a multi-label classifier with 16 classes representing health guidelines is proposed. Also, for the attitude problem, a multi-class classifier with three classes (positive, negative and neutral) is proposed.

Labeling quality has a direct relationship with the performance of the final model, the authors calculated the inter-rater reliability using the Krippendorf alpha coefficient, which shows the reliability of the assessment in both problems. In the problem of knowledge and practice, 87% and in the problem of people’s attitude, 95% agreement was reached. The high agreement obtained indicates the reliability of the dataset and warrants the assessment. The proposed models in both problems were evaluated with some metrics, which shows that both proposed models perform better than the common methods. Our analyses for KAP are more efficient than questionnaire methods. Our method has solved many shortcomings of questionnaires, the most important of which is increasing the speed of evaluation, increasing the studied population and receiving reliable opinions to get accurate results.

Our research is based on social network datasets. This data cannot provide the possibility to discover the public information of users definitively. Addressing this limitation can have a lot of complexity and little certainty, so in this research, the authors presented our final analysis independent of the public information of users.

Combining recurrent neural networks with methods based on the attention mechanism improves the performance of the model and solves the need for large training data. Also, using these methods is effective in the process of improving the implementation of KAP research and eliminating its shortcomings. These results can be used in other text processing tasks and cause their improvement. The results of the analysis on the attitude, practice and knowledge of people regarding the health guidelines lead to the effective planning and implementation of health decisions and interventions and required training by health institutions. The results of this research show the effective relationship between attitude, practice and knowledge. People are better at following health guidelines than being aware of COVID-19. Despite many tensions during the epidemic, most people still discuss the issue with a positive attitude.

To the best of our knowledge, so far, no text processing-based method has been proposed to perform KAP research. Also, our method benefits from the most valuable data of today’s era (i.e. social networks), which is the expression of people’s experiences, facts and free opinions. Therefore, our final analysis provides more realistic results.

Sarcasm is a linguistic expression that usually carries the opposite meaning of what is being said by words, thus making it difficult for machines to discover the actual meaning. It is mainly distinguished by the inflection with which it is spoken, with an undercurrent of irony, and is largely dependent on context, which makes it a difficult task for computational analysis. Moreover, sarcasm expresses negative sentiments using positive words, allowing it to easily confuse sentiment analysis models. This paper aims to demonstrate the task of sarcasm detection using the approach of machine and deep learning.

For the purpose of sarcasm detection, machine and deep learning models were used on a data set consisting of 1.3 million social media comments, including both sarcastic and non-sarcastic comments. The data set was pre-processed using natural language processing methods, and additional features were extracted and analysed. Several machine learning models, including logistic regression, ridge regression, linear support vector and support vector machines, along with two deep learning models based on bidirectional long short-term memory and one bidirectional encoder representations from transformers (BERT)-based model, were implemented, evaluated and compared.

The performance of machine and deep learning models was compared in the task of sarcasm detection, and possible ways of improvement were discussed. Deep learning models showed more promise, performance-wise, for this type of task. Specifically, a state-of-the-art model in natural language processing, namely, BERT-based model, outperformed other machine and deep learning models.

This study compared the performance of the various machine and deep learning models in the task of sarcasm detection using the data set of 1.3 million comments from social media.

The purpose of this study is to investigate the effectiveness of qualitative information extracted from firm’s annual report in predicting corporate credit rating. Qualitative information represented by published reports or management interview has been known as an important source in addition to quantitative information represented by financial values in assigning corporate credit rating in practice. Nevertheless, prior studies have room for further research in that they rarely employed qualitative information in developing prediction model of corporate credit rating.

This study adopted three document vectorization methods, Bag-Of-Words (BOW), Word to Vector (Word2Vec) and Document to Vector (Doc2Vec), to transform an unstructured textual data into a numeric vector, so that Machine Learning (ML) algorithms accept it as an input. For the experiments, we used the corpus of Management’s Discussion and Analysis (MD&A) section in 10-K financial reports as well as financial variables and corporate credit rating data.

Experimental results from a series of multi-class classification experiments show the predictive models trained by both financial variables and vectors extracted from MD&A data outperform the benchmark models trained only by traditional financial variables.

This study proposed a new approach for corporate credit rating prediction by using qualitative information extracted from MD&A documents as an input to ML-based prediction models. Also, this research adopted and compared three textual vectorization methods in the domain of corporate credit rating prediction and showed that BOW mostly outperformed Word2Vec and Doc2Vec.

Digital platforms have grown significantly in recent years. Although high platform failure risks (PFR) have plagued the industry, the literature has only given this issue scant treatment. Customer sentiments are crucial for platforms and have a growing body of knowledge on its analysis. However, previous studies have overlooked rich contextual information emb`edded in user-generated content (UGC). Confronting the research gap of digital platform failure and drawbacks of customer sentiment analysis, we aim to detect signals of PFR based on our advanced customer sentiment analysis approach for UGC and to illustrate how customer sentiments could predict PFR.

We develop a deep-learning based approach to improve the accuracy of customer sentiment analysis for further predicting PFR. We leverage a unique dataset of online P2P lending, i.e., a typical setting of transactional digital platforms, including 97,876 pieces of UGC for 2,467 platforms from 2011 to 2018.

Our results show that the proposed approach can improve the accuracy of measuring customer sentiment by integrating word embedding technique and bidirectional long short-term memory (Bi-LSTM). On top of that, we show that customer sentiment can improve the accuracy for predicting PFR by 10.96%. Additionally, we do not only focus on a single type of customer sentiment in a static view. We discuss how the predictive power varies across positive, neutral, negative customer sentiments, and during different time periods.

Our research results contribute to the literature stream on digital platform failure with online information processing and offer implications for digital platform risk management with advanced customer sentiment analysis.

This paper aims to propose a novel approach which integrates collocations and domain concepts for Chinese cosmetic word of mouth (WOM) sentiment classification. Most sentiment analysis works by collecting sentiment scores from each unigram or bigram. However, not every unigram or bigram in a WOM document contains sentiments. Chinese collocations consist of the main sentiments of WOM. This paper reduces the complexity of the document dimensionality and makes an improvement for sentiment classification.

This paper builds two contextual lexicons for feature words and sentiment words, respectively. Based on these contextual lexicons, this paper uses the techniques of associated rules and mutual information to build possible Chinese collocation sets. This paper applies preference vector modelling as the vector representation approach to catch the relationship between Chinese collocations and their associated concepts.

This paper compares the proposed preference vector models with benchmarks, using three classification techniques (i.e. support vector machine, J48 decision tree and multilayer perceptron). According to the experimental results, the proposed models outperform all benchmarks evaluated by the criterion of accuracy.

This paper focuses on Chinese collocations and proposes a novel research approach for sentiment classification. The Chinese collocations used in this paper are adaptable to the content and domains. Finally, this paper integrates collocations with the preference vector modelling approach, which not only achieves a better sentiment classification performance for Chinese WOM documents but also avoids the curse of dimensionality.

Decision-making in human beings is affected by emotions and sentiments. The affective computing takes this into account, intending to tailor decision support to the emotional states of people. However, the representation and classification of emotions is a very challenging task. The study used customized methods of deep learning models to aid in the accurate classification of emotions and sentiments.

The present study presents affective computing model using both text and image data. The text-based affective computing was conducted on four standard datasets using three deep learning customized models, namely LSTM, GRU and CNN. The study used four variants of deep learning including the LSTM model, LSTM model with GloVe embeddings, Bi-directional LSTM model and LSTM model with attention layer.

The result suggests that the proposed method outperforms the earlier methods. For image-based affective computing, the data was extracted from Instagram, and Facial emotion recognition was carried out using three deep learning models, namely CNN, transfer learning with VGG-19 model and transfer learning with ResNet-18 model. The results suggest that the proposed methods for both text and image can be used for affective computing and aid in decision-making.

The study used deep learning for affective computing. Earlier studies have used machine learning algorithms for affective computing. However, the present study uses deep learning for affective computing.

The purpose of this paper is to develop a new method to extract medical temporal information from online health communities.

The authors trained a conditional random-filed model for the extraction of temporal expressions. The temporal relation identification is considered as a classification task and several support vector machine classifiers are built in the proposed method. For the model training, the authors extracted some high-level semantic features including co-reference relationship of medical concepts and the semantic similarity among words.

For the extraction of TIMEX, the authors find that well-formatted expressions are easy to recognize, and the main challenge is the relative TIMEX such as “three days after onset”. It also shows the same difficulty for normalization of absolute date or well-formatted duration, whereas frequency is easier to be normalized. For the identification of DocTimeRel, the result is fairly well, and the relation is difficult to identify when it involves a relative TIMEX or a hypothetical concept.

The authors proposed a new method to extract temporal information from the online clinical data and evaluated the usefulness of different level of syntactic features in this task.