[1] Peng, W., Li, W. (co-first author), Hu Y. (2023, July). Leader-Generator Net: Dividing Skill and Implicitness for Conquering FairytaleQA. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR ’23). Association for Computing Machinery, New York, NY, USA, 791–801. (CCF A)

• Machine reading comprehension is a generally important domain within question answering. However, question answering involves many fine-grained skills - including the explicitness of the question or implicitness - i.e. whether the answer is directly contained in the passage or inferred.
• Mimicking human reading processes, We propose a new QA approach, Leader-Generator Network, that explicitly models both the fine-grained reading skills as well as the implicitness/explicitness of the question being asked.
• Specifically, the proposed Skill Leader accurately captures the difference and semantic representation of fine-grained reading skills with contrastive learning. And the implicitness-aware pointer-generator adaptively extracts or generates the answer based on the implicitness or explicitness of the question.
• To validate the generalizability of the methodology, we annotate a new dataset named “NarrativeQA 1.1”. Experiments on the “FairytaleQA” and “NarrativeQA 1.1” show that the proposed model achieves the state-of-the-art performance (about 5% gain on Rouge-L) on the question answering task.
• PAPER & Code

 
[2] Li, W., Wang, X., & Feng, Q. (2021, March). Final Prediction of Product Quality in Batch Process based on Bidirectional Neural Network Algorithm. In IOP Conference Series: Earth and Environmental Science (Vol. 692, No. 3, p. 032091). IOP Publishing.

• There is a huge amount of alarm information in the secondary equipment of the power system; however, most of them are useless information and interference information, which caused a lot of trouble to accurately find the correct alarm information.
• Due to the low efficiency and bad accuracy of the commonly used manual experience screening method, we built up the LSTM deep learning NLP network screening model, which greatly improved the performance of alarm information screening, and can accurately locate specific alarm signals.
• PAPER

 
[3] Li, W., Liu, Y., & Jiang, W. (2021, Feb). Research on the Method of Screening Alarm Information of Secondary Equipment in Power System Based on Deep Learning NLP Technology. Basic & Clinical Pharmacology & Toxicology, 128(S1), 89–90. (Impact factor: 4.080)

• Based on the analysis of time series characteristics of the production process on common batch process endpoint quality prediction, a predictive model with bidirectional gated loop neural network is proposed to predict final product quality for unequal interval batch processes.
• Based on the requirement of forecasting value in actual production, loss function adapted to batch process is constructed, which makes model meet forecasting requirement under guaranteed prediction precision, thus obtaining greater production benefit.
• Compared with MPLS, SVR and GRU algorithms, the BiGRU model-based batch process product quality prediction method achieves better prediction results than the traditional algorithm, which verifies that the bidirectional gated circulation unit neural network has better prediction ability for industrial batch process data.
• PAPER