20results about How to "Improve recommendation quality" patented technology

The invention belongs to the technical field of e-commerce, and particularly relates to a commodity category recommendation management system based on an e-commerce applet, and the system comprises the steps: analyzing historical sales records, and understanding the sales-free categories and commodities; the method comprises the steps of recommending similar or related commodity categories according to commodity categories browsed or purchased by users in the past, recommending commodity categories in which the users are interested according to behavior data of similar users, classifying the users according to different dimensions, and providing customized recommendation for different groups; setting different category priorities for the recommendation system according to different recommendation strategies, and performing feedback based on real-time data; according to a recommendation system, commodity categories seen when a user enters an applet every time are dynamically generated, the conversion rate of recommendation, the click rate of the user and the sales are evaluated through a data analysis tool, behaviors of the user on a recommendation page are analyzed, and the method has the advantages that changes of user requirements can be quickly responded, recommendation quality is continuously optimized, and user experience is improved. And it is ensured that the e-commerce platform keeps leading advantages in intense market competition.

The invention provides a neural collaborative filtering recommendation model considering multi-dimensional user preferences, which comprises the following steps: acquiring and cleaning data information from a database to obtain an original user item score matrix; based on a Picture fuzzy set theory and a designed filling strategy, a user item scoring matrix is divided into four matrixes including a membership degree P, a non-membership degree N, a hesitation degree H and a rejection degree R so as to correspond to four interest preference states including user liking, disliking, hesitation and rejection; fitting the four preference matrixes by using a de-noising auto-encoder model, learning interests and preferences of the user in four dimensions, and constructing a user interest and preference model; performing normalization processing on the Picture fuzzy number (interests and preferences in four dimensions) of the user on the item obtained in the previous step, and obtaining an interests and preferences value of the user on the item through a scoring function; learning user behavior preferences by using a residual neural network fused with an attention mechanism, and generating a user behavior preference value by taking historical score information of a user on an article as input; and fusing the user interest preference and the user behavior preference in a weighting manner to form multi-dimensional user preference predicted values, and sorting the multi-dimensional user preference predicted values to generate a recommended item list.

The invention discloses an archive data management system and method in a distributed environment, and relates to the technical field of data analysis. Paper archives are converted into electronic archives through a scanning instrument, an electronic archive system is established, after user registration and identity verification, query conditions are input to generate query records, and historical query records are obtained; the method comprises the steps of collecting recommendation parameters and calculating recommendation scores, determining abnormal recommendation records, assigning user permissions, collecting user information in the abnormal recommendation records, calculating user personal scores, determining user levels, calculating abnormal frequencies of the levels, determining abnormal user levels, setting a training period, and calculating the recommendation success rate of the abnormal user levels in a certain training time period. And determining a characteristic time period of the abnormal user level, calculating a real-time recommendation success rate in the characteristic time period of the abnormal user level, and if the real-time recommendation success rate is lower than a recommendation success rate threshold, reminding a manager to optimize a recommendation method and improve recommendation accuracy.

The invention discloses an interest point recommendation method, a storage medium and equipment, and relates to the technical field of location-based social network data mining, and the method comprises the following steps: obtaining a sign-in record item sequence of a user from a location-based social network; generating item embedding vectors for the sign-in record items, constructing position embedding vectors to represent the sequence of the items, constructing frequency embedding vectors to represent the importance of repeated items, and splicing the item embedding vectors, the position embedding vectors and the frequency embedding vectors into combined embedding vectors; inputting the combined embedded vector into a bidirectional self-attention layer to obtain the representation of the combined embedded vector; according to the expressions of the item embedding vector and the combination embedding vector, using a gating vector to obtain a user intention expression; key user intention representation is obtained; on the basis of the item embedding vector, the key user intention representation is decoded and then aggregated, and interest points of the target user are generated for recommendation according to aggregated item distribution. The method provided by the invention has a remarkable effect when coping with long-sequence and multi-intention user interaction.

The invention provides a cross-domain recommendation method and related equipment. The method comprises the following steps: acquiring data sets of a plurality of fields; wherein the data set of each field comprises data related to a plurality of entities in the field; the plurality of entities comprise a target entity; on the basis of the data sets of the multiple fields, feature extraction is carried out on the target entities, and multiple target entity features corresponding to the multiple fields are obtained; determining the weight of the plurality of fields relative to any target field in the plurality of fields according to the data density of the entities contained in the data set of the plurality of fields; and based on the weights of the plurality of fields, performing weighted fusion on the plurality of target entity features, and based on the target entity features after weighted fusion, executing a recommendation task related to the target entity in the target field.

The invention discloses an evaluation method for vision-language model alignment limitation. The evaluation method comprises the following steps: respectively extracting an image and embedding text features by using a text and an image encoder of a pre-training model CLIP; calculating the similarity between the images or the texts based on the cosine similarity, and setting a threshold value to screen data meeting conditions; constructing a reference data set by adopting TT2I and II2T strategies; testing the performance of the vision-language model on Flickr30K and MSCOCO data sets, evaluating by using a recall rate R (at) K index, and analyzing the performance difference of the model on an alignment limit reference and an original data set; according to the method, the AlignVLM method shows the excellent ability in the aspect of detecting vision-language model alignment defects, and verifies the performance of different models and strategies in complex alignment tasks. An effective tool is provided for developing a more robust vision-language model, and the generalization ability of the model is improved.

The embodiment of the invention discloses an investment attraction information recommendation method, electronic equipment and a readable storage medium. The method comprises the steps that investment attraction party data and investor party data are collected; constructing an investment attracting party and investor knowledge graph, an investment attracting party attribute graph and an investor attribute graph; an investment attraction information recommendation model is constructed, the investment attraction information recommendation model comprises an input layer, a graph convolution layer and a prediction layer, and the input layer fuses internal interaction information and cross interaction information of an investment attraction party attribute graph and an investor attribute graph to obtain embedded representations of an investment attraction party and an investor; embedding representations of the investment attracting party and the investor are input into a graph convolution layer, the graph convolution layer obtains final representations of the investment attracting party and the investor through multi-layer aggregation operation, the final representations are input into a prediction layer, and the prediction layer obtains correlation scores of the investment attracting party and the investor through a three-layer multi-layer perceptron; and selecting a preset number of investors with the highest score to recommend to the investment attracting party.

The invention relates to the technical field of code completion, and discloses an intelligent code completion method under a programming education scene, which comprises the following steps: simulating answering codes of different styles of students through a general large language model in which programming learner identities are preset; each symbol of an answer code in the extended data set is coded, an obtained coding sequence is input into the large language model, the large language model predicts the next symbol of the coding sequence according to the previous part of the coding sequence, the prediction precision of the large language model is optimized by minimizing cross entropy loss, and a complementation model is obtained; and inputting answer codes of the students into the completion model which completes training to obtain an initial recommendation list, selecting three symbols which are most related to the scene information in the initial list, and increasing the order of the three symbols in the initial recommendation list to generate a final recommendation list. According to the method, the recommendation list is optimized by means of the scene information, and the recommendation quality under the programming education scene can be effectively improved.

The application provides a catering recommendation method and system applied to a digital multimedia online platform, relates to the technical field of the digital multimedia online platform, and first acquires user catering interaction intention data and platform internal catering content supply time sequence data; constructs a user catering intention evolution model based on the user catering interaction intention data and the platform internal catering content supply time sequence data, forms an intention evolution link; establishes intention supply time sequence adaptation logic in combination with the model and the supply time sequence data; screens content from a platform catering content library according to the adaptation logic to generate a dynamic recommendation scheme; pushes the dynamic recommendation scheme to a user interface and captures new interaction intention data, integrates the original data to update the model and the adaptation logic, so that the dynamic change of the user intention can be accurately grasped, dynamic recommendation is combined with supply time sequence characteristics, and the user consumption experience and platform competitiveness are improved.

The invention relates to a multimedia resource playing method and device, equipment and a storage medium, and relates to the technical field of multimedia. According to the method, an attribute setting interface is provided in a target application, the attribute setting interface provides a plurality of attribute options of a plurality of crowd attributes, and crowd attribute information is generated based on the attribute options subjectively selected by a terminal login object, so that recommended multimedia resources conform to subjective appeals of the terminal login object, and the user experience is improved. And the recommended multimedia resources are accurate. And the recommended first multimedia resources are associated with attribute options under various crowd attributes instead of multimedia resources of a certain single preference dimension, so that the recommended multimedia resources are richer and more diversified, and the recommendation quality is improved. And the first multimedia resource is the multimedia resource meeting the first condition in the plurality of multimedia resources, so that the hit rate of the first multimedia resource is higher, and the recommendation quality is further improved.

This application discloses a content recommendation method and related products, relating to the field of artificial intelligence technology. The method involves: obtaining a first sequence of operation information of a target object on a specified type of historical content in the current terminal device; obtaining a second sequence of operation information of the target object on a specified type of historical content under the current account; obtaining a first attention enhancement vector for the first operation information sequence through a first attention mechanism; obtaining a second attention enhancement vector for the second operation information sequence through a second attention mechanism; fusing the first and second attention enhancement vectors to obtain an attention fusion vector; performing feature cross-processing on the first, second, and attention fusion vectors to obtain a preference vector; and determining target content to be recommended to the target object from a candidate content library based on the similarity between the candidate content vector and the preference vector. This application can improve the quality of content recommendation.

The invention provides an interest point recommendation method and device, electronic equipment, a computer readable storage medium and a computer program product, and relates to the technical field of artificial intelligence such as data processing, interest point analysis and intelligent recommendation. A specific embodiment of the method comprises: generating an associated predicted value based on characteristic characterization and prompt characterization of a point of interest and characteristic characterization and prompt characterization of a candidate point of interest, the prompt characterization being generated based on the characteristic characterization; and determining the candidate interest point as a recommended interest point of the interest point in response to the condition that the associated prediction value is greater than or equal to a prediction value threshold.

This invention discloses a method, apparatus, device, and storage medium for determining a course recommendation list. The method includes acquiring sample data; obtaining a target basic data dictionary based on the sample data; obtaining a target sparse vector based on the sample data and the target basic data dictionary; acquiring an initial feature vector of the object to be recommended; obtaining a target feature vector of the object to be recommended based on the target basic data dictionary, the target sparse vector, and the initial feature vector of the object to be recommended; and determining a course recommendation list for the object to be recommended based on the target basic data dictionary and the target feature vector. The technical solution of this invention provides a new method for determining a course recommendation list, fully considering the characteristic information of employees from various dimensions, effectively avoiding the scenario of only recommending courses of personal interest during subsequent course recommendations, and improving the quality and accuracy of course recommendations.

This invention provides a method and apparatus for training a recommended works determination model, relating to the field of data processing technology. The method includes: acquiring historical behavior data of sample users; extracting sample content feature information and sample cover image feature information to obtain sample user feature information; generating sample feature information based on the sample content feature information, sample cover image feature information, and sample user feature information; inputting the sample feature information into a recommended works determination model to obtain an evaluation score; calculating the sample loss of the recommended works determination model based on the evaluation score and the historical behavior data of the sample users; adjusting the parameters of the recommended works determination model based on the sample loss; and returning to the step of inputting the sample feature information into the recommended works determination model until a preset training convergence condition is reached, resulting in a trained recommended works determination model. The solution provided by this invention can be used to train a model capable of recommending works to users.

The invention discloses an archive data management system and method in a distributed environment, and relates to the technical field of data analysis. Paper archives are converted into electronic archives by scanning instruments, an electronic archive system is established, and after user registration and identity authentication, query conditions are input to generate query records, historical query records are obtained, recommendation parameters are collected and recommendation scores are calculated, abnormal recommendation records are determined, user permissions are assigned, user information in abnormal recommendation records is collected, user personal scores are calculated, user levels are determined, abnormal frequencies of levels are calculated, abnormal user levels are determined, training cycles are set, recommendation success rates of abnormal user levels in a certain training time period are calculated, characteristic time periods of abnormal user levels are determined, and real-time recommendation success rates in characteristic time periods of abnormal user levels are calculated. If the rate is lower than a recommendation success rate threshold, management personnel are reminded to optimize recommendation methods and improve recommendation accuracy.

The invention belongs to the field of artificial intelligence and big data, and relates to a product recommendation test set generation method, which comprises the following steps: obtaining product description of a product; dividing feature categories to which the product descriptions belong, and generating a product category corresponding relationship between the related product descriptions and the feature categories; extracting feature keywords from the product description based on the feature categories, and generating category feature corresponding relations between the related feature categories and the feature keywords; generalizing the at least part of the feature keywords to obtain generalized feature keywords; updating a category feature corresponding relation based on the generalized feature keyword; generating a product feature relationship matrix of related products and feature keywords based on the product category corresponding relationship and the updated category feature corresponding relationship; and in combination with a preset rule, generating a test set based on the product feature relation matrix. The invention further provides a related device, computer equipment and a storage medium. The test set can be automatically generated so as to expand the coverage range of the test set.

The present application relates to natural language processing technology in the field of artificial intelligence, and discloses a prediction method and computer device for word semantics based on semantics. The method, based on HowNet, constructs a semantic space of words based on semantics, and represents known words in HowNet in the semantic space. The vector representation of words in the semantic space consists of two parts, one part of which is determined by the category semantics, and the other part of which is determined by the extended description and the relationship between the extended description and the category semantics. Since the meaning of the semantics is clear, the vector used in this application to represent words is interpretable; in the face of newly added words, the correlation between the newly added words and the known words can be calculated, and the vector representation of the newly added words can be derived from the vector representation of the known words in the semantic space. The newly added words predicted by this method have strong representation capabilities. The method of this application can be combined with a recommendation system to solve the problem of understanding and representing new words in the recommendation system and improve the quality of recommendations.

The invention relates to the technical field of natural language processing, and discloses a session method, a nonvolatile readable storage medium and electronic equipment. The method comprises the steps of obtaining historical session content and a plurality of candidate reply contents recalled by a preset recall model, determining a text vector feature corresponding to each candidate reply content according to the historical session content and each candidate reply content, determining a constraint vector feature according to the historical session content, fusing the text vector feature and the constraint vector feature, and obtaining a candidate reply content model; and obtaining session vector features, and recommending each candidate reply content according to the session vector features and a preset scoring model. According to the embodiment, the score of each candidate reply content can be adjusted in a multi-feature manner by combining the session vector feature and the constraint vector feature, so that the accuracy and reliability of the score of each candidate reply content can be improved, and the recommendation quality and session quality of each candidate reply content are further improved.

The invention relates to a privacy protection recommendation method and system based on matrix decomposition, and belongs to the technical field of privacy protection recommendation. Comprises: obtaining a historical score matrix of a user; performing matrix decomposition on the historical score matrix to obtain a shared score matrix; sending the shared scoring matrix and the scored item set to a recommendation server, and generating public parameters, a decomposition matrix and a decomposition matrix shared value; generating an inner product triple and a multiplication triple by using a triple server; calculating an error between the real score and the predicted score by using the recommendation server, and generating an error sharing matrix; the error sharing matrixes are optimized until the error sum of squares of all the error sharing matrixes is smaller than a threshold value; calculating a partial derivative of the loss function to the decomposition matrix to obtain a gradient sharing matrix; updating the decomposition matrix to obtain a user decomposition shared matrix; and generating a privacy protection recommendation result of the user by using the user decomposition sharing matrix. According to the invention, powerful guarantee is provided in the aspect of protecting user privacy.

The invention belongs to the technical field of natural language processing, and particularly relates to a library resource intelligent recommendation method and system based on natural language processing, and the method comprises the steps: obtaining historical reading information of a user; performing LDA subject analysis on each library resource in the historical reading list to obtain a subject category list; obtaining theme category preference information according to the theme category list and the reading duration information; obtaining first average reading duration information according to the theme category list and the reading duration information; obtaining recommendation values of all library resources according to the theme category preference information and the first average reading duration information, and obtaining a recommendation list of the user according to the recommendation values of all library resources; detecting feedback information of a user on the recommendation list; and updating the recommendation list according to the feedback information. The method can improve user experience and recommendation quality, and solves the problem of repeated and fixed recommendation of library resources.