Aspect-Based Sentiment Analysis (ABSA) offers granular insights into opinions but often suffers from the scarcity of diverse, labeled datasets that reflect real-world conversational nuances. This paper presents an approach for generating synthetic ABSA data using Large Language Models (LLMs) to address this gap. We detail the generation process aimed at producing data with consistent topic and sentiment distributions across multiple domains using GPT-4o. The quality and utility of the generated data were evaluated by assessing the performance of three state-of-the-art LLMs (Gemini 1.5 Pro, Claude 3.5 Sonnet, and DeepSeek-R1) on topic and sentiment classification tasks. Our results demonstrate the effectiveness of the synthetic data, revealing distinct performance trade-offs among the models: DeepSeek-R1 showed higher precision, Gemini 1.5 Pro and Claude 3.5 Sonnet exhibited strong recall, and Gemini 1.5 Pro offered significantly faster inference. We conclude that LLM-based synthetic data generation is a viable and flexible method for creating valuable ABSA resources, facilitating research and model evaluation without reliance on limited or inaccessible real-world labeled data.

This paper presents a compact model that achieves strong sentiment analysis capabilities through targeted distillation from advanced large language models (LLMs). Our methodology decouples the distillation target into two key components: sentiment-related knowledge and task alignment. To transfer these components, we propose a two-stage distillation framework. The first stage, knowledge-driven distillation (\textsc{KnowDist}), transfers sentiment-related knowledge to enhance fundamental sentiment analysis capabilities. The second stage, in-context learning distillation (\textsc{ICLDist}), transfers task-specific prompt-following abilities to optimize task alignment. For evaluation, we introduce \textsc{SentiBench}, a comprehensive sentiment analysis benchmark comprising 3 task categories across 12 datasets. Experiments on this benchmark demonstrate that our model effectively balances model size and performance, showing strong competitiveness compared to existing small-scale LLMs.

Aspect-based sentiment analysis (ABSA), a fine-grained sentiment classification task, has received much attention recently. Many works investigate sentiment information through opinion words, such as ''good'' and ''bad''. However, implicit sentiment widely exists in the ABSA dataset, which refers to the sentence containing no distinct opinion words but still expresses sentiment to the aspect term. To deal with implicit sentiment, this paper proposes an ABSA method that integrates explicit sentiment augmentations. And we propose an ABSA-specific augmentation method to create such augmentations. Specifically, we post-trains T5 by rule-based data. We employ Syntax Distance Weighting and Unlikelihood Contrastive Regularization in the training procedure to guide the model to generate an explicit sentiment. Meanwhile, we utilize the Constrained Beam Search to ensure the augmentation sentence contains the aspect terms. We test ABSA-ESA on two of the most popular benchmarks of ABSA. The results show that ABSA-ESA outperforms the SOTA baselines on implicit and explicit sentiment accuracy.

Dialogue understanding has always been a bottleneck for many conversational tasks, such as dialogue response generation and conversational question answering. To expedite the progress in this area, we introduce the task of conversational aspect sentiment analysis (CASA) that can provide useful fine-grained sentiment information for dialogue understanding and planning. Overall, this task extends the standard aspect-based sentiment analysis to the conversational scenario with several major adaptations. To aid the training and evaluation of data-driven methods, we annotate 3,000 chit-chat dialogues (27,198 sentences) with fine-grained sentiment information, including all sentiment expressions, their polarities and the corresponding target mentions. We also annotate an out-of-domain test set of 200 dialogues for robustness evaluation. Besides, we develop multiple baselines based on either pretrained BERT or self-attention for preliminary study. Experimental results show that our BERT-based model has strong performances for both in-domain and out-of-domain datasets, and thorough analysis indicates several potential directions for further improvements.

Aspect-based sentiment analysis aims to identify the sentiment polarity of a specific aspect in product reviews. We notice that about 30% of reviews do not contain obvious opinion words, but still convey clear human-aware sentiment orientation, which is known as implicit sentiment. However, recent neural network-based approaches paid little attention to implicit sentiment entailed in the reviews. To overcome this issue, we adopt Supervised Contrastive Pre-training on large-scale sentiment-annotated corpora retrieved from in-domain language resources. By aligning the representation of implicit sentiment expressions to those with the same sentiment label, the pre-training process leads to better capture of both implicit and explicit sentiment orientation towards aspects in reviews. Experimental results show that our method achieves state-of-the-art performance on SemEval2014 benchmarks, and comprehensive analysis validates its effectiveness on learning implicit sentiment.

The problem of automatic extraction of sentiment expressions from informal text, as in microblogs such as tweets is a recent area of investigation. Compared to formal text, such as in product reviews or news articles, one of the key challenges lies in the wide diversity and informal nature of sentiment expressions that cannot be trivially enumerated or captured using predefined lexical patterns. In this work, we present an optimization-based approach to automatically extract sentiment expressions for a given target (e.g., movie, or person) from a corpus of unlabeled tweets. Specifically, we make three contributions: (i) we recognize a diverse and richer set of sentiment-bearing expressions in tweets, including formal and slang words/phrases, not limited to pre-specified syntactic patterns; (ii) instead of associating sentiment with an entire tweet, we assess the target-dependent polarity of each sentiment expression. The polarity of sentiment expression is determined by the nature of its target; (iii) we provide a novel formulation of assigning polarity to a sentiment expression as a constrained optimization problem over the tweet corpus. Experiments conducted on two domains, tweets mentioning movie and person entities, show that our approach improves accuracy in comparison with several baseline methods, and that the improvement becomes more prominent with increasing corpus sizes.