人机对话理解中联合学习技术的研究

joint learning model is established for the slot filling and intention detection tasks in spoken dialogue understanding, the conditional random fields and bi-directional long short-term memory models are applied to the joint learning model. The method of adding the intention identifier at the end of the sentence is determined by experiments, and the effect of the joint learning model is improved by using the feResearch on Joint Learning Technology in Spoken Language Understandingature of the dialogue history information. The experiment results of different size of training set indicate that the joint learning model based on conditional random fields is better under the current corpus scale, but with the increase of the size of the corpus, the effect growth trend of joint learning model based bi-directional long short- term memory is more obvious.