To BERT or Not to BERT Dealing with Possible BERT Failures in an Entailment Task

Text classification algorithms investigate the intricate relationships between words or phrases and attempt to deduce the document's interpretation. In the last few years, these algorithms have progressed tremendously. Transformer architecture and sentence encoders have proven to give superior results on natural language processing tasks. But a major limitation of these architectures is their applicability for text no longer than a few hundred words. In this paper, we explore hierarchical transfer learning approaches for long document classification. We employ pre-trained Universal Sentence Encoder (USE) and Bidirectional Encoder Representations from Transformers (BERT) in a hierarchical setup to capture better representations efficiently. Our proposed models are conceptually simple where we divide the input data into chunks and then pass this through base models of BERT and USE. Then output representation for each chunk is then propagated through a shallow neural network comprising of LSTMs or CNNs for classifying the text data. These extensions are evaluated on 6 benchmark datasets. We show that USE + CNN/LSTM performs better than its stand-alone baseline. Whereas the BERT + CNN/LSTM performs on par with its stand-alone counterpart. However, the hierarchical BERT models are still desirable as it avoids the quadratic complexity of the attention mechanism in BERT. Along with the hierarchical approaches, this work also provides a comparison of different deep learning algorithms like USE, BERT, HAN, Longformer, and BigBird for long document classification. The Longformer approach consistently performs well on most of the datasets.

Fenomena bullying di media sosial, khususnya di Twitter, telah menjadi isu yang semakin memprihatinkan dengan dampak signifikan terhadap kesehatan mental pengguna. Dalam rangka mengatasi masalah ini, deteksi otomatis tweet yang mengandung konten bullying menjadi sangat penting. Penelitian ini bertujuan untuk membandingkan performa dua model pemrosesan bahasa alami terbaru, yaitu BERT (Bidirectional Encoder Representations from Transformers) dan XLNet, dalam klasifikasi tweet yang mengandung bullying. Metodologi penelitian ini melibatkan pengumpulan dataset tweet yang telah dilabeli sebagai bullying atau non-bullying. Proses preprocessing teks dilakukan untuk membersihkan dan menyiapkan data sebelum digunakan dalam pelatihan model. Kedua model, BERT dan XLNet, dilatih dan diuji menggunakan dataset yang sama. Evaluasi performa dilakukan dengan menggunakan metrik akurasi, presisi, recall, dan F1-score. Hasil penelitian menunjukkan bahwa kedua model memiliki kemampuan yang baik dalam mengidentifikasi tweet bullying, akan tetapi XLNet menunjukkan performa yang lebih unggul dibandingkan BERT dengan tingkat akurasi sebesar 95%. Dengan nilai presisi = 100%, recall = 0,87%, dan F1-score = 0,88%. XLNet mampu menangkap konteks dan nuansa bahasa yang lebih kompleks dalam tweet, yang berkontribusi pada akurasi klasifikasi yang lebih tinggi. Penelitian ini memberikan kontribusi penting dalam bidang deteksi bullying di media sosial dengan menunjukkan bahwa penggunaan model XLNet lebih efektif dibandingkan BERT. Temuan ini dapat membantu platform seperti Twitter dalam mengidentifikasi dan mencegah konten bullying, sehingga menciptakan lingkungan online yang lebih aman bagi pengguna, serta dapat digunakan sebagai dasar untuk pengembangan sistem deteksi bullying yang lebih canggih dan efisien di masa depan. Abstract The phenomenon of bullying on social media, particularly on Twitter, has become an increasingly concerning issue with significant impacts on users' mental health. In order to address this issue, automatic detection of tweets containing bullying content is crucial. This study aims to compare the performance of two recent natural language processing models, namely BERT (Bidirectional Encoder Representations from Transformers) and XLNet, in the classification of tweets containing bullying. The research methodology involves collecting a dataset of tweets that have been labelled as bullying or non-bullying. Text preprocessing is done to clean and prepare the data before it is used in model training. Both models, BERT and XLNet, were trained and tested using the same dataset. Performance evaluation was conducted using accuracy, precision, recall, and F1-score metrics. The results show that both models have a good ability to identify bullying tweets, but XLNet shows superior performance compared to BERT with an accuracy rate of 95%. With precision = 100%, recall = 0.87%, and F1-score = 0.88%. XLNet is able to capture more complex context and language nuances in tweets, which contributes to higher classification accuracy. This research makes an important contribution to the field of bullying detection on social media by showing that the use of the XLNet model is more effective than BERT. These findings can help platforms like Twitter identify and prevent bullying content, thereby creating a safer online environment for users, and can be used as a basis for the development of more sophisticated and efficient bullying detection systems in the future.

In this study, we explore the potential of machine learning (ML) models after different text representation methods on the balanced IMDB dataset, which is widely regarded as a gold standard in sentiment analysis, one of the Natural Language processing (NLP) tasks. On the open source IMDB movie reviews dataset, we first undertake data cleaning and text representation with data preprocessing steps. Then, we apply sentiment classification using different ML models. In order to evaluate the models, we used precision (P), recall (R), F1-score (F1), and area under curve (AUC), as well as receiver operating characteristic (ROC). It is worth noting that text feature extraction with Bidirectional Encoder Representations from Transformers (BERT) provided the highest performance in all models, with the SVM model offering particularly promising results. In this model, we observed the following results: ACC 0.9033, F1 0.9308, R 0.9015, R 0.9015, P 0.9072, AUC 0.9638, and ROC 0.96. These findings suggest that NLP techniques and, in particular, machine learning models that employ BERT may offer high levels of accuracy and reliability in text classification problems. It would be beneficial for future studies to validate these findings using BERT on different NLP tasks. This would help to evaluate the effectiveness and applicability of the models in practice.

Bidirectional Encoder Representations from Transformers (BERT) and BERT-based approaches are the current state-of-the-art in many natural language processing (NLP) tasks; however, their application to document classification on long clinical texts is limited. In this work, we introduce four methods to scale BERT, which by default can only handle input sequences up to approximately 400 words long, to perform document classification on clinical texts several thousand words long. We compare these methods against two much simpler architectures - a word-level convolutional neural network and a hierarchical self-attention network - and show that BERT often cannot beat these simpler baselines when classifying MIMIC-III discharge summaries and SEER cancer pathology reports. In our analysis, we show that two key components of BERT - pretraining and WordPiece tokenization - may actually be inhibiting BERT's performance on clinical text classification tasks where the input document is several thousand words long and where correctly identifying labels may depend more on identifying a few key words or phrases rather than understanding the contextual meaning of sequences of text.

We study the problem of incorporating prior knowledge into a deep Transformer-based model,i.e.,Bidirectional Encoder Representations from Transformers (BERT), to enhance its performance on semantic textual matching tasks. By probing and analyzing what BERT has already known when solving this task, we obtain better understanding of what task-specific knowledge BERT needs the most and where it is most needed. The analysis further motivates us to take a different approach than most existing works. Instead of using prior knowledge to create a new training task for fine-tuning BERT, we directly inject knowledge into BERT's multi-head attention mechanism. This leads us to a simple yet effective approach that enjoys fast training stage as it saves the model from training on additional data or tasks other than the main task. Extensive experiments demonstrate that the proposed knowledge-enhanced BERT is able to consistently improve semantic textual matching performance over the original BERT model, and the performance benefit is most salient when training data is scarce.

Relying on large pretrained language models such as Bidirectional Encoder Representations from Transformers (BERT) for encoding and adding a simple prediction layer has led to impressive performance in many clinical natural language processing (NLP) tasks. In this work, we present a novel extension to the Transformer architecture, by incorporating signature transform with the self-attention model. This architecture is added between embedding and prediction layers. Experiments on a new Swedish prescription data show the proposed architecture to be superior in two of the three information extraction tasks, comparing to baseline models. Finally, we evaluate two different embedding approaches between applying Multilingual BERT and translating the Swedish text to English then encode with a BERT model pretrained on clinical notes.

Natural Language Processing (NLP) is a pivotal area in artificial intelligence, aiming to make computers capable of understanding and generating human language. This study evaluates and compares three prominent NLP models—the Generative Pre-trained Transformer (GPT) model, Bidirectional Encoder Representations from Transformers (BERT) model, and Generalized Autoregressive Pretraining for Language Understanding (XLNet)—to determine their strengths, limitations, and suitability for various tasks. The research involves a comprehensive analysis of these models, utilizing well-established datasets such as the Stanford Question Answering Dataset (SQuAD), General Language Understanding Evaluation (GLUE), Reading Comprehension from Examinations (RACE), and the Situations with Adversarial Generations (SWAG). The study explores each model's architecture, pre-training, and fine-tuning processes: GPT’s unidirectional approach is assessed for its language generation and handling of long-range dependencies; Bidirectional encoding is examined for its effectiveness in context understanding, and XLNet permutation-based training is analyzed for its robust contextual comprehension. The experimental results reveal that GPT excels in generative tasks but is constrained by its unidirectional nature. BERT achieves superior accuracy in comprehension tasks but is computationally demanding and susceptible to pre-training bias. XLNet outperforms both GPT and BERT in accuracy and contextual understanding, though at the cost of increased complexity. The results offer a significant understanding of the effectiveness and applicability of these models, suggesting future research directions such as hybrid models and improvements in efficiency.

Recently, people used social media to convey their emotions in the text. Text classification is required for group a text containing emotion or not and identify the type of emotion. Most of the research in emotion classification use machine learning methods. So this research offers the use of deep learning for the classification of emotions in text. One of the algorithms that is widely used for deep learning is BERT (Bidirectional Encoder Representations from Transformers). For the experiment, classification is divided into two phases, namely the first phase groups emotion or not emotion tweets, and the second phase groups emotion tweets into five types of emotions, namely happy, angry, sad, scared and surprised. The experiment uses three data scenarios, namely the commuter line, transjakarta and combination data. The BERT method was compared with the BiLSTM and LSTM methods for emotion classification. The result of this study is BERT generates the highest accuracy compared to other methods for the three data scenarios with values of 87.24%, 86.27%, and 86.99% in the first phase. Then, the next accuracy value followed by obtained BiLSTM, and the last is LSTM. In the second phase, BERT also obtained the highest accuracy values for the three data scenarios with values of 94.26%, 93.71 %, and 93.98%. Then, the evaluation using Precision, Recall and F1-Measure results that BERT provides the best overall method. In this study, BERT has good performance for emotion classification.

Given the foundational role of system requirements in design projects, designers can benefit from classifying, comparing, and observing connections between requirements. Manually undertaking these processes, however, can be laborious and time-consuming. Previous studies have employed Bidirectional Encoder Representations from Transformers (BERT), a state-of-the-art natural language processing (NLP) deep neural network model, to automatically analyze written requirements. Yet, it remains unclear whether BERT can sufficiently capture the nuances that differentiate requirements between and within design documents. This work evaluates BERT’s performance on two requirement classification tasks (one inter- document and one intra-document) executed on a corpus of 1,303 requirements sourced from five system design projects. First, in the “parent document classification” task, a BERT model is fine-tuned to classify requirements according to their originating project. A separate BERT model is then fine-tuned on a “functional classification” task where each requirement is classified as either functional or nonfunctional. Our results also include a comparison with a baseline model, Word2Vec, and demonstrate that our model achieves higher classification accuracy. When evaluated on test sets, the former model receives a Matthews correlation coefficient (MCC) of 0.95, while the latter receives an MCC of 0.82, indicating BERT’s ability to reliably distinguish requirements. This work then explores the application of BERT’s representations, known as embeddings, to identify similar requirements and predict requirement change.

BackgroundThe accumulation of vast electronic medical records (EMRs) through medical informatization creates significant research value, particularly in obstetrics. Diagnostic standardization across different health care institutions and regions is vital for medical data analysis. Large language models (LLMs) have been extensively used for various medical tasks. Prompt engineering is key to use LLMs effectively.ObjectiveThis study aims to evaluate and compare the performance of LLMs with various prompt engineering techniques on the task of standardizing obstetric diagnostic terminology using real-world obstetric data.MethodsThe paper describes a 4-step approach used for mapping diagnoses in electronic medical records to the International Classification of Diseases, 10th revision, observation domain. First, similarity measures were used for mapping the diagnoses. Second, candidate mapping terms were collected based on similarity scores above a threshold, to be used as the training data set. For generating optimal mapping terms, we used two LLMs (ChatGLM2 and Qwen-14B-Chat [QWEN]) for zero-shot learning in step 3. Finally, a performance comparison was conducted by using 3 pretrained bidirectional encoder representations from transformers (BERTs), including BERT, whole word masking BERT, and momentum contrastive learning with BERT (MC-BERT), for unsupervised optimal mapping term generation in the fourth step.ResultsLLMs and BERT demonstrated comparable performance at their respective optimal levels. LLMs showed clear advantages in terms of performance and efficiency in unsupervised settings. Interestingly, the performance of the LLMs varied significantly across different prompt engineering setups. For instance, when applying the self-consistency approach in QWEN, the F1-score improved by 5%, with precision increasing by 7.9%, outperforming the zero-shot method. Likewise, ChatGLM2 delivered similar rates of accurately generated responses. During the analysis, the BERT series served as a comparative model with comparable results. Among the 3 models, MC-BERT demonstrated the highest level of performance. However, the differences among the versions of BERT in this study were relatively insignificant.ConclusionsAfter applying LLMs to standardize diagnoses and designing 4 different prompts, we compared the results to those generated by the BERT model. Our findings indicate that QWEN prompts largely outperformed the other prompts, with precision comparable to that of the BERT model. These results demonstrate the potential of unsupervised approaches in improving the efficiency of aligning diagnostic terms in daily research and uncovering hidden information values in patient data.

Bidirectional Encoder Representations from Transformers (BERT) models achieve state-of-the-art performance on a number of Natural Language Processing tasks. However, their model size on disk often exceeds 1 GB and the process of fine-tuning them and using them to run inference consumes significant hardware resources and runtime. This makes them hard to deploy to production environments. This paper fine-tunes DistilBERT, a lightweight deep learning model, on medical text for the named entity recognition task of Protected Health Information (PHI) and medical concepts. This work provides a full assessment of the performance of DistilBERT in comparison with BERT models that were pre-trained on medical text. For Named Entity Recognition task of PHI, DistilBERT achieved almost the same results as medical versions of BERT in terms of F1 score at almost half the runtime and consuming approximately half the disk space. On the other hand, for the detection of medical concepts, DistilBERT's F1 score was lower by 4 points on average than medical BERT variants.

With the popularity of large-scale corpora, statistics-based models have become mainstream model in Natural Language Processing (NLP). The Bidirectional Encoder Representations from Transformers (BERT), as one of those models, has achieved excellent results in various tasks of NLP since its emergence. But it still has shortcomings, such as poor capability of extracting local features and exploding of training gradients. After analyzing the shortcomings of BERT, this paper proposed an Information-gain Association Vector Semi-supervised Bidirectional Encoder Representations from Transformers (IAS-BERT) model, which improves the capability of capturing local features. Considering the influence of feature's polarity to overall sentiment and association between two word-embeddings, we use information gain on the training corpus. And then, the information gain results are used as an annotation of training corpus to generate a new word embedding. At the same time, we use forward-matching to optimize the computational overhead of IAS-BERT. We experiment the model on dataset of sentiment analysis, and it have achieved good results.

Sentence Boundary Disambiguation (SBD) is crucial for building datasets for tasks such as machine translation, syntactic analysis, and semantic analysis. Currently, most automatic sentence segmentation in Tibetan adopts the methods of rule-based and statistical learning, as well as the combination of the two, which have high requirements on the corpus and the linguistic foundation of the researchers and are more costly to annotate manually. In this study, we explore Tibetan SBD using deep learning technology. Initially, we analyze Tibetan characteristics and various subword techniques, selecting Byte Pair Encoding (BPE) and Sentencepiece (SP) for text segmentation and training the Bidirectional Encoder Representations from Transformers (BERT) pre-trained language model. Secondly, we studied the Tibetan SBD based on different BERT pre-trained language models, which mainly learns the ambiguity of the shad (“།”) in different positions in modern Tibetan texts and determines through the model whether the shad (“།”) in the texts has the function of segmenting sentences. Meanwhile, this study introduces four models, BERT-CNN, BERT-RNN, BERT-RCNN, and BERT-DPCNN, based on the BERT model for performance comparison. Finally, to verify the performance of the pre-trained language models on the SBD task, this study conducts SBD experiments on both the publicly available Tibetan pre-trained language model TiBERT and the multilingual pre-trained language model (Multi-BERT). The experimental results show that the F1 score of the BERT (BPE) model trained in this study reaches 95.32% on 465,669 Tibetan sentences, nearly five percentage points higher than BERT (SP) and Multi-BERT. The SBD method based on pre-trained language models in this study lays the foundation for establishing datasets for the later tasks of Tibetan pre-training, summary extraction, and machine translation.

PurposeSystems equipped with natural language (NLP) processing can reduce missed radiological findings by physicians, but the annotation costs are burden in the development. This study aimed to compare the effects of active learning (AL) algorithms in NLP for estimating the significance of head computed tomography (CT) reports using bidirectional encoder representations from transformers (BERT).MethodsA total of 3728 head CT reports annotated with five categories of importance were used and UTH-BERT was adopted as the pre-trained BERT model. We assumed that 64% (2385 reports) of the data were initially in the unlabeled data pool (UDP), while the labeled data set (LD) used to train the model was empty. Twenty-five reports were repeatedly selected from the UDP and added to the LD, based on seven metrices: random sampling (RS: control), four uncertainty sampling (US) methods (least confidence (LC), margin sampling (MS), ratio of confidence (RC), and entropy sampling (ES)), and two distance-based sampling (DS) methods (cosine distance (CD) and Euclidian distance (ED)). The transition of accuracy of the model was evaluated using the test dataset.ResultsThe accuracy of the models with US was significantly higher than RS when reports in LD were < 1800, whereas DS methods were significantly lower than RS. Among the US methods, MS and RC were even better than the others. With the US methods, the required labeled data decreased by 15.4–40.5%, and most efficient in RC. In addition, in the US methods, data for minor categories tended to be added to LD earlier than RS and DS.ConclusionsIn the classification task for the importance of head CT reports, US methods, especially RC and MS can lead to the effective fine-tuning of BERT models and reduce the imbalance of categories. AL can contribute to other studies on larger datasets by providing effective annotation.

Online medias are currently the dominant source of Information due to not being limited by time and place, fast and wide distributions. However, inaccurate news, or often referred as fake news is a major problem in news dissemination for online medias. Inaccurate news is information that is not true, that is engineered to cover the real information and has no factual basis. Usually, inaccurate news is made in the form of news that has mass appeal and is presented in the guise of genuine and legitimate news nuances to deceive or change the reader's mind or opinion. Identification of inaccurate news from real news can be done with natural language processing (NLP) technologies. In this paper, we proposed bidirectional encoder representations from transformers (BERT) for inaccurate news identification. BERT is a language model based on deep learning technologies and it has found effective for many NLP tasks. In this study, we use transfer learning and fine-tuning to adapt BERT for inaccurate news identification. The experiments show that our method could achieve accuracy of 99.23%, recall 99.46%, precision 98.86%, and F-Score of 99.15%. It is largely better than traditional method for the same tasks.