Search results

The purpose of this paper is to describe a new approach to sentence representation learning leading to text classification using Bidirectional Encoder Representations from Transformers (BERT) embeddings. This work proposes a novel BERT-convolutional neural network (CNN)-based model for sentence representation learning and text classification. The proposed model can be used by industries that work in the area of classification of similarity scores between the texts and sentiments and opinion analysis.

The approach developed is based on the use of the BERT model to provide distinct features from its transformer encoder layers to the CNNs to achieve multi-layer feature fusion. To achieve multi-layer feature fusion, the distinct feature vectors of the last three layers of the BERT are passed to three separate CNN layers to generate a rich feature representation that can be used for extracting the keywords in the sentences. For sentence representation learning and text classification, the proposed model is trained and tested on the Stanford Sentiment Treebank-2 (SST-2) data set for sentiment analysis and the Quora Question Pair (QQP) data set for sentence classification. To obtain benchmark results, a selective training approach has been applied with the proposed model.

On the SST-2 data set, the proposed model achieved an accuracy of 92.90%, whereas, on the QQP data set, it achieved an accuracy of 91.51%. For other evaluation metrics such as precision, recall and F1 Score, the results obtained are overwhelming. The results with the proposed model are 1.17%–1.2% better as compared to the original BERT model on the SST-2 and QQP data sets.

The novelty of the proposed model lies in the multi-layer feature fusion between the last three layers of the BERT model with CNN layers and the selective training approach based on gated pruning to achieve benchmark results.

The purpose of this study is to examine public perceptions and attitudes toward using artificial intelligence (AI) in the US criminal justice system.

The authors took a quantitative approach and administered an online survey using the Amazon Mechanical Turk platform. The instrument was developed by integrating prior literature to create multiple scales for measuring public perceptions and attitudes.

The findings suggest that despite the various attempts, there are still significant perceptions of sociodemographic bias in the criminal justice system and technology alone cannot alleviate them. However, AI can assist judges in making fairer and more objective decisions by using triangulation – offering additional data points to offset individual biases.

Other scholars can build upon the findings and extend the work to shed more light on some problems of growing concern for society – bias and inequality in criminal sentencing. AI can be a valuable tool to assist judges in the decision-making process by offering diverse viewpoints. Furthermore, the authors bridge the gap between the fields of technology and criminal justice and demonstrate how the two can be successfully integrated for the benefit of society.

To the best of the authors’ knowledge, this is among the first studies to examine a complex societal problem like the introduction of technology in a high-stakes environment – the US criminal justice system. Understanding how AI is perceived by society is necessary to develop more transparent and unbiased algorithms for assisting judges in making fair and equitable sentencing decisions. In addition, the authors developed and validated a new scale that can be used to further examine this novel approach to criminal sentencing in the future.

Question answering (QA) answers the questions asked by people in the form of natural language. In the QA, due to the subjectivity of users, the questions they query have different expressions, which increases the difficulty of text retrieval. Therefore, the purpose of this paper is to explore new query rewriting method for QA that integrates multiple related questions (RQs) to form an optimal question. Moreover, it is important to generate a new dataset of the original query (OQ) with multiple RQs.

This study collects a new dataset SQuAD_extend by crawling the QA community and uses word-graph to model the collected OQs. Next, Beam search finds the best path to get the best question. To deeply represent the features of the question, pretrained model BERT is used to model sentences.

The experimental results show three outstanding findings. (1) The quality of the answers is better after adding the RQs of the OQs. (2) The word-graph that is used to model the problem and choose the optimal path is conducive to finding the best question. (3) Finally, BERT can deeply characterize the semantics of the exact problem.

The proposed method can use word-graph to construct multiple questions and select the optimal path for rewriting the question, and the quality of answers is better than the baseline. In practice, the research results can help guide users to clarify their query intentions and finally achieve the best answer.

Drug sales facilitated through digital communication on the surface web and on darknet cryptomarkets have increased during the past two decades. This has resulted in an increase in drug law enforcement efforts to combat these markets and a subsequent increase in judicial sentencing of people selling drugs online. The aim of this study was to analyze how Swedish courts describe sentenced sellers and how the courts apply case law.

The empirical material consists of 71 sentencing documents produced by Swedish courts in cases of online drug selling between January 1, 2010 and January 1, 2020. In total, 99 sentenced persons occur in the documents. Using a qualitative research design, the authors analyzed the material through thematic text analysis.

Overall, in their descriptions of online drug sale operations, the courts’ characterizations of the concepts of street capital and digital capital show a dichotomy. These forms of capital are situationally described as both aggravating and mitigating aspects in the application of case law, indicating that it may be fruitful to view both street and digital capital as resources used on contemporary drug markets in general.

Very little research exists into how judicial systems describe and perceive the developing phenomenon of online drug sales. Using a relatively large sample from a decade of sentencing, the authors provide an analysis of how Swedish courts view and valuate capital forms in the online drugs trade.

Whether automatically generated summaries of health social media can assist users in appropriately managing their diseases and ensuring better communication with health professionals becomes an important issue. This paper aims to develop a novel deep learning-based summarization approach for obtaining the most informative summaries from online patient reviews accurately and effectively.

This paper proposes a framework to generate summaries that integrates a domain-specific pre-trained embedding model and a deep neural extractive summary approach by considering content features, text sentiment, review influence and readability features. Representative health-related summaries were identified, and user judgements were analysed.

Experimental results on the three real-world health forum data sets indicate that awarding sentences without incorporating all the adopted features leads to declining summarization performance. The proposed summarizer significantly outperformed the comparison baseline. User judgement through the questionnaire provides realistic and concrete evidence of crucial features that remarkably influence patient forum review summaries.

This study contributes to health analytics and management literature by exploring users’ expressions and opinions through the health deep learning summarization model. The research also developed an innovative mindset to design summarization weighting methods from user-created content on health topics.

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.

According to the Indian Sign Language Research and Training Centre (ISLRTC), India has approximately 300 certified human interpreters to help people with hearing loss. This paper aims to address the issue of Indian Sign Language (ISL) sentence recognition and translation into semantically equivalent English text in a signer-independent mode.

This study presents an approach that translates ISL sentences into English text using the MobileNetV2 model and Neural Machine Translation (NMT). The authors have created an ISL corpus from the Brown corpus using ISL grammar rules to perform machine translation. The authors’ approach converts ISL videos of the newly created dataset into ISL gloss sequences using the MobileNetV2 model and the recognized ISL gloss sequence is then fed to a machine translation module that generates an English sentence for each ISL sentence.

As per the experimental results, pretrained MobileNetV2 model was proven the best-suited model for the recognition of ISL sentences and NMT provided better results than Statistical Machine Translation (SMT) to convert ISL text into English text. The automatic and human evaluation of the proposed approach yielded accuracies of 83.3 and 86.1%, respectively.

It can be seen that the neural machine translation systems produced translations with repetitions of other translated words, strange translations when the total number of words per sentence is increased and one or more unexpected terms that had no relation to the source text on occasion. The most common type of error is the mistranslation of places, numbers and dates. Although this has little effect on the overall structure of the translated sentence, it indicates that the embedding learned for these few words could be improved.

Sign language recognition and translation is a crucial step toward improving communication between the deaf and the rest of society. Because of the shortage of human interpreters, an alternative approach is desired to help people achieve smooth communication with the Deaf. To motivate research in this field, the authors generated an ISL corpus of 13,720 sentences and a video dataset of 47,880 ISL videos. As there is no public dataset available for ISl videos incorporating signs released by ISLRTC, the authors created a new video dataset and ISL corpus.

Over a decade since the Special Criminal Court (SCC) was established in Cameroon, hundreds of individuals have been indicted, tried and convicted. Sentences have been imposed, most of which include a term of imprisonment (principal punishment/penalty) and confiscation as accessory penalty or punishment. Research focus has not been directed at the sentences which, as argued in this paper, are inconsistent, incommensurate with the amounts of money stolen and a significant departure from the Penal Code. This paper aims to explore the aspect of sentencing by the SCC.

To identify, highlight and discuss the issue of sentencing, the paper looks at a blend of primary and secondary materials: primary materials here include but not limited to the judgements of the SCC and other courts in Cameroon and the Penal Code. Secondary materials shall include the works of scholars in the fields of criminal law, criminal justice and penal reform.

A few findings were made: first, the judges are inconsistent in the manner in which they determine the appropriate sentence. Second, in making that determination, the judges would have been oblivious to the prescripts in the Penal Code, which provides the term of imprisonment, and in the event of a mitigating circumstance, the prescribed minimum to be applied. Yet, the default imposition of an aggravating circumstance (being a civil servant) was not explored by the SCC. Finally, whether the sentences imposed are commensurate with the amounts of monies stolen.

This research unravels key insights into the functioning of the SCC. It advances the knowledge thereon and adds to the literature on corruption in Cameroon.

The prosecution and judges at the SCC should deepen their knowledge of Cameroonian criminal law, especially on the nature of liberty given to judges to determine within the prescribed range of the sentence to be imposed but also consider the existence of an aggravating factor – civil servant. They must also consider whether the sentences imposed befit the crime for which they are convicted.

The paper is an original contribution with new insights on the manner in which sentencing should be approached by the SCC.

The purpose of this paper is to assess if federal judges have sentenced criminal corporations to fines that are consistent with the seriousness of the offense and the blameworthiness of the organization, which would be in line with the directives from the US Sentencing Guidelines. This paper will also use the focal concerns framework to measure organizational blameworthiness.

This paper uses secondary data from federal sentencing documents, collected by the US Sentencing Commission, for cases that were adjudicated between October 1, 2010 and September 30, 2017.

Results showed that the focal concerns framework can be used to define potential constructs for blameworthiness and that an organization’s culpability score was a significant predictor in whether the company received a higher fine.

The data are unable to examine two of the three measures of focal concerns. Cross-sectional data limits the ability to draw conclusions regarding cause and effect between blameworthiness and monetary fines.

Results imply that judges are sentencing corporations that have higher culpability scores to more severe fines, in accordance with both the federal Sentencing Guidelines and focal concerns framework.

This study is one of the first to apply the focal concerns framework, usually used to examine the sentencing of individuals, to the sentencing of corporations. It is also one of the first to attempt to empirically define blameworthiness.

Customer reviews are one important source that contains valuable information for quality evaluation of products or services. Review sentences contain sentiment words that show whether a user’s opinion is positive or negative. When review sentence has mix opinions, having sentiment words of both polarities, it is difficult to conclude whether it is positive or negative opinion. The purpose of this study is to improve the detection of polarity in such situation.

In this research, methods such as part-of-speech tagging, polarity analysis and rules selection are used to identify the polarity. A set of rules called contrast and conditional polarity rules (CCPR) has been created to improve the polarity detection in cases when there is mixture of sentiment words used in contrast and conditional type of review sentences. The experiment is conducted with data sets from three domains, i.e. restaurant, electronic and Tripadvisor.

The experimental result confirms that CCPR rules have higher baseline of the polarity aggression. In restaurant domain, CCPR rules (62.07%) have increased 13.79% compared with the Pol_Agg_MPQA baseline (48.28%) and 13.79% compared with Pol_Agg_Senti baseline (48.28%). In electronic domain, CCPR rule (79.17%) is higher by 12.50% compared with the Pol_Agg_MPQA baseline (66.67%) and 16.67% compared with Pol_Agg_Senti baseline (62.50%). Another one, CCPR rule (70.83%) is higher by 8.33% compared with the Pol_Agg_MPQA baseline (62.50%) and 12.50% compared with Pol_Agg_Senti baseline (58.33%). In conclusion, result of experiment shows promising outcome with improvement in detecting the positivity and negativity of indirect sentence, especially for the case of sentence with indirect polarity.

To address the problem of mix opinions in terms of polarities, this paper presents a rule-based approach to improve the result of identifying positivity and negativity in sentence with indirect polarities.