A5032690266- Topic Modeling
- Natural Language Processing Techniques
- Advanced Graph Neural Networks
- Text and Document Classification Technologies
- Recommender Systems and Techniques
- Information and Cyber Security
- Network Security and Intrusion Detection
- Advanced Malware Detection Techniques
- Speech and dialogue systems
- Adversarial Robustness in Machine Learning
- Sentiment Analysis and Opinion Mining
- Medical Imaging Techniques and Applications
- Advanced X-ray and CT Imaging
- Cardiac Imaging and Diagnostics
Tianjin University
2023
National University of Defense Technology
2021-2023
Job-resume matching (JRM) is the core of online recruitment services for predicting degree between a job post and resume. Most existing methods JRM achieve promising performance by simplifying this task as free-text attributes in However, they neglect contributions semistructured multivariate such education salary, which will result an unsuccessful prediction. To address issue, we propose novel approach to comprehensively explore Internal EXternal InTeractions JRM, i.e., InEXIT. In detail,...
Log-based insider threat detection (ITD) detects malicious user activities by auditing log entries. Recently, large language models (LLMs) with strong common sense knowledge have emerged in the domain of ITD. Nevertheless, diverse activity types and overlong files pose a significant challenge for LLMs directly discerning ones within myriads normal activities. Furthermore, faithfulness hallucination issue from aggravates its application difficulty ITD, as generated conclusion may not align...
This study aimed to develop and validate an AI (artificial intelligence)-aid method in myocardial perfusion imaging (MPI) differentiate ischemia coronary artery disease.We retrospectively selected 599 patients who had received gated-MPI protocol. Images were acquired using hybrid SPECT-CT systems. A training set was used train the neural network a validation test predictive ability of network. We learning technique named "YOLO" carry out process. compared accuracy with that physician...
Text classification aims to assign predefined labels unlabeled sentences, which tend struggle in real-world applications when only a few annotated samples are available. Previous works generally focus on using the paradigm of meta-learning overcome difficulties brought by insufficient data, where set auxiliary tasks is given. Accordingly, prompt-based approaches proposed deal with low-resource issue. However, existing methods mainly English tasks, apply pretrained language models that can...
Generating diversified and affected context-consistent responses is key to intelligent dialogue systems. Previous works about emotional response generation mainly refer incorporate the category into decoding process, ignoring historical emotion in conversation context. In addition, existing methods fail consider diversity of categories simultaneously. However, crucial improve quality human-machine conversations. this paper, we propose an Emo-CV Aemodel that able capture variations signal...
Label representation aims to generate a so-called verbalizer an input text, which has broad application in the field of text classification, event detection, question answering, etc. Previous works on label representation, especially few-shot setting, mainly define verbalizers manually, is accurate but time-consuming. Other models fail correctly produce antonymous for two semantically opposite classes. Thus, this paper, we propose metric sentiment learning framework (MSeLF) automatically,...
Fact verification is a recently introduced task, which quired to judge the authenticity of claim. Previous works mainly leverage extracting semantic relations claim and evidences, e.g., using graph model their relation. However, these graph-based models ignore features evolution over different step propagation. In addition, they fail capture sequence evidence graph. To deal with above issues, we propose dynamic attention recurrent neural network for fact verification. Specifically, design...