News information pushing method and device
A news and information technology, applied in the field of knowledge graphs, can solve problems such as manual selection and unrecognizable news information
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
example 1
[0046] In this example, the event name extraction model is pre-trained, and the event name extraction model learns the correspondence between news titles and event names in advance. During the training process, such as figure 2 As shown, a large amount of sample data is obtained in advance, each sample data includes a marked pair, and an event name extraction model is generated according to the marked pair training, and then, in this implementation In the example, in order to further improve the efficiency of determining the event name, such as figure 2 As shown, multiple candidate events in the news title can also be extracted according to the preset event name extraction strategy (such as segmenting news titles to obtain multiple word segments, and using part-of-speech word segments such as nouns as candidate event names according to the part-of-speech of multiple word segments) Input a plurality of candidate event names into the event name extraction model, and determin...
example 2
[0048] In this example, the event name extraction neural network is trained in advance based on a large number of sample data, and the word segmentation features of each word in the news title are extracted. The output probability value determines whether each token is an event name.
[0049] Step 102, if at least one event name is extracted, match each event name in the at least one event name with a plurality of event clusters in the preset event library, wherein each event cluster in the plurality of event clusters contains the same Multiple news items under one event.
[0050] In one embodiment of the present application, if the event name cannot be extracted, the news information is sent to the manual platform for processing.
[0051] It can be understood that the events to which each generated news information is sorted out, so that for the published historical news information, multiple event clusters are finally sorted out according to the events to which they belong,...
Embodiment 1
[0068] In this example, if Figure 7 As shown, after the above step 104, the method also includes:
[0069] In step 201, a news reading request is obtained, and a reading event corresponding to the news reading request is determined.
[0070] In practical applications, users will search for news in browsers, etc. In order to provide users with more relevant news information, in this embodiment, a news reading request is obtained and a reading event corresponding to the news reading request is determined. The event extraction method may be to extract the event name in the reading request. The event name extraction method may refer to the event name extraction method in the above embodiment. In this embodiment, the event name may be used as the reading event.
[0071] Step 202, judging whether the read event matches the event corresponding to the target event cluster.
[0072] As a possible implementation, the event name can be matched with the theme event name in the event cl...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More