DIAC2019-DQD-Based-on-Adversarial-Attack | 5st place solution for competition Duplication Question

DIAC2019-DQD-Based-on-Adversarial-Attack is a Python library typically used in Institutions, Learning, Education applications. DIAC2019-DQD-Based-on-Adversarial-Attack has no bugs, it has no vulnerabilities and it has low support. However DIAC2019-DQD-Based-on-Adversarial-Attack build file is not available. You can download it from GitHub.

Although the QA system has made great progress in recent years, how to accurately determine whether the user's input is a semantic equivalent of a given question is still the key of a QA system (such as law, government affairs, etc.). For example, "What departments does the municipal government govern?" and "Which departments are under the jurisdiction of the municipal government?" can be considered as semantically equivalent issues, while "what departments does the municipal government govern?" and "Which departments do the mayor govern?" is different questions. For Duplication Question Detection, in addition to the accuracy of the system, the robustness of the system is also important, but it is often neglected. For example, although a deep neural network model can often achieve satisfactory accuracy on a given train set and test set, a slight change to the test set (Adversarial Attack) may cause a large decrease in overall accuracy.