網路輿情面向分析與知識圖譜建構系統之開發研究

The existing public opinion system is mainly in the monitoring and volume statistics of brands and specific entity names. In this project, we will make more detailed and accurate judgments for different aspects of named entities, on the one hand, save the manpower for labeling cross-domain training materials. At the same time, it can save the manpower of interpretation in different fields of analysis. The second year of the program will focus on the dialogue understanding module required for smart devices. We will take the music industry as an example to construct a knowledge map of the music industry, so as to provide a broader search function (such as the name of a lyricist or director) to enhance the intelligence of video search. Therefore, in the second year's plan, we will focus on the development of dialogue intention recognition and slot filling technology, so as to construct a dialogue understanding module for smart devices.