CN-121980095-A - Method, device and computer program product for ordering interest points

CN121980095ACN 121980095 ACN121980095 ACN 121980095ACN-121980095-A

Abstract

The embodiment of the application discloses a method and a related device for ordering interest points. The method comprises the steps of obtaining behavior data of a target user and attribute information of candidate interest points, wherein the behavior data comprise an online behavior sequence and an offline behavior sequence of the target user, the online behavior sequence comprises interest point information clicked by the target user in a first time sequence, the offline behavior sequence comprises interest point information resided by the target user in a second time sequence, generating user characteristic representation of the target user and interest point characteristic representation of the candidate interest points according to the behavior data and the attribute information of the candidate interest points, and inputting the user characteristic representation and the interest point characteristic representation into a sorting model to obtain a sorting result of the candidate interest points. The method and the device can capture interest preferences of the user more accurately, so that user requirements can be matched more accurately during sorting, sorting accuracy is improved, and user experience is enhanced.

Inventors

XI JIUZHOU
YANG ZHEN
SONG JIAN
LI XIN

Assignees

北京高德云图科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260123

Claims (10)

1. A method of ordering points of interest, the method comprising: Acquiring behavior data of a target user and attribute information of candidate points of interest, wherein the behavior data comprises an online behavior sequence and an offline behavior sequence of the target user, the online behavior sequence comprises point of interest information clicked by the target user in a first time sequence, and the offline behavior sequence comprises point of interest information resided by the target user in a second time sequence; Generating a user characteristic representation of the target user and a point of interest characteristic representation of the candidate point of interest according to the behavior data and the attribute information of the candidate point of interest; and inputting the user characteristic representation and the interest point characteristic representation into a sequencing model to obtain a sequencing result of the candidate interest points.
2. The method of claim 1, wherein generating the user characteristic representation of the target user and the point of interest characteristic representation of the candidate point of interest based on the behavioral data and the attribute information of the candidate point of interest comprises: and inputting the behavior data and the attribute information of the candidate interest points into a fine tuning model to obtain the user characteristic representation and the interest point characteristic representation of the candidate interest points.
3. The method of claim 2, further comprising obtaining industry information to which the candidate point of interest belongs, wherein inputting the behavior data and the attribute information of the candidate point of interest into a fine tuning model to obtain the user feature representation and the point of interest feature representation of the candidate point of interest comprises: Inputting the industry information, the behavior data and the attribute information of the candidate interest points into the fine tuning model, generating the user feature representation by the fine tuning model based on the industry information and the behavior data, and generating the interest point feature representation based on the attribute information of the candidate interest points.
4. The method of claim 3, wherein the generating the user characteristic representation based on the industry information and the behavioral data comprises: encoding the industry information to obtain an industry prompt vector; respectively encoding the online behavior sequence and the offline behavior sequence, and splicing encoding results into a behavior vector sequence; fusing the industry prompt vector and the behavior vector sequence to obtain an industry enhanced behavior vector sequence; the user feature representation is generated based on the industry enhanced behavior vector sequence.
5. The method according to any one of claims 2 to 4, wherein the fine tuning model is obtained by: Acquiring training data comprising a plurality of training samples, wherein the training samples at least comprise a behavior data sample of a user, attribute information of a candidate interest point sample and behavior label true values corresponding to the candidate interest point sample, the behavior data sample comprises an online behavior sequence and an offline behavior sequence of the user, the online behavior sequence comprises interest point information clicked by the user in a first time sequence, and the offline behavior sequence comprises interest point information resided by the user in a second time sequence; and performing fine adjustment on the pre-training large model based on the training data to obtain a fine adjustment model, wherein the fine adjustment comprises the steps of inputting attribute information of the behavior data sample and the candidate interest point sample into the pre-training large model, obtaining a behavior prediction result, and updating model parameters of the pre-training large model by using a loss function value corresponding to a training target, wherein the training target comprises the step of minimizing the difference between the behavior prediction result and the behavior label true value.
6. The method of claim 5, wherein the training sample further comprises industry information to which the candidate sample of interest belongs; Inputting the attribute information of the behavior data sample and the candidate interest point sample into the pre-training large model to obtain a behavior prediction result, wherein the method comprises the following steps: And inputting the industry information of the candidate interest sample specimen, the behavior data sample and the attribute information of the candidate interest sample specimen into the pre-training large model to obtain a behavior prediction result.
7. The method of claim 6, wherein the obtaining the behavioral prediction comprises: generating the user characteristic representation based on the industry information to which the candidate interest sample specimen belongs and the behavior data sample; generating the feature representation of the interest point based on the attribute information of the candidate interest point; And generating the behavior prediction result based on the user characteristic representation and the interest point characteristic representation.
8. The method of claim 5, wherein the pre-trained large model comprises an embedded layer, a backbone network, and a predictive network, and the fine-tuning model is composed of the embedded layer and the backbone network after fine-tuning.
9. A point of interest ordering apparatus, the apparatus comprising: The system comprises an acquisition unit, a storage unit and a display unit, wherein the acquisition unit is configured to acquire behavior data of a target user and attribute information of candidate points of interest, the behavior data comprises an online behavior sequence and an offline behavior sequence of the target user, the online behavior sequence comprises point of interest information clicked by the target user in a first time sequence, and the offline behavior sequence comprises point of interest information resided by the target user in a second time sequence; a generation unit configured to generate a user feature representation of the target user and a point of interest feature representation of the candidate point of interest according to the behavior data and attribute information of the candidate point of interest; and the ranking unit is configured to input the user characteristic representation and the interest point characteristic representation into a ranking model to obtain a ranking result of the candidate interest points.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 8.

Description

Method, device and computer program product for ordering interest points Technical Field The present application relates to the field of search recommendation technologies, and in particular, to a method, an apparatus, and a computer program product for ordering points of interest. Background With the rapid development of the mobile internet, people are increasingly used to query information by using a search function, which greatly facilitates the life of people. For example, when a user searches a certain interest point (Point of Interest, POI) in the input box, the search system searches based on the interest point to obtain a plurality of candidate interest points, and then selects the candidate interest points to meet the interest preference of the user by using the search precision ranking model and performs ranking display. However, current search ranking models understand the user's interest preferences primarily from the user's online behavior sequences and related statistical features. The method is difficult to comprehensively capture the actual interests and preferences of the users, so that the information of the users is lost, and the accuracy of the sequencing results is affected. Disclosure of Invention The application provides a method, a device and a computer program product for ordering interest points, which are used for capturing interest preferences of users more accurately, so that user requirements can be matched more accurately during ordering, ordering accuracy is improved, and user experience is enhanced. The application provides the following scheme: according to a first aspect, there is provided a method of ordering points of interest, the method comprising: Acquiring behavior data of a target user and attribute information of candidate points of interest, wherein the behavior data comprises an online behavior sequence and an offline behavior sequence of the target user, the online behavior sequence comprises point of interest information clicked by the target user in a first time sequence, and the offline behavior sequence comprises point of interest information resided by the target user in a second time sequence; Generating a user characteristic representation of the target user and a point of interest characteristic representation of the candidate point of interest according to the behavior data and the attribute information of the candidate point of interest; and inputting the user characteristic representation and the interest point characteristic representation into a sequencing model to obtain a sequencing result of the candidate interest points. According to a second aspect, there is provided a point of interest ordering apparatus, the apparatus comprising: The system comprises an acquisition unit, a storage unit and a display unit, wherein the acquisition unit is configured to acquire behavior data of a target user and attribute information of candidate points of interest, the behavior data comprises an online behavior sequence and an offline behavior sequence of the target user, the online behavior sequence comprises point of interest information clicked by the target user in a first time sequence, and the offline behavior sequence comprises point of interest information resided by the target user in a second time sequence; a generation unit configured to generate a user feature representation of the target user and a point of interest feature representation of the candidate point of interest according to the behavior data and attribute information of the candidate point of interest; and the ranking unit is configured to input the user characteristic representation and the interest point characteristic representation into a ranking model to obtain a ranking result of the candidate interest points. According to a third aspect, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the steps of the method described in the first aspect above. According to the specific embodiment provided by the application, the application discloses the following technical effects: The embodiment of the application generates the user characteristic representation by combining the online behavior sequence and the offline behavior sequence of the target user, generates the interest point characteristic representation of the candidate interest point based on the attribute information of the candidate interest point, and inputs the user characteristic representation and the interest point characteristic representation into the sequencing model to predict the sequencing result of the candidate interest point. Compared with the prior art, the technical scheme comprehensively considers the online behavior and the offline behavior of the user, and can capture the interest preference of the target user more accurately and comprehensively, so that the user requirements can be matched more accurately during sorting, the sorting acc