International Journal of Software Engineering and Knowledge Engineering

In the contemporary social environment, social crisis events occur frequently with significant impacts. Effective management of these events requires comprehensive group intention mining, which encompasses intention detection and intention attribution. Knowledge graph inference facilitates the detection of group intention in crisis events. This is supported by the construction of crisis knowledge graphs, which organize crisis elements and inter-element relations into structured semantic information. This paper provides a comprehensive overview of the research about knowledge graph in social crisis management, focusing on three key areas: knowledge graph construction and inference, knowledge graph-based interpretable crisis attribution, and risk management. Specifically, the interpretable semantics in crisis knowledge graphs enables attribution of intention. To illustrate the significance of knowledge graphs in group intention mining, the COVID-19 and China–US game events are selected as two case studies. Finally, the paper proposes future research directions to solve the limitations of existing knowledge graph-related methods in social crises.

Dialogue generation, as a crucial subtask of dialogue systems, is garnering increasing attention in the field of Natural Language Processing (NLP). The success of dialogue generation relies on effectively utilizing context information to ensure coherent and diverse responses. However, current approaches heavily rely on external sources rather than leveraging the inherent dialogue content. We propose a new approach to address this challenge by introducing semantic segmentation from the field of image processing into NLP. Our contribution lies in the development of a Dialogue Generation model with Hierarchical Encoding and Semantic segmentation of dialogue Context, which is called DGHESC. This model is topic and speaker-aware, capturing the flow of topic and speaker information within the dialogue context using a hierarchical transformer-based framework. Specifically, we extract semantic information at the word-level for each utterance, segment the dialogue context based on topic and speaker semantics, and employ attention mechanisms to model the context at the utterance-level. Experimental results on two open-domain datasets demonstrate the effectiveness of DGHESC. It enhances response quality and achieves state-of-the-art performances on the datasets.

With the increasing number of patent applications over the years, instances of patent infringement cases have become more frequent. However, traditional manual patent infringement detection models are no longer suitable for large-scale infringement detection. Existing automated models mainly focus on detecting one-to-one patent infringements, but neglect the many-to-one scenarios. The many-to-one patent infringement detection model faces some major challenges. First, the diversity of patent domains, complexity of content and ambiguity of features make it difficult to extract and represent patent features. Second, patent infringement detection relies on the correlation between patents and the comparison of contextual information as the key factors, but modeling the process and drawing conclusions present challenges. To address these challenges, we propose a many-to-one patent graph (MPG) embedding base infringement detection model. Our model extracts the relationship between keywords and patents, as well as association relation between keywords from many-to-one patent texts (MPTs), to construct a MPG. We obtain patent infringement features through graph embedding of MPG. By using these embedding features as input, the many-to-one infringement detection (MOID) model outputs the conclusion on whether a patent is infringed or not. The comparative experimental results indicate that our model improves accuracy, precision and F-measure by 3.81%, 11.82% and 5.37%, respectively, when compared to the state-of-the-art method.

Continuous reinforcement learning carries potential security risks when applied in real-world scenarios, which could have significant societal implications. While its field of application is expanding, the majority of applications still remain confined to virtual environments. If only a single continuous learning method is applied to an unmanned system, it will still forget previously learned experiences, and retraining will be required when it encounters unknown environments. This reduces the learning efficiency of the unmanned system. To address these issues, some scholars have suggested prioritizing the experience playback pool and using transfer learning to apply previously learned strategies to new environments. However, these methods only alleviate the speed at which the unmanned system forgets its experiences and do not fundamentally solve the problem. Additionally, they cannot prevent dangerous actions and falling into local optima. Therefore, we propose a dual decision-making continuous learning method based on simulation to reality (Sim2Real). This method employs a knowledge body to eliminate the local optimal dilemma, and corrects bad strategies in a timely manner to ensure that the unmanned system makes the best decision every time. Our experimental results demonstrate that our method has a 30% higher success rate than other state-of-the-art methods, and the model transfer to real scenes is still highly effective.

With the explosion of 5G network scale, the network structure becomes increasingly complex. During the operation of the network devices, the probability of anomalies or faults increases accordingly. Network faults may lead to the disappearance of important information and cause unpredictable losses. The prediction of network faults can enhance the quality of network services and reduce economic loss. In this paper, we propose the concept of 4D features and use the BERT algorithm to extract semantic features, the graph neural network algorithm to extract network topology information, and the Temporal Convolutional Network (TCN) algorithm to extract time series. Based on this, we propose Fault Prediction based on GraphSage and TCN (GTFP), an end-to-end solution of network fault alarm prediction, which is based on GraphSage and TCN (GTCN), a hybrid algorithm of a graph neural network and the TCN model. Our solution takes the historical alarm data as input. First, we filter out the alarm noises irrelevant to the faults through data cleaning. Then, we employ feature engineering to extract the valid alarm features, including the statistical features of the network alarm information, the semantic features of the alarm texts, the sequential features of the alarms and the network topology features of the nodes where the alarms are located. Finally, we use GTCN to predict future fault alarms based on the extracted features. Experiments on the alarm data of a real service system show that GTFP performs better than the state-of-the-art algorithms of fault alarm prediction.

Interpretation is important for adopting software defect prediction in practice. Model-agnostic techniques such as Local Interpretable Model-agnostic Explanation (LIME) can help practitioners understand the factors which contribute to the prediction. They are effective and useful for models constructed on tabular data with traditional features. However, when they are applied on source code-based models, they cannot differentiate the contribution of code tokens in different locations for deep learning-based models with Bag-of-Word features. Besides, only using limited features as explanation may result in information loss about actual riskiness. Such limitations may lead to inaccurate explanation for source code-based models, and make model-agnostic techniques not useful and helpful as expected. Thus, we apply a perturbation-based approach Randomized Input Sampling Explanation (RISE) for source code-based defect prediction. Besides, to fill the gap that there lacks a systematical evaluation on model-agnostic techniques on source code-based defect models, we also conduct an extensive case study on the model-agnostic techniques on both token frequency-based and deep learning-based models. We find that (1) model-agnostic techniques are effective to identify the most important code tokens for an individual prediction and predict defective lines based on the importance scores, (2) using limited features (code tokens) for explanation may result in information loss about actual riskiness, and (3) RISE is more effective than others as it can generate more accurate explanation, achieve better cost-effectiveness for line-level prediction, and result in less information loss about actual riskiness. Based on such findings, we suggest that model-agnostic techniques can be a supplement to file-level source code-based defect models, while such explanations should be used with caution as actual risky tokens may be ignored. Also, compared with LIME, we would recommend RISE for a more effective explanation.