8results about How to "Improve recommendations" patented technology

The invention relates to an optimization method and device of a fine arrangement model, a storage medium and electronic equipment. The method comprises the following steps: acquiring user behavior data, marking an object clicked by a target user in the user behavior data as a positive sample, and marking an object which is not clicked as a negative sample; obtaining a model prediction probability, an exposure frequency and a sample feature of each negative sample, and calculating a dynamic weight of each negative sample; inputting all browsing objects and object information thereof in the user behavior data into a fine arrangement model to perform recommendation probability prediction on all the browsing objects to obtain a prediction result, and calculating to obtain a basic loss function according to the prediction result and click information of all the browsing objects; and performing weighted calculation on the basic loss function according to the dynamic weight of each negative sample to obtain a weighted loss function, and iteratively updating parameters of the fine arrangement model according to the weighted loss function. The technical problems that an existing fine ranking model does not make full use of difficult-to-load samples, and the judgment and ranking precision is limited are solved.

This application is intended to provide a user recommendation method and device. In comparison to the prior art, in this application, regional activity record information about a target user is determined according to connection record information (S11) about the target user establishing a connection to a wireless access point, a matching user corresponding to the target user is determined based on overlap of activity regions of different users, and then the matching user is provided to the target user (S12). According to this application, a user is recommended based on connection record information about the user establishing a connection to a wireless access point, to help the user develop a related social relationship chain, thereby improving user experience. Further, in this application, the matching user corresponding to the target user is determined according to the regional activity record information about the target user and in combination with a plurality of user matching rules. According to this application, users are classified according to a plurality of flexible user matching rules, so that the users are more appropriately recommended, and user experience is optimized.

An article recommendation method, a computer device and a computer readable storage medium, the method comprising: obtaining a plurality of evaluation texts; performing word segmentation processing on each evaluation text to obtain a corresponding sentence set; extracting sentiment evaluation words of each evaluation text using a preset sentiment dictionary to obtain a corresponding sentiment evaluation word set; obtaining a corresponding sentiment evaluation unit set based on a preset evaluation object dictionary, a preset adverb dictionary and the sentence set and the sentiment evaluation word set of each evaluation text; calculating an evaluation score of the evaluation text for the article recorded by the evaluation text based on the sentiment evaluation unit set of the evaluation text, and obtaining an evaluation value of each article by aggregating the evaluation scores of each article; and selecting at least one article from the articles according to the evaluation value of each article to recommend to a target user. The present application constructs sentiment evaluation units based on evaluation texts, and then obtains the evaluation score of the evaluation text for the article to recommend the article, so that the article recommendation effect is better.

This invention provides a method, apparatus, device, and storage medium for recommending technologies that link supply and demand. The method includes: obtaining technical information from technology suppliers through a patent trading platform; constructing a technology library based on the technical information, which includes patent application texts, application dates, and IPC classification numbers; obtaining demand information from technology demanders through the patent trading platform and classifying the demands, with each demand classification associated with the IPC classification number; filtering technical information in the technology library according to set conversion criteria to obtain technologies to be converted; classifying the technologies to be converted based on the IPC classification number to generate a list of technologies to be converted, and pushing this list to the technology demanders corresponding to the demand categories associated with the IPC classification numbers. This invention can break down information barriers between patent technology suppliers and demanders, enabling targeted recommendations of patent technologies, improving patent transfer efficiency, and meeting the needs of both parties.

The application provides a method for analyzing user level intention from the space-time angle in conversation recommendation, and the training steps include an intention perception step, a time intention decoupling step, a feature transformation step, an item prediction step and a cross-scale contrast learning step. Compared with traditional conversation recommendation models, the application innovatively introduces the concept of hierarchical intention, uses a time intention decoupling module to decouple long-term intention and short-term intention from the original conversation, thereby avoiding mutual interference in the learning process. Secondly, in order to utilize the compatibility of spatial dimension intention, HearInt also proposes a cross-scale contrast learning loss calculation method, which is beneficial to the model to integrate large-scale intention information into small-scale intention when perceiving intention, thereby enriching the information amount of small-scale intention. The application not only enriches the information amount of the current conversation representation, but also fully excavates the user intention which can achieve the best recommendation effect in the conversation recommendation reference.

This application discloses a content recommendation method, apparatus, storage medium, and device, belonging to the field of artificial intelligence technology. It includes: for the t-th time step of the current recommendation round, acquiring user data of the target user to be recommended; generating state data for the t-th time step based on the first t-1 recommended content generated in the previous t-1 time steps and the user data acquired at the t-th time step; determining the recommended content for the t-th time step from the candidate content set based on the state data of the t-th time step; repeatedly executing the steps of generating recommended content at each time step until N recommended content items are obtained; the reward value corresponding to any recommended content item is related to the recommended content item and other recommended content items ranked after it; and outputting a list of recommended content items to the target user, where the N recommended content items in the list are arranged in order. This application can recommend a list that maximizes the overall reward, achieving the expected recommendation effect.

The application discloses a kind of interest point recommendation methods based on cross-city knowledge migration, first, three kinds of graph structures are constructed for source city and target city respectively, include: user-interest interaction graph, user-user association graph, interest point-interest point association graph, wherein user-interest point interaction graph is as main view, the rest two are as auxiliary view;And utilize multilayer graph neural network on each view carries out information propagation and feature learning;Second, design cross-view contrast learning task, maximize the consistency of the representation of the same entity object in the main view and the corresponding auxiliary view, enhance the discriminability and robustness of the model under the condition of data sparsity;Then, a double adapter transfer architecture is constructed, a two-stage training strategy is used.The method of the application can effectively alleviate the data sparsity problem of the target city, and provides strong support for improving the interest point recommendation effect in data scarce cities.

The application discloses a commodity list file generation method, device and system. The method is used for inferring the certainty degree of a customer to a commodity by searching the number of times of the customer to the commodity in a historical time, and inferring the interest degree of the customer to other commodities based on the certainty degree. Finally, the commodity list file of the customer is generated by setting a priority, so that the server can recommend commodities to the customer according to the commodity list file. The device and the system are used for executing the commodity list file generation method. The application is mainly used in the technical field of e-commerce.