This paper presents a comprehensive overview of the Comparative Opinion Mining from Vietnamese Product Reviews shared task (ComOM), held as part of the 10$^{th}$ International Workshop on Vietnamese Language and Speech Processing (VLSP 2023). The primary objective of this shared task is to advance the field of natural language processing by developing techniques that proficiently extract comparative opinions from Vietnamese product reviews. Participants are challenged to propose models that adeptly extract a comparative "quintuple" from a comparative sentence, encompassing Subject, Object, Aspect, Predicate, and Comparison Type Label. We construct a human-annotated dataset comprising $120$ documents, encompassing $7427$ non-comparative sentences and $2468$ comparisons within $1798$ sentences. Participating models undergo evaluation and ranking based on the Exact match macro-averaged quintuple F1 score.

Sentiment analysis (SA) has been, and is still, a thriving research area. However, the task of Arabic sentiment analysis (ASA) is still underrepresented in the body of research. This study offers the first in-depth and in-breadth analysis of existing ASA studies of textual content and identifies their common themes, domains of application, methods, approaches, technologies and algorithms used. The in-depth study manually analyses 133 ASA papers published in the English language between 2002 and 2020 from four academic databases (SAGE, IEEE, Springer, WILEY) and from Google Scholar. The in-breadth study uses modern, automatic machine learning techniques, such as topic modelling and temporal analysis, on Open Access resources, to reinforce themes and trends identified by the prior study, on 2297 ASA publications between 2010-2020. The main findings show the different approaches used for ASA: machine learning, lexicon-based and hybrid approaches. Other findings include ASA 'winning' algorithms (SVM, NB, hybrid methods). Deep learning methods, such as LSTM can provide higher accuracy, but for ASA sometimes the corpora are not large enough to support them. Additionally, whilst there are some ASA corpora and lexicons, more are required. Specifically, Arabic tweets corpora and datasets are currently only moderately sized. Moreover, Arabic lexicons that have high coverage contain only Modern Standard Arabic (MSA) words, and those with Arabic dialects are quite small. Thus, new corpora need to be created. On the other hand, ASA tools are stringently lacking. There is a need to develop ASA tools that can be used in industry, as well as in academia, for Arabic text SA. Hence, our study offers insights into the challenges associated with ASA research and provides suggestions for ways to move the field forward such as lack of Dialectical Arabic resource, Arabic tweets, corpora and data sets for SA.

In today's digital world, social media plays a significant role in facilitating communication and content sharing. However, the exponential rise in user-generated content has led to challenges in maintaining a respectful online environment. In some cases, users have taken advantage of anonymity in order to use harmful language, which can negatively affect the user experience and pose serious social problems. Recognizing the limitations of manual moderation, automatic detection systems have been developed to tackle this problem. Nevertheless, several obstacles persist, including the absence of a universal definition for harmful language, inadequate datasets across languages, the need for detailed annotation guideline, and most importantly, a comprehensive framework. This study aims to address these challenges by introducing, for the first time, a detailed framework adaptable to any language. This framework encompasses various aspects of harmful language detection. A key component of the framework is the development of a general and detailed annotation guideline. Additionally, the integration of sentiment analysis represents a novel approach to enhancing harmful language detection. Also, a definition of harmful language based on the review of different related concepts is presented. To demonstrate the effectiveness of the proposed framework, its implementation in a challenging low-resource language is conducted. We collected a Persian dataset and applied the annotation guideline for harmful detection and sentiment analysis. Next, we present baseline experiments utilizing machine and deep learning methods to set benchmarks. Results prove the framework's high performance, achieving an accuracy of 99.4% in offensive language detection and 66.2% in sentiment analysis.

Information extraction (IE) aims to extract complex structured information from the text. Numerous datasets have been constructed for various IE tasks, leading to time-consuming and labor-intensive data annotations. Nevertheless, most prevailing methods focus on training task-specific models, while the common knowledge among different IE tasks is not explicitly modeled. Moreover, the same phrase may have inconsistent labels in different tasks, which poses a big challenge for knowledge transfer using a unified model. In this study, we propose a regularization-based transfer learning method for IE (TIE) via an instructed graph decoder. Specifically, we first construct an instruction pool for datasets from all well-known IE tasks, and then present an instructed graph decoder, which decodes various complex structures into a graph uniformly based on corresponding instructions. In this way, the common knowledge shared with existing datasets can be learned and transferred to a new dataset with new labels. Furthermore, to alleviate the label inconsistency problem among various IE tasks, we introduce a task-specific regularization strategy, which does not update the gradients of two tasks with'opposite direction'. We conduct extensive experiments on 12 datasets spanning four IE tasks, and the results demonstrate the great advantages of our proposed method.

It is important to be able to analyze the emotional state of people around the globe. There are 7100+ active languages spoken around the world and building emotion classification for each language is labor intensive. Particularly for low-resource and endangered languages, building emotion classification can be quite challenging. We present a cross-lingual emotion classifier, where we train an emotion classifier with resource-rich languages (i.e. \textit{English} in our work) and transfer the learning to low and moderate resource languages. We compare and contrast two approaches of transfer learning from a high-resource language to a low or moderate-resource language. One approach projects the annotation from a high-resource language to low and moderate-resource language in parallel corpora and the other one uses direct transfer from high-resource language to the other languages. We show the efficacy of our approaches on 6 languages: Farsi, Arabic, Spanish, Ilocano, Odia, and Azerbaijani. Our results indicate that our approaches outperform random baselines and transfer emotions across languages successfully. For all languages, the direct cross-lingual transfer of emotion yields better results. We also create annotated emotion-labeled resources for four languages: Farsi, Azerbaijani, Ilocano and Odia.

Gradient-based explanation methods are increasingly used to interpret neural models in natural language processing (NLP) due to their high fidelity. Such methods determine word-level importance using dimension-level gradient values through a norm function, often presuming equal significance for all gradient dimensions. However, in the context of Aspect-based Sentiment Analysis (ABSA), our preliminary research suggests that only specific dimensions are pertinent. To address this, we propose the Information Bottleneck-based Gradient (\texttt{IBG}) explanation framework for ABSA. This framework leverages an information bottleneck to refine word embeddings into a concise intrinsic dimension, maintaining essential features and omitting unrelated information. Comprehensive tests show that our \texttt{IBG} approach considerably improves both the models' performance and interpretability by identifying sentiment-aware features.

In this paper we explore the challenges of measuring sentiment in relation to Environmental, Social and Governance (ESG) social media. ESG has grown in importance in recent years with a surge in interest from the financial sector and the performance of many businesses has become based in part on their ESG related reputations. The use of sentiment analysis to measure ESG related reputation has developed and with it interest in the use of machines to do so. The era of digital media has created an explosion of new media sources, driven by the growth of social media platforms. This growing data environment has become an excellent source for behavioural insight studies across many disciplines that includes politics, healthcare and market research. Our study seeks to compare human performance with the cutting edge in machine performance in the measurement of ESG related sentiment. To this end researchers classify the sentiment of 150 tweets and a reliability measure is made. A gold standard data set is then established based on the consensus of 3 researchers and this data set is then used to measure the performance of different machine approaches: one based on the VADER dictionary approach to sentiment classification and then multiple language model approaches, including Llama2, T5, Mistral, Mixtral, FINBERT, GPT3.5 and GPT4.

In recent years, transformer models have revolutionized Natural Language Processing (NLP), achieving exceptional results across various tasks, including Sentiment Analysis (SA). As such, current state-of-the-art approaches for SA predominantly rely on transformer models alone, achieving impressive accuracy levels on benchmark datasets. In this paper, we show that the key for further improving the accuracy of such ensembles for SA is to include not only transformers, but also traditional NLP models, despite the inferiority of the latter compared to transformer models. However, as we empirically show, this necessitates a change in how the ensemble is constructed, specifically relying on the Hierarchical Ensemble Construction (HEC) algorithm we present. Our empirical studies across eight canonical SA datasets reveal that ensembles incorporating a mix of model types, structured via HEC, significantly outperform traditional ensembles. Finally, we provide a comparative analysis of the performance of the HEC and GPT-4, demonstrating that while GPT-4 closely approaches state-of-the-art SA methods, it remains outperformed by our proposed ensemble strategy.

Aspect-Based Sentiment Analysis (ABSA) stands as a crucial task in predicting the sentiment polarity associated with identified aspects within text. However, a notable challenge in ABSA lies in precisely determining the aspects' boundaries (start and end indices), especially for long ones, due to users' colloquial expressions. We propose DiffusionABSA, a novel diffusion model tailored for ABSA, which extracts the aspects progressively step by step. Particularly, DiffusionABSA gradually adds noise to the aspect terms in the training process, subsequently learning a denoising process that progressively restores these terms in a reverse manner. To estimate the boundaries, we design a denoising neural network enhanced by a syntax-aware temporal attention mechanism to chronologically capture the interplay between aspects and surrounding text. Empirical evaluations conducted on eight benchmark datasets underscore the compelling advantages offered by DiffusionABSA when compared against robust baseline models.

The effectiveness of various pretrained language models, including BERT [Devlin et al., 2018], XLNet [Yang et al., 2019], BART [Lewis et al., 2020], and GPT-3, in sentiment classification, a significant downstream task, has been extensively studied. Current research in sentiment classification often focuses on identifying sentiment polarities at the aspect level, leading to the emergence of aspect-based sentiment classification (ABSC). Many studies have achieved impressive results and introduced innovative approaches to tackle the ABSC task. For instance, Sun et al. [2019] utilized BERT to transform ABSC tasks into sentence-pair classification, which has influenced subsequent methodologies [Hu et al., 2022]. Additionally, generative models like BART [Lewis et al., 2020] have been employed by Yan et al. [2021] to convert ABSC tasks into sequence-to-sequence tasks, enabling the prediction of token sequences representing identified aspects and associated sentiments. Furthermore, Li et al. [2021a] reframed ABSC tasks as masked language modeling tasks, effectively bridging the performance gap between pre-training and ABSC tasks. Despite numerous attempts to address aspect-level sentiment classification, the primary focus has been on improving aspect-level sentiment polarity performance through specialized datasets and training methodologies. However, it is equally crucial for models to predict not only the in-context polarity of aspects but also their aspect polarity.