Adversarial Examples Against a BERT ABSA Model – Fooling Bert With L33T, Misspellign, and Punctuation,

Text classification is a basic task in natural language processing, but the small character perturbations in words can greatly decrease the effectiveness of text classification models, which is called character-level adversarial example attack. There are two main challenges in character-level adversarial examples defense, which are out-of-vocabulary words in word embedding model and the distribution difference between training and inference. Both of these two challenges make the character-level adversarial examples difficult to defend. In this paper, we propose a framework which jointly uses the character embedding and the adversarial stability training to overcome these two challenges. Our experimental results on five text classification data sets show that the models based on our framework can effectively defend character-level adversarial examples, and our models can defend 93.19% gradient-based adversarial examples and 94.83% natural adversarial examples, which outperforms the state-of-the-art defense models.

Deep learning based systems are susceptible to adversarial attacks, where a small, imperceptible change at the input alters the model prediction. However, to date the majority of the approaches to detect these attacks have been designed for image processing systems. Many popular image adversarial detection approaches are able to identify adversarial examples from embedding feature spaces, whilst in the NLP domain existing state of the art detection approaches solely focus on input text features, without consideration of model embedding spaces. This work examines what differences result when porting these image designed strategies to Natural Language Processing (NLP) tasks - these detectors are found to not port over well. This is expected as NLP systems have a very different form of input: discrete and sequential in nature, rather than the continuous and fixed size inputs for images. As an equivalent model-focused NLP detection approach, this work proposes a simple sentence-embedding "residue" based detector to identify adversarial examples. On many tasks, it out-performs ported image domain detectors and recent state of the art NLP specific detectors.

The advances in deep learning have led to great achievements in many Natural Language Processing (NLP) tasks. With the nature of language, i.e., sequential data, most NLP tasks can be framed into the sequence learning framework, such as text generation. As one of the most important foundations for modern NLP techniques, autoregressive generation models have achieved dominant performance in a great deal of NLP tasks. Therefore, this thesis emphasizes improving the autoregressive generation model for different NLP tasks. While many tasks can naturally fit into the sequence learning framework, some of them, e.g., building discourse parsing tree, require sophisticated designs to fit into neural models. Therefore, this thesis firstly emphasizes a novel unified framework for discourse parsing, which builds a discourse tree in a top-down depth-first manner, and it frames the task as an autoregressive generation task with the goal of each step being the prediction of the node position given a piece of text. The proposed approach is proven effective with extensive empirical experiments. In addition, I extend the above framework by proposing a hierarchical decoder, which leverages the information from parents and siblings of the nodes that are currently processed. The proposed decoder utilizes the nature of the tree structure and further improves the experiment performance on both discourse parsing and dependency parsing tasks. On the other hand, the de facto strategies, i.e., cross entropy loss and teacher forcing, for training the autoregressive generation models have been shown problematic in certain aspects. For example, cross entropy loss, which is one of the widely leveraged training objective functions, often leads to text degeneration in text generation, and teacher forcing suffers from the exposure bias problem, where there exists a mismatch between the training and testing setup. For text degeneration, I introduce a class of diminishing attentions, which enforces the submodularity of the coverage calculated by cross attention in the sequence-to-sequence model. The proposed diminishing attentions achieve notable improvement on several neural text generation tasks, including text summarization, machine translation, and image paragraph generation. Further, I propose a novel training objective, ScaleGrad, to replace cross entropy, which significantly reduces the degeneration problem in different text generation tasks. In fact, ScaleGrad can be extended to problems beyond text degeneration. It provides wide flexibility to inject different inductive biases into the text generation model by directly modifying the gradient information in the output layer. Next, for the exposure bias problem, this thesis introduces a novel type of scheduled sampling based on training accuracy, which requires only minimal hyper-parameter tuning compared to existing scheduled sampling methods. Additionally, a novel imitation loss is proposed to further enforce the model’s generative behavior to match the teacher-forced behavior. Moreover, this thesis demonstrates that reducing exposure bias can improve the robustness of language models against repetition and toxic errors.

According to data from the U.S. Center for Disease Control and Prevention, as of June 2020, a significant number of African Americans had been infected with the coronavirus disease, experiencing disproportionately higher death rates compared to other demographic groups. These disparities highlight the urgent need to examine the experiences, behaviors, and opinions of the African American population in relation to the COVID-19 pandemic. By understanding their unique challenges in navigating matters of health and well-being, we can work towards promoting health equity, eliminating disparities, and addressing persistent barriers to care. Since Twitter data has shown significant promise as a representation of human behavior and for opinion mining, this study leverages Twitter data published in 2020 to characterize the pandemic-related experiences of the United States’ African American population using aspect-based sentiment analysis. Sentiment analysis is a common task in natural language processing that identifies the emotional tone (i.e., positive, negative, or neutral) of a text sample. Aspect-based sentiment analysis increases the granularity of sentiment analysis by also extracting the aspect for which sentiment is expressed. We developed a machine learning pipeline consisting of image and language-based classification models to filter out tweets not related to COVID-19 and those unlikely published by African American Twitter subscribers, leading to an analysis of nearly 4 million tweets. Overall, our results show that the majority of tweets had a negative tone, and that the days with larger numbers of published tweets often coincided with major U.S. events related to the pandemic as suggested by major news headlines (e.g., vaccine rollout). We also show how word usage evolved throughout the year (e.g., outbreak to pandemic and coronavirus to covid). This work also points to important issues like food insecurity and vaccine hesitation, along with exposing semantic relationships between words, such as covid and exhausted. As such, this work furthers understanding of how the nationwide progression of the pandemic may have impacted the narratives of African American Twitter users.

Aspect-Based Sentiment Analysis (ABSA) is one of the highly challenging tasks in natural language processing. It extracts fine-grained sentiment information in user-generated reviews, as it aims at predicting the polarities towards predefined aspect categories or relevant entities in free text. Previous deep learning approaches usually rely on large-scale pre-trained language models and the attention mechanism, which applies the complete computed attention weights and does not place any restriction on the attention assignment. We argue that the original attention mechanism is not the ideal configuration for ABSA, as for most of the time only a small portion of terms are strongly related to the sentiment polarity of an aspect or entity. In this paper, we propose a masked attention mechanism customized for ABSA, with two different approaches to generate the mask. The first method sets an attention weight threshold that is determined by the maximum of all weights, and keeps only attention scores above the threshold. The second selects the top words with the highest weights. Both remove the lower score parts that are assumed to be less relevant to the aspect of focus. By ignoring part of input that is claimed irrelevant, a large proportion of input noise is removed, keeping the downstream model more focused and reducing calculation cost. Experiments on the Multi-Aspect Multi-Sentiment (MAMS) and SemEval-2014 datasets show significant improvements over state-of-the-art pre-trained language models with full attention, which displays the value of the masked attention mechanism. Recent work shows that simple self-attention in Transformer quickly degenerates to a rank-1 matrix, and masked attention may be another cure for that trend.

Sentiment analysis, a critical task in natural language processing, aims to automatically identify and classify the sentiment expressed in textual data. Aspect-level sentiment analysis focuses on determining sentiment at a more granular level, targeting specific aspects or features within a piece of text. In this paper, we explore various techniques for sentiment analysis, including traditional machine learning approaches and state-of-the-art deep learning models. Additionally, deep learning techniques has been utilized to identifying and extracting specific aspects from text, addressing aspect-level ambiguity, and capturing nuanced sentiments for each aspect. These datasets are valuable for conducting aspect-level sentiment analysis. In this article, we explore a language model based on pre-trained deep neural networks. This model can analyze sequences of text to classify sentiments as positive, negative, or neutral without explicit human labeling. To evaluate these models, data from Twitter's US airlines sentiment database was utilized. Experiments on this dataset reveal that the BERT, RoBERTA and DistilBERT model outperforms than the ML based model in accuracy and is more efficient in terms of training time. Notably, our findings showcase significant advancements over previous state-of-the-art methods that rely on supervised feature learning, bridging existing gaps in sentiment analysis methodologies. Our findings shed light on the advancements and challenges in sentiment analysis, offering insights for future research directions and practical applications in areas such as customer feedback analysis, social media monitoring, and opinion mining.

Aspect-Based Sentiment Analysis (ABSA) is a Natural Language Processing (NLP) task that extracts referred aspects from text and assigns polarities to opinions about those aspects. Most research on ABSA focuses on English. Only a few ABSA works deal with the Portuguese language. In this work, we used BERTimbau to create a Question-Answer approach to ABSA in Portuguese. First, we post-trained this model with text from the same domain as our target corpus. Then, we constructed an auxiliary sentence from the aspect and converted ABSA to a sentence-pair classification task, such as question answering (QA) and natural language inference (NLI). Our experiments show that ABSA based on BERT for Portuguese achieved Balanced Accuracy (BACC) of 77% on a corpus of reviews about the accommodation sector using a post-trained model with QA approach.

Sentiment analysis is becoming rapidly important for exploring social media Bangla text. The lack of sufficient resources is considered to be an important challenge for Aspect Based Sentiment Analysis (ABSA) of the Bangla language. The ABSA is a technique that divides the text and defines its sentiment based on its aspects. In this paper, we developed a high-quality Bangla ABSA annotated dataset namely BANGLA_ABSA. The datasets are labelled with aspects category and their respective sentiment polarity to do the ABSA task in Bangla. Four open domains namely Restaurant, Movie, Mobile phone, and Car are considered to make the dataset. The datasets are called Restaurant_ABSA, Movie_ABSA, Mobile_phone_ABSA, and Car_ABSA respectively that contain 801, 800, 975, and 1149 comments. All the comments are either complex or compound sentences. We created the dataset manually and annotated the same by exerting opinions. We organized the dataset as three tuples in Excel format namely 〈Id, Comment, {Aspect category, Sentiment Polarity}〉. These data are very important that facilitate the efficient handling of sentiment for any machine learning and deep learning research, especially for Bangla text.

Aspect Sentiment Triplet Extraction (ASTE) is a complex and challenging task in Natural Language Processing (NLP). It aims to extract the triplet of aspect term, opinion term, and their associated sentiment polarity, which is a more fine-grained study in Aspect Based Sentiment Analysis. Furthermore, there have been a large number of approaches being proposed to handle this relevant task. However, existing methods for ASTE suffer from powerless interactions between different sources of textual features, and they usually exert an equal impact on each type of feature, which is quite unreasonable while building contextual representation. Therefore, in this paper, we propose a novel Multi-Branch GCN (MBGCN)-based ASTE model to solve this problem. Specifically, our model first generates the enhanced semantic features via the structure-biased BERT, which takes the position of tokens into the transformation of self-attention. Then, a biaffine attention module is utilized to further obtain the specific semantic feature maps. In addition, to enhance the dependency among words in the sentence, four types of linguistic relations are defined, namely part-of-speech combination, syntactic dependency type, tree-based distance, and relative position distance of each word pair, which are further embedded as adjacent matrices. Then, the widely used Graph Convolutional Network (GCN) module is utilized to complete the work of integrating the semantic feature and linguistic feature, which is operated on four types of dependency relations repeatedly. Additionally, an effective refining strategy is employed to detect whether word pairs match or not, which is conducted after the operation of each branch GCN. At last, a shallow interaction layer is designed to achieve the final textual representation by fusing the four branch features with different weights. To validate the effectiveness of MBGCNs, extensive experiments have been conducted on four public and available datasets. Furthermore, the results demonstrate the effectiveness and robustness of MBGCNs, which obviously outperform state-of-the-art approaches.

In the ever-expanding digital landscape, the abundance of user-generated content on consumer platforms such as Booking and TripAdvisor offers a rich source of information for both travellers and hoteliers. Sentiment analysis, a fundamental research task of natural language processing (NLP) is used for mining sentiments and opinions within this vast reservoir of text reviews. A more specific type of sentiment analysis, i.e., aspect-based sentiment analysis (ABSA), is used when processing customer reviews is required. In ABSA, we aim to capture aspect-level sentiments and intricate relationships between various aspects within reviews. This article proposes a novel approach to ABSA by introducing a novel technique of word sense disambiguation (WSD) and integrating it with the Transformer architecture bidirectional encoder representations from Transformers (BERT) and graph convolutional networks (GCNs). The proposed approach resolves the intriguing ambiguities of the words and represents the review data as a complex graph structure, facilitating the modeling of intricate relationships between different aspects. The combination of bidirectional long short-term memory (BiLSTM) and GCN proves effective in capturing inter-dependencies among various aspects, providing a nuanced understanding of customer sentiments. The experiments are conducted on the RABSA dataset (an enhanced and richer hotel review data collection), and results demonstrate that our approach outperforms previous baselines, showcasing the effectiveness of integrating WSD in ABSA. Furthermore, an ablation study confirms the significant contribution of the WSD module to the overall performance. Moreover, we explore different similarity measures and find that cosine similarity yields the best results when identifying the real sense of a word in a given sentence using WordNet. The findings of our work and future work related to our work create lots of interest for people in the tourism and hospitality industry. This research gives another boost to the concept of the potential of NLP techniques in sentiment analysis. It emphasizes that if we combine the potential of NLP techniques along with state-of-the-art machine learning frameworks, we can shape the future of this field.

Text extraction is an important part of natural language processing (NLP) tasks. Most NLP tasks like text classification, machine translation, text-to-speech, text-based language identification, text summarization, and named-entity recognition involve the use of textual data. Such data is limited for low-resourced languages making it difficult to experiment advanced NLP techniques on these languages. This paper presents a Python-based toolkit for text analysis and text extraction from different types of images, documents, and audio files. The toolkit is built as a library that has functions that can be imported and utilized for text extraction.

Entity disambiguation is an important task in natural language processing that addresses situations where entity names mentioned in text may correspond to multiple real-world entities, with the goal of disambiguation by mapping entity names mentioned in text to unique and unambiguous entity identifiers. At present, most researches focus on entity disambiguation based on the concept level, which mainly uses contextual information to achieve entity disambiguation. However, the above methods still have some shortcomings: the word vector representation of the traditional model can not solve the polysemy problem, and the disambiguation terms can not consider the context globally to achieve entity disambiguation. In order to solve the above problems, this paper proposes a short text entity disambiguation method based on BERT model and shortest path algorithm. In this method, BiLSTM-CRF model is used in the ambiguous entity recognition stage, cosine similarity is used to perform optimal segmentation of short texts in the text segmentation stage, BERT pre-trained model is used to construct word vectors in the semantic vector construction stage, and the shortest path algorithm (SPA) is used to screen out the correct meaning of ambiguous entities in the current context in the entity disambiguation stage. The experimental results show that the three evaluation indicators of the proposed method, namely accuracy rate, recall rate and F1 score, have increased by an average of 23.37%, 26.47% and 24.98% respectively compared with the other three entity disambiguation methods. In conclusion, the proposed method can further improve the performance of entity disambiguation.

Many tasks in natural language processing like offensive, toxic, or emotional text classification are subjective by nature. Humans tend to perceive textual content in their own individual way. Existing methods commonly rely on the agreed output values, the same for all consumers. Here, we propose personalized solutions to subjective tasks. Our four new deep learning models take into account not only the content but also the specificity of a given human. The models represent different approaches to learning the representation and processing data about text readers. The experiments were carried out on four datasets: Wikipedia discussion texts labelled with attack, aggression, and toxicity, as well as opinions annotated with ten numerical emotional categories. Emotional data was considered as multivariate regression (multitask), whereas Wikipedia data as independent classifications. All our models based on human biases and their representations significantly improve the prediction quality in subjective tasks evaluated from the individual’s perspective.

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.

Long Short-Term Memory (LSTM) networks are unique to exercise data in its memory cell with long-term memory as Natural Language Processing (NLP) tasks have inklings of intensive time and computational power due to their complex structures like magnitude language model Transformer required to pre-train and learn billions of data performing different NLP tasks. In this paper, a dynamic chaotic model is proposed for the objective of transforming neurons states in network with neural dynamic characteristics by restructuring LSTM as Chaotic Neural Oscillatory-Long-Short Term Memory (CNO-LSTM), where neurons in LSTM memory cells are weighed in substitutes by oscillatory neurons to speed up computational training of language model and improve text classification accuracy for real-world applications. From the implementation perspective, five popular datasets of general text classification including binary, multi classification and multi-label classification are used to compare with mainstream baseline models on NLP tasks. Results showed that the performance of CNO-LSTM, a simplified model structure and oscillatory neurons state in exercising different types of text classification tasks are above baseline models in terms of evaluation index such as Accuracy, Precision, Recall and F1. The main contributions are time reduction and improved accuracy. It achieved approximately 46.76% of the highest reduction training time and 2.55% accuracy compared with vanilla LSTM model. Further, it achieved approximately 35.86% in time reduction compared with attention model without oscillatory indicating that the model restructure has reduced GPU dependency to improve training accuracy.