WSDM is an international top conference in the field of information retrieval and data mining. It is organized by SIGIR, SIGKDD, SIGMOD and SIGWEB, and enjoys a high academic reputation in the field of Internet search and data mining. 202 1 1 Up to now, Gaoying Institute of Artificial Intelligence has published or been employed in 67 CCF A international journals and conference papers and 34 CCF B journals and papers.
In conversation search, it is very important to use the historical interaction between users and search engines to improve the efficiency of document retrieval. But not all historical information is helpful to the ranking of candidate documents. In fact, users often express their preferences in the process of modifying each query, which can help us capture useful information in historical interaction.
As an important resource, knowledge map has been widely used in information retrieval, recommendation system, natural language processing and other fields. However, knowledge maps are usually faced with incomplete phenomena. Knowledge completion is to use the structural information of knowledge map to predict the missing triplets in knowledge map, which has become a research hotspot in the field of knowledge map. And entity type prediction is an effective means to complete knowledge map.
An efficient and universal time series representation method
1, dictionary method, find the eigenvalue of time series segmentation.
2. Shape method to find the special waveform of time series segmentation.
3. Clustering method is used to find the classification features of time series segmentation.
Based on the above background, in order to describe the dynamic information of time series and provide an interpretable model representation for anomaly detection, this paper attempts to map Shapelet back to time series, explore the sensitivity of position, accumulate the transition relationship with time, and construct a graph to represent it, thus forming a rational and interpretable time series modeling and anomaly analysis method.