Methods and apparatus for identifying critical developments in events

Abstract

The invention provides a method and device for recognizing event key progress. The method includes the steps that an event search word cluster is obtained based on event key words; event key progress recognition is conducted on the event search word cluster based on the media attention meter, so that a key progress point set of a first event is obtained; event key progress recognition is conducted on the event key words based on news hot search words, so that a key progress point set of a second event is obtained; fusion and duplicate removal are carried out on the key progress point set of the first event and the key progress point set of the second event, so that a key progress point set of a third event is obtained; the key progress point set of the third event is optimized, and the key progress context of a final event is obtained. According to the method and device for recognizing the event key progress, apparent event context can be provided, the event attention demands of users are met, the user experience is improved, manual editing is needless, and special subject generation cost is greatly lowered.

Description

technical field [0001] The present invention relates to a news topic tracking technology, and more specifically, to a method and device for identifying key progress of an event by determining the occurrence time point of an important sub-event without manual labeling. Background technique [0002] With the rapid development of network technology, whether using a mobile terminal (for example, a mobile phone, etc.) or a non-mobile terminal (for example, a desktop computer, etc.) to browse news on the Internet has become the most common way of leisure for people. According to a survey conducted by Tencent Technology, 61.67% of surveyed users use mobile phones to surf the Internet mainly to browse news. In news portal websites, events are generally called topics, and a popular event (or topic) usually consists of several sub-events. Each event has a process of generation, development, climax, and end. The important sub-events in the whole process are connected in series to form...

AI Technical Summary

Problems solved by technology

However, event correlation often causes topic drift due to the influence of event lace news

Another way is to use document clustering to associate all the clusters obtained in the previous stage with the topic clusters clustered in the current stage. Since document clustering belongs to unsupervised learning, the cost of clustering is high.

At the same time, the way of event tracking is difficult to deal with the cold start problem, and all current topics need to be associated and matched with all historical topics. The later development costs are relatively large and the development cycle is long.

Images