Sarcastic remarks often appear in social media and e-commerce platforms to express almost exclusively negative emotions and opinions on certain instances, such as dissatisfaction with a purchased product or service. Thus, the detection of sarcasm allows merchants to timely resolve users’ complaints. However, detecting sarcastic remarks is difficult because of its common form of using counterfactual statements. The few studies that are dedicated to detecting sarcasm largely ignore what sparks these sarcastic remarks, which could be because of an empty promise of a merchant’s product description. This study formulates a novel problem of sarcasm cause detection that leverages domain information, dialogue context information, and sarcasm sentences by proposing a pretrained language model-based approach equipped with a novel hybrid multihead fusion-attention mechanism that combines self-attention, target-attention, and a feed-forward neural network. The domain information and the dialogue context information are then interactively fused to obtain the domain-specific dialogue context representation, and bidirectionally enhanced sarcasm-cause pair representations are generated for detecting sarcasm spark. Experimental results on real-world data sets demonstrate the efficacy of the proposed model. The findings of this study contribute to the literature on sarcasm cause detection and provide business value to relevant stakeholders and consumers. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: This work was partially supported by the National Natural Science Foundation of China [Grants 72293575, 62071467, and 62141608] and the Research Grant Council of the Hong Kong Special Administrative Region, China [Grants 11500322 and 11500421]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2022.0285 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2022.0285 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .

User demand mining aims to identify the implicit demand from the e-commerce reviews, which are always irregular, vague and diverse. Existing sentiment analysis research mainly focuses on aspect-opinion-sentiment triplet extraction, while the deeper user demands remain unexplored. In this paper, we formulate a novel research question of jointly mining aspect-opinion-sentiment-demand, and propose a Mutually Enhanced Bidirectional Extraction (MEMB) framework for capturing the dynamic interaction among different types of information. Finally, experiments on Chinese e-commerce data demonstrate the efficacy of the proposed model.

Accurately predicting hotspot reprint paragraphs can timely provide valuable clues for topic selection, thereby improving the influence of the disseminated content. Most existing works in media reprint analysis focus on mining reprint relationships and reprint patterns. Meanwhile, few works predict the hotspot reprint paragraph from a fine-grained level. The writing style reflects the structure and semantic logic of the article to some extent. Thus, the challenge is to determine how to effectively incorporate writing style features into the semantic analysis while also reasoning deeply about the semantic relevance between sections of the article. This paper proposes a multi-perspective relevance collaborative modeling method called MPRCM-TS. It integrates writing styles of titles into the semantic representations and deeply mines the multi-perspective semantic relevance between the title and paragraphs on the basis of the attention mechanism. Simultaneously, multiple loss functions collaborate to enhance the parameter optimization ability. We evaluate the performance of the proposed model on a real-world dataset, and the experimental results demonstrate the efficacy.

Performing target-oriented opinion word extraction (TOWE) from online travel reviews is a valuable reference for both tourists and attraction administration department. This paper formulates a novel research topic of identifying target-opinion pair from Chinese travel review corpus. Learning target-oriented representation accurately, locating the opinion word and extracting the complete opinion are three major challenges. Hence, we leverage aspect-based query, pos-tag and relative position and devise appropriate structure to fuse them in an encoder-decoder framework. Specifically, in the encoder, the target-fused (aspect, review) pair and the pos-tag label are encoded by transformers to model the global dependency, in the decoder, a BiLSTM is adopted to enhance contextual representation by incorporating relative position information. A real-world Chinese travel dataset for TOWE task is constructed, and the experimental results demonstrate the efficacy of the proposed model. Extensive ablation experiments are also conducted to study the effect of different components of the model.

Mining organization-related topics is helpful to analyze the information dissemination situation. Existing methods based on graph neural networks mainly consider the association between words and documents, they ignore the semantic interactions between documents, and do not consider the heterogeneity of edges which are difficult to solve the challenge of blurred topic boundaries in real scenarios, resulting in performance loss. This paper proposes a BERT-based Heterogeneous Graph Convolution Network (BERT-HGCN) approach for semi-supervised topic mining that comprehensively considers multi-semantic relations between words and documents. It deeply combines the advantages of transductive learning with pre-training models. We model documents as graph-structured data and capture multiple semantic dependencies among word-word, word-doc, and doc-doc via information propagation mechanism. During the model learning process, a two-stream encoding mechanism is used to learn the structural and semantic representations, which combines a hierarchical graph convolution network (HGCN) and a BERT-based auto-encoder. It considers both edges heterogeneity and semantics of original documents. Finally, a dual-supervision loss is used to train the classifier based on graph nodes and semantic representations for topic mining. We empirically evaluate the performance of the proposed model on a real-world organization-related dataset, and the experimental results demonstrate the efficacy of the model.

Mining users’ opinions towards the rising and falling trends of the stocks may help the management department estimate the risk and make timely decision. Existing methods ignore the effective fusion of domain information and pre-trained language models, hindering mining implicit semantic information. This paper proposes a hybrid method that adopts masked language modeling to obtain a domain-information-enhanced language model. Firstly, it generates an attention-mechanism-oriented masking based on words’ importance, word-level polarity and terminology. Then, the masked words and their corresponding knowledge are predicted to acquire domain-aware language representation. Experimental results on two public financial sentiment analysis datasets show the efficacy of the proposed model.