CONCLUSION WE HAVE PRESENTED IN THIS PAPER AN EMPIRICAL STUDY ON QU...

6. Conclusion

We have presented in this paper an empirical study on question analysis, the first and

crucial step towards an automatic Vietnamese question answering system in the

education domain. By integrating traditional statistical models and deep neural

networks which can utilize both manually engineered and automatically learned

features, our proposed models can accurately extract fourteen types of important

information from Vietnamese questions. Our work, however, has some limitations

that we discuss in the following. First, our work is institute-specific, i.e., VNU

International School, and domain-specific, i.e., the education domain. The dataset

needs to be updated if we want to build a similar system for other schools/universities

or a system that is expected to answer questions from multidisciplinary domains.

Second, due to budget limit, our annotated corpus is quite small with 3,600 sentences.

The system could be better if we had a larger dataset which covers a wide range of

questions. Finally, our work focuses only on question analysis, not a full question

answering system. As future work, we plan to improve the performance of extraction

models with state-of-the-art deep neural networks such as attention-based

architectures. We also aim at building a QA system, which can automatically answer

questions from Vietnamese students at Vietnam National University, Hanoi.

R e f e r e n c e s