Knowledge information recommendation method, device and equipment

Abstract

The present application provides a method, device and equipment for recommending knowledge information, and relates to the technical field of recommendation systems. The method includes: obtaining the portrait tag of the current user; searching for knowledge tags matching the portrait tags in a preset tag library; wherein the tag library includes the correspondence between the portrait tags and the knowledge tags, and the correspondence between the knowledge tags and the knowledge information; according to the correspondence between the knowledge tags and the knowledge information, determining the knowledge information to be recommended corresponding to the knowledge tags, and pushing the knowledge information to the user. The method of the present application breaks the semantic barrier between the library where the portrait tags are located and the library where the knowledge tags are located, and integrates the library where the portrait tags are located with the library where the knowledge tags are located. It can be flexibly applied to multiple business fields, and is convenient for recommending knowledge tags related to the portrait to the user in a more targeted manner based on the user's portrait tags, and solves the technical problem that the independent tag system has low targeting in recommending knowledge information to the user.

Description

Technical Field

[0001] The present application relates to the technical field of recommendation systems, and in particular to a knowledge information recommendation method, device and equipment. Background Art

[0002] At present, in practical applications, after a label system is established based on the knowledge base itself, knowledge information can be recommended to users based on the label system.

[0003] In the prior art, when recommending knowledge information to users based on a tag system, different business fields will establish a tag system based on their own knowledge bases, and then recommend knowledge information to users based on their own tag systems.

[0004] However, in the prior art, since the label systems of different business fields are independent, the label system of each business field can only serve its own business, which will result in less knowledge information recommended to users by the label system. Summary of the invention

[0005] The present application provides a knowledge information recommendation method, device and equipment to solve the technical problem that an independent tag system has low pertinence in recommending knowledge information to users.

[0006] In a first aspect, the present application provides a knowledge information recommendation method, comprising:

[0007] Get the current user's portrait tag;

[0008] Searching for a knowledge tag matching the portrait tag in a preset tag library; wherein the tag library includes a correspondence between the portrait tag and the knowledge tag, and a correspondence between the knowledge tag and the knowledge information;

[0009] According to the correspondence between the knowledge tag and the knowledge information, the knowledge information to be recommended corresponding to the knowledge tag is determined, and the knowledge information is pushed to the user.

[0010] Furthermore, searching a preset tag library for a knowledge tag matching the portrait tag includes:

[0011] According to the correspondence between the preset portrait tags and knowledge tags, the knowledge tags matching the portrait tags are searched in the tag library.

[0012] Furthermore, the method further comprises:

[0013] Obtain multiple portrait tags, knowledge tags and knowledge information;

[0014] A first corresponding relationship between the knowledge tag and the knowledge information is established, a second corresponding relationship between the portrait tag and the knowledge tag is established, and a tag library is established based on the first corresponding relationship and the second corresponding relationship.

[0015] Furthermore, the knowledge tags include core knowledge tags and generalized knowledge tags; obtaining multiple core knowledge tags includes:

[0016] Acquire a plurality of first knowledge texts and second knowledge texts to be detected;

[0017] Performing word segmentation and screening processing on the first knowledge text to obtain a plurality of text words;

[0018] Annotating the text words according to a preset annotation type to obtain annotated text words;

[0019] The neural network model is trained according to the annotated text words to obtain a label classification model;

[0020] The second knowledge text is detected through a label classification model to obtain multiple core knowledge labels.

[0021] Furthermore, multiple generalized knowledge labels are obtained, including:

[0022] Determine the core distribution data of the core knowledge tags in each second knowledge text;

[0023] Randomly assign a core knowledge label to each text word, and determine the text word distribution data of each text word; each of the core knowledge labels corresponds to a core number;

[0024] Determine and update a new core number corresponding to the text word using a Gibbs sampling formula, and update the new core number to the text word number of the text word until Gibbs sampling converges;

[0025] According to the text word numbers of the text words, the core knowledge label to which each text word belongs is determined, and the text word corresponding to the core knowledge label is used as the generalized knowledge label corresponding to the core knowledge label.

[0026] Furthermore, establishing a first corresponding relationship between the knowledge tag and the knowledge information includes:

[0027] Analyze and process each sentence in the second knowledge text according to the dependency syntax method to obtain the grammatical relationship between the knowledge label and the knowledge information in each sentence;

[0028] According to the grammatical relationship between the knowledge tags and the knowledge information, a first corresponding relationship between the knowledge tags and the knowledge information is established and stored; wherein, multiple groups of knowledge tags and knowledge information generate a knowledge tag graph.

[0029] Furthermore, establishing a second corresponding relationship between the portrait label and the knowledge label includes:

[0030] Extracting knowledge tags in the knowledge tag graph through a preset tag classification model, converting the extracted multiple knowledge tags into knowledge tag vectors, and converting the multiple portrait tags into portrait tag vectors;

[0031] The portrait label vector and the knowledge label vector are embedded in the same matrix, and the matrix is ​​calculated using a preset cosine similarity method to obtain a semantic similarity score between each portrait label vector and each knowledge label vector;

[0032] Determine the association value between each knowledge label vector and each image label vector;

[0033] According to the correspondence between each portrait label vector, each knowledge label vector, and the semantic similarity score, and the correspondence between each portrait label vector, each knowledge label vector, and the association value, the product of the semantic similarity score corresponding to each knowledge label vector and each portrait label vector and the association value is calculated and obtained;

[0034] According to the product, a second corresponding relationship between the knowledge label corresponding to each knowledge label vector and the image label corresponding to each image label vector is determined and established.

[0035] Furthermore, determining the association value between each knowledge label vector and each portrait label vector includes:

[0036] Determine a first probability that the knowledge label vector and the portrait label vector both appear in the plurality of second knowledge texts at the same time, and a second probability that the knowledge label vector and the portrait label vector both appear separately in the plurality of second knowledge texts;

[0037] A first expectation corresponding to the first probability and a second expectation corresponding to the second probability are determined, and an association value between each knowledge label vector and each portrait label vector is determined based on the first expectation and the second expectation.

[0038] In a second aspect, the present application provides a knowledge information recommendation device, comprising:

[0039] A first acquisition unit, used to acquire a portrait tag of the current user;

[0040] A search unit, used to search for a knowledge tag matching the portrait tag in a preset tag library; wherein the tag library includes a correspondence between the portrait tag and the knowledge tag, and a correspondence between the knowledge tag and the knowledge information;

[0041] A determination unit, configured to determine the knowledge information to be recommended corresponding to the knowledge tag according to the correspondence between the knowledge tag and the knowledge information;

[0042] The push unit is used to push the knowledge information to the user.

[0043] Furthermore, the search unit is specifically used to:

[0044] According to the correspondence between the preset portrait tags and knowledge tags, the knowledge tags matching the portrait tags are searched in the tag library.

[0045] Furthermore, the device also includes:

[0046] A second acquisition unit, used to acquire a plurality of knowledge tags;

[0047] A third acquisition unit is used to acquire multiple portrait tags and knowledge information;

[0048] A first establishing unit, used to establish a first corresponding relationship between the knowledge tag and the knowledge information;

[0049] A second establishing unit, used to establish a second corresponding relationship between the portrait label and the knowledge label;

[0050] The third establishing unit is used to establish a label library according to the first corresponding relationship and the second corresponding relationship.

[0051] Furthermore, the knowledge tags include core knowledge tags and generalized knowledge tags; and the second acquisition unit includes:

[0052] An acquisition module, used for acquiring a plurality of first knowledge texts and second knowledge texts to be detected;

[0053] A processing module, used for performing word segmentation and screening on the first knowledge text to obtain a plurality of text words;

[0054] A marking module, used to mark the text words according to a preset marking type to obtain marked text words;

[0055] A training module is used to train the neural network model according to the annotated text words to obtain a label classification model;

[0056] The detection module is used to detect the second knowledge text through a label classification model to obtain multiple core knowledge labels.

[0057] Furthermore, the second acquiring unit includes:

[0058] A core distribution determination module, used to determine core distribution data of core knowledge tags in each second knowledge text;

[0059] A text word distribution determination module is used to randomly assign a core knowledge label to each text word and determine the text word distribution data of each text word; each of the core knowledge labels corresponds to a core number;

[0060] A text word number updating module is used to determine and update a new core number corresponding to the text word by using a Gibbs sampling formula, and update the new core number to the text word number of the text word until the Gibbs sampling converges;

[0061] The generalized knowledge label determination module is used to determine the core knowledge label to which each text word belongs according to the text word number of the text word, and use the text word corresponding to the core knowledge label as the generalized knowledge label corresponding to the core knowledge label.

[0062] Furthermore, the first establishing unit includes:

[0063] An analysis module, used to analyze and process each sentence in the second knowledge text according to a dependency syntax method to obtain a grammatical relationship between the knowledge label and the knowledge information in each sentence;

[0064] The first establishing module is used to establish and store a first corresponding relationship between the knowledge tag and the knowledge information according to the grammatical relationship between the knowledge tag and the knowledge information; wherein, multiple groups of knowledge tags and knowledge information generate a knowledge tag map.

[0065] Furthermore, the second establishing unit includes:

[0066] A conversion module, used to extract knowledge tags in the knowledge tag graph through a preset tag classification model, convert the extracted multiple knowledge tags into knowledge tag vectors, and convert the multiple portrait tags into portrait tag vectors;

[0067] A score calculation module, used for embedding the portrait label vector and the knowledge label vector into the same matrix, and calculating the matrix using a preset cosine similarity method to obtain a semantic similarity score between each portrait label vector and each knowledge label vector;

[0068] A correlation value determination module, used to determine the correlation value between each knowledge label vector and each portrait label vector;

[0069] A product calculation module is used to calculate and obtain the product of the semantic similarity score and the association value corresponding to each knowledge label vector and each portrait label vector according to the corresponding relationship between each portrait label vector, each knowledge label vector, and the semantic similarity score, and the corresponding relationship between each portrait label vector, each knowledge label vector, and the association value;

[0070] The second establishing module is used to determine and establish a second corresponding relationship between the knowledge label corresponding to each knowledge label vector and the image label corresponding to each image label vector according to the product.

[0071] Furthermore, the module for determining the associated value includes:

[0072] A determination probability submodule, used to determine a first probability that the knowledge label vector and the portrait label vector both appear in a plurality of second knowledge texts at the same time, and a second probability that the knowledge label vector and the portrait label vector both appear separately in a plurality of second knowledge texts;

[0073] The association value determination submodule is used to determine a first expectation corresponding to the first probability and a second expectation corresponding to the second probability, and determine the association value between each knowledge label vector and each portrait label vector based on the first expectation and the second expectation.

[0074] In a third aspect, the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0075] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method described in the first aspect.

[0076] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the method described in the first aspect when executed by a processor.

[0077] The present application provides a method, device and equipment for recommending knowledge information, which obtains the portrait tag of the current user; searches for knowledge tags matching the portrait tag in a preset tag library; wherein the tag library includes the matching relationship between the portrait tag and the knowledge tag, and the corresponding relationship between the knowledge tag and the knowledge information; according to the corresponding relationship between the knowledge tag and the knowledge information, determines the knowledge information to be recommended corresponding to the knowledge tag, and pushes the knowledge information to the user. In this scheme, since the preset tag library includes the matching relationship between the portrait tag and the knowledge tag, and the corresponding relationship between the knowledge tag and the knowledge information, it is possible to search for the knowledge tag matching the portrait tag in the tag library according to the matching relationship between the portrait tag and the knowledge tag, and then determine the knowledge information to be recommended corresponding to the knowledge tag in the tag library according to the corresponding relationship between the knowledge tag and the knowledge information, and push the knowledge information to the user. Therefore, the present application breaks the semantic barrier between the library where the portrait tag is located and the library where the knowledge tag is located, integrates the library where the portrait tag is located and the library where the knowledge tag is located, and can be flexibly applied to multiple business fields, so as to facilitate the recommendation of knowledge tags related to the portrait to the user with strong pertinence according to the user's portrait tag, and solves the technical problem that the independent tag system has low pertinence in recommending knowledge information to the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0079] Figure 1 A flowchart of a knowledge information recommendation method provided in an embodiment of the present application;

[0080] Figure 2 A flowchart of another knowledge information recommendation method provided in an embodiment of the present application;

[0081] Figure 3 A schematic diagram of a scenario of a knowledge information recommendation method provided in an embodiment of the present application;

[0082] Figure 4 A schematic diagram of another scenario of a knowledge information recommendation method provided in an embodiment of the present application;

[0083] Figure 5 A schematic diagram of the structure of a knowledge information recommendation device provided in an embodiment of the present application;

[0084] Figure 6 A schematic diagram of the structure of another knowledge information recommendation device provided in an embodiment of the present application;

[0085] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0086] Figure 8 A block diagram of an electronic device provided in an embodiment of the present application.

[0087] The above drawings show clear embodiments of the present disclosure, which will be described in more detail below. These drawings and text descriptions are not intended to limit the scope of the present disclosure in any way, but to illustrate the concepts of the present disclosure to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0088] Here, exemplary embodiments will be described in detail, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present disclosure.

[0089] In one example, in actual applications, after a label system is established based on the knowledge base itself, knowledge information can be recommended to users based on the label system. In the prior art, when recommending knowledge information to users based on the label system, different business fields will establish a label system based on their own knowledge bases, and then recommend knowledge information to users based on their own label systems. However, in the prior art, since the label systems of different business fields are independent, the label system of each business field can only serve its own business, which will result in less knowledge information recommended to users by the label system.

[0090] The knowledge information recommendation method, device and equipment provided in this application are intended to solve the above technical problems in the prior art.

[0091] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0092] Figure 1 A flowchart of a knowledge information recommendation method provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0093] 101. Get the portrait tag of the current user.

[0094] For example, the execution subject of this embodiment may be an electronic device, or a terminal device, or a knowledge information recommendation device or device, or other devices or equipment that can execute this embodiment, and there is no limitation to this. In this embodiment, the execution subject is introduced as an electronic device.

[0095] First, you need to obtain the current user's portrait tag. When a user registers the system, they need to fill in their personal information, which may include information such as portrait tags. When the user logs into the system, they can obtain the current user's portrait tag. Alternatively, if the personal information does not include a portrait tag, the electronic device can analyze the personal information to obtain a portrait tag corresponding to the personal information, and there is no restriction on this.

[0096] 102. Search for a knowledge tag that matches the portrait tag in a preset tag library; wherein the tag library includes a correspondence between the portrait tag and the knowledge tag, and a correspondence between the knowledge tag and the knowledge information.

[0097] Exemplarily, the preset tag library includes multiple portrait tags and knowledge tags, and there is a corresponding relationship between the portrait tags and the knowledge tags. Therefore, the electronic device can search for the knowledge tag that matches the portrait tag in the preset tag library based on the matching relationship between the portrait tag and the knowledge tag, and then determine the knowledge tag that matches the portrait tag.

[0098] 103. According to the correspondence between the knowledge tag and the knowledge information, determine the knowledge information to be recommended corresponding to the knowledge tag, and push the knowledge information to the user.

[0099] Exemplarily, the preset tag library includes the correspondence between knowledge tags and knowledge information, so the electronic device can search for the knowledge information to be recommended corresponding to the knowledge tags in the tag library according to the correspondence between the knowledge tags and the knowledge information, and then push the knowledge information to the user.

[0100] In the embodiment of the present application, the portrait tag of the current user is obtained. The knowledge tag matching the portrait tag is searched in the preset tag library; wherein the tag library includes the corresponding relationship between the portrait tag and the knowledge tag, and the corresponding relationship between the knowledge tag and the knowledge information. According to the corresponding relationship between the knowledge tag and the knowledge information, the knowledge information to be recommended corresponding to the knowledge tag is determined, and the knowledge information is pushed to the user. In this scheme, since the preset tag library includes the matching relationship between the portrait tag and the knowledge tag, and the corresponding relationship between the knowledge tag and the knowledge information, the knowledge tag matching the portrait tag can be searched in the tag library according to the matching relationship between the portrait tag and the knowledge tag, and then the knowledge information to be recommended corresponding to the knowledge tag is determined in the tag library according to the corresponding relationship between the knowledge tag and the knowledge information, and the knowledge information is pushed to the user. Therefore, the present application breaks the semantic barrier between the library where the portrait tag is located and the library where the knowledge tag is located, and integrates the library where the portrait tag is located with the library where the knowledge tag is located, which can be flexibly applied to multiple business fields, so as to facilitate the recommendation of knowledge tags related to the portrait to the user in a more targeted manner according to the user's portrait tag, and solves the technical problem that the independent tag system recommends knowledge information to the user with low targeting.

[0101] Figure 2 A flowchart of another knowledge information recommendation method provided in an embodiment of the present application is shown as follows: Figure 2 As shown, the method includes:

[0102] 201. Obtain multiple portrait tags, knowledge tags and knowledge information.

[0103] In one example, knowledge labels include core knowledge labels and generalized knowledge labels; obtaining multiple core knowledge labels includes: obtaining multiple first knowledge texts and second knowledge texts to be detected; performing word segmentation and screening on the first knowledge text to obtain multiple text words; annotating the text words according to a preset annotation type to obtain annotated text words; training a neural network model based on the annotated text words to obtain a label classification model; and detecting the second knowledge text through the label classification model to obtain multiple core knowledge labels.

[0104] In one example, multiple generalized knowledge labels are obtained, including: determining the core distribution data of the core knowledge label in each second knowledge text; randomly assigning a core knowledge label to each text word, and determining the text word distribution data of each text word; each core knowledge label corresponds to a core number; using the Gibbs sampling formula to determine and update the new core number corresponding to the text word, and updating the new core number to the text word number of the text word until Gibbs sampling converges; according to the text word number of the text word, determining the core knowledge label to which each text word belongs, and using the text word corresponding to the core knowledge label as the generalized knowledge label corresponding to the core knowledge label.

[0105] Exemplarily, the portrait tag indicates the user's identity and identity-related information, for example, the identities include: students, white-collar workers, high-net-worth customers, etc., and the identity-related information includes wealth planning, asset isolation, etc. Knowledge tags include multiple types of tags, for example, the types of knowledge tags include: location, product, card, activity, etc. Knowledge information indicates information related to the knowledge tag, for example: knowledge information includes: open-end financial products related to the product type knowledge tag, value-added service packages, private banking appointment gifts, etc. related to the activity type knowledge tag.

[0106] For example, knowledge tags include core knowledge tags and generalized knowledge tags. When an electronic device obtains multiple core knowledge tags, it needs to first obtain multiple first knowledge texts and second knowledge texts to be detected, and perform word segmentation and screening on the first knowledge texts, specifically screening out stop words. Stop words are roughly divided into two categories. One category is functional words contained in the language. Compared with other words, functional words have no actual meaning, such as this, that, in, he, him, etc. Another category of words includes vocabulary words, such as think, do, optimize, etc. These words will be removed from the text during screening, thereby improving the accuracy of label extraction; in addition, regular expressions are used to match punctuation, spaces, line breaks, and special characters, and punctuation, spaces, line breaks, and special characters are filtered out to obtain multiple remaining text words, and then the text words are annotated according to the preset annotation type. The preset annotation type includes: annotating different text words with different letters, for example, using B- to mark the beginning of the knowledge label, I- to mark the rest of the knowledge label, and O- to mark other words except the knowledge label, to obtain annotated text words, and train the neural network model according to the annotated text words. During the training process, the nn.LSTM function of the pytorch open source package can also be used to adjust the parameters of the neural network model to obtain a label classification model, wherein PyTorch is an open source Python machine learning library based on Torch for applications such as natural language processing. Finally, the second knowledge text is detected through the label classification model to obtain multiple core knowledge labels. Then the pytorch open source package calls the conditional random field algorithm (CRF function) to process multiple categories of core knowledge labels. Each classification result has a score for the category it belongs to, and human intervention and tuning are performed to filter out some unreasonable classifications. At this time, multiple filtered core knowledge labels are obtained, and the filtered multiple core knowledge labels can form a core knowledge label set.

[0107] When the electronic device obtains multiple generalized knowledge tags, it is necessary to first determine the core distribution data of the core knowledge tags in any knowledge text. Any knowledge can be each second knowledge text. The core distribution data θd of the core knowledge tags is: Among them, α is the hyperparameter of the distribution, which is a K-dimensional vector. K represents the number of core knowledge tags. α is assigned a random value when initialized. Then a core knowledge tag is randomly assigned to each text word in the corpus. Each core knowledge tag generates a corresponding core number. Each text word in the corpus is a word other than the knowledge tag marked by O-in the above text, and the text word distribution data βk of each text word is determined: μ is a hyperparameter of the distribution, which is a V-dimensional vector. V represents the number of all text words in the vocabulary. μ is initialized to a random value. Finally, the corpus is rescanned. For each text word, the Gibbs sampling formula is used to determine and update the new core knowledge label corresponding to the text word, that is, to update the new core number corresponding to the text word, and update the new core number to the text word number of the text word until Gibbs sampling converges. According to the text word number of the text word, the core knowledge label to which each text word belongs is determined. The distribution of the core knowledge labels in the second knowledge text can be obtained. The text word corresponding to the core knowledge label can be used as the generalized knowledge label corresponding to the core knowledge label. The ternary method can also be used.<w1,r,w2> Describe and store core knowledge labels and generalized knowledge labels. w1 represents the core knowledge label, i.e., entity word, w2 represents the generalized knowledge label corresponding to the core knowledge label, and r represents the relationship between the core knowledge label and the generalized knowledge label. Therefore, more generalized knowledge labels related to the core knowledge label are excavated, which greatly enriches the generalization ability of entity words and can be better integrated with other label systems.

[0108] 202. Analyze and process each sentence in the second knowledge text according to the dependency syntax method to obtain the grammatical relationship between the knowledge label and the knowledge information in each sentence.

[0109] Exemplarily, the electronic device analyzes and processes each sentence in the second knowledge text according to the dependency syntax method, first obtains the core knowledge label and knowledge information in each sentence, and then obtains the grammatical relationship between the core knowledge label and the knowledge information. And / or, the electronic device analyzes and processes each sentence in the second knowledge text according to the dependency syntax method, first obtains the generalized knowledge label and knowledge information in each sentence, and then obtains the grammatical relationship between the generalized knowledge label and the knowledge information. And / or, since the core knowledge label corresponds to the generalized knowledge label, the generalized knowledge label and the core knowledge label can be regarded as knowledge labels. The electronic device analyzes and processes each sentence in the second knowledge text according to the dependency syntax method, first obtains the core knowledge label, the generalized knowledge label, and the knowledge information in each sentence, and then obtains the grammatical relationship between the core knowledge label and the knowledge information, and the grammatical relationship between the generalized knowledge label and the knowledge information. 203. According to the grammatical relationship between the knowledge label and the knowledge information, establish and store the first corresponding relationship between the knowledge label and the knowledge information; wherein, multiple groups of knowledge labels and knowledge information generate a knowledge label map.

[0110] Exemplarily, the electronic device can establish and store the first correspondence between the knowledge tag and the knowledge information according to the grammatical relationship between the knowledge tag and the knowledge information, and can also generate a knowledge tag map according to multiple sets of knowledge tags and knowledge information. For example, after being analyzed and processed by the dependency syntax method, the result is converted into a map format:

[0111] {leather shoe manufacturing enterprise, mortgage, quick loan, subject-predicate relationship, verb-object relationship},

[0112] {Auto Sales Co., Ltd., recommended, inclusive financial products, referral relationship}

[0113] Among them, knowledge labels include leather shoe manufacturers and automobile sales companies, and knowledge information includes quick loans and inclusive financial products. The electronic device can establish and store the first correspondence between knowledge labels and knowledge information based on the subject-object relationship between leather shoe manufacturers and quick loans, and the preposition-object relationship between automobile sales companies and inclusive financial products, and generate a knowledge label map based on multiple sets of knowledge labels and knowledge information.

[0114] The storage strategy for storing knowledge tag graphs is as follows:

[0115] <node id, node name, relationship, related node id, related node name, attribute, distance>. Among them, node id is the unique code of knowledge label, node name is knowledge label, relationship is grammatical relationship, related node id is the unique code of knowledge information, related node name is knowledge information, attribute is used to distinguish different label types (such as location, product, card, activity, etc.), distance indicates the similarity between nodes, and the default value of distance is null. Therefore, each relationship is stored as a piece of data. Each time the relationship between nodes is updated, two pieces of data need to be updated. Although the update resource consumption increases, the query efficiency is improved.

[0116] 204. Extract knowledge tags from the knowledge tag graph through a preset tag classification model, convert the extracted multiple knowledge tags into knowledge tag vectors, and convert the multiple portrait tags into portrait tag vectors.

[0117] Exemplarily, the electronic device inputs the knowledge tag map into a preset tag classification model, and can extract multiple knowledge tags in the knowledge tag map, the multiple knowledge tags including core knowledge tags and generalized knowledge tags, and then converts the multiple knowledge tags into knowledge tag vectors, and also converts the multiple portrait tags into portrait tag vectors.

[0118] 205. Embed the portrait label vector and the knowledge label vector into the same matrix, and use the preset cosine similarity method to calculate the matrix to obtain the semantic similarity score between each portrait label vector and each knowledge label vector.

[0119] Exemplarily, the electronic device initializes the portrait label vector and the knowledge label vector and embeds them into the same matrix Z, and uses the preset cosine similarity method cosine to calculate the matrix Z to obtain the structural similarity matrix Ms between each portrait label vector and each knowledge label vector. In this matrix, the rows represent knowledge labels, the columns represent customer portrait labels, and the values ​​in the matrix represent the similarity scores between the two. The embedding of the names of all labels can be defined as a matrix N, and the cosine similarity is used to capture the degree of similarity between entities. The semantic similarity matrix is ​​represented as Mn. After 3 iterative calculations, the relationship combination with the maximum similarity of the labels is obtained.

[0120] 206. Determine the association value between each knowledge label vector and each portrait label vector.

[0121] In one example, a first probability that a knowledge label vector and a portrait label vector appear simultaneously in multiple second knowledge texts, and a second probability that the knowledge label vector and the portrait label vector appear separately in multiple second knowledge texts are determined; a first expectation corresponding to the first probability and a second expectation corresponding to the second probability are determined, and an association value between each knowledge label vector and each portrait label vector is determined based on the first expectation and the second expectation.

[0122] Exemplarily, the electronic device may first determine a first probability that the knowledge tag vector Tag and the portrait tag vector Label both appear simultaneously in a plurality of second knowledge texts, the first probability being expressed as P(x, y), the plurality of second knowledge texts being texts stored in a text library, and a second probability P(x) that the knowledge tag vector appears alone in the plurality of second knowledge texts, and a second probability P(y) that the portrait tag vector Label appears alone in the plurality of second knowledge texts, and then calculate the mutual information between the knowledge tag vector Tag and the portrait tag vector Label according to a preset formula (1), the preset formula (1) being as follows:

[0123]

[0124] Among them, x represents the knowledge label vector Tag, y represents the image label vector Label, I(X;Y) represents the mutual information between the knowledge label vector Tag and the image label vector Label, and the mutual information represents the association information between Tag and Label.

[0125] Then determine the first expected EP corresponding to the first probability P(x, y) [ g(x, y)], the first expectation represents the expectation that the knowledge tag vector Tag and the image tag vector Label appear in the same knowledge text, and the second expectation E corresponding to the second probability P(x) N , the second expectation represents the expectation that the knowledge tag vector Tag and the image tag vector Label do not appear in the same knowledge text E N [g(x′, y′)], and then the maximum mutual information between the knowledge tag vector Tag and the image tag vector Label can be calculated according to the preset formula (2), and the preset formula (2) is as follows:

[0126] MI(X, Y) ≥ E P [g(x,y)]-E N |[g(x′,y′)]

[0127] Among them, g(x, y) represents the result value of the binary classification function input by the vector product of Tag and Label, x represents a specific Tag, y represents a specific Label, g(x′, y′) represents the result value of the binary classification input by the vector product of a random Tag and a random Label, and the g(.) function uses the softmax function for binary classification. MI(X, Y) represents the association value between each knowledge label vector and each portrait label vector, and then the association value greater than or equal to the preset value can be taken to obtain the knowledge label vector and portrait label vector corresponding to the association value.

[0128] 207. According to the correspondence between each portrait label vector, each knowledge label vector, and the semantic similarity, as well as the correspondence between each portrait label vector, each knowledge label vector, and the associated value, the product of the semantic similarity score and the associated value corresponding to each knowledge label vector and each portrait label vector is calculated and obtained.

[0129] Exemplarily, based on the correspondence between each portrait label vector, each knowledge label vector, and the semantic similarity score, as well as the correspondence between each portrait label vector, each knowledge label vector, and the association value, the electronic device can multiply the association value greater than or equal to a preset value by the semantic similarity score corresponding to the association value to obtain the product, and finally obtain the numerical value of the relationship between the portrait label vector and the knowledge label vector.

[0130] 208. According to the product, determine and establish a second corresponding relationship between the knowledge label corresponding to each knowledge label vector and the image label corresponding to each image label vector.

[0131] Exemplarily, the electronic device determines the corresponding relationship between each knowledge label vector and each portrait label vector based on the product, and further determines the second corresponding relationship between each knowledge label and each portrait label.

[0132] 209. Establish a label library according to the first corresponding relationship and the second corresponding relationship.

[0133] Exemplarily, the electronic device establishes a label library according to the first corresponding relationship and the second corresponding relationship. For example, the label library stores a knowledge label set, a portrait label set, and a knowledge information set.

[0134] 210. Get the portrait tag of the current user.

[0135] For example, this step can be referred to Figure 1 Step 101 in the above is not described in detail.

[0136] 211. According to the preset correspondence between the portrait tag and the knowledge tag, search for the knowledge tag matching the portrait tag in the tag library; wherein the tag library includes the correspondence between the portrait tag and the knowledge tag, and the correspondence between the knowledge tag and the knowledge information.

[0137] Exemplarily, the knowledge tag set in the preset tag library includes multiple knowledge tags, the portrait tag set includes multiple portrait tags, and there is a corresponding relationship between the portrait tags and the knowledge tags. Therefore, the electronic device can search for the knowledge tag that matches the portrait tag in the knowledge tag set based on the matching relationship between the portrait tag and the knowledge tag, and then determine the knowledge tag that matches the portrait tag.

[0138] 212. According to the correspondence between the knowledge tag and the knowledge information, determine the knowledge information to be recommended corresponding to the knowledge tag, and push the knowledge information to the user.

[0139] Exemplarily, the knowledge information set in the preset tag library includes multiple knowledge information, and there is a corresponding relationship between the knowledge information and the knowledge tags. Therefore, the electronic device can search for the knowledge information to be recommended corresponding to the knowledge tag in the knowledge information set according to the corresponding relationship between the knowledge tags and the knowledge information, and then push the knowledge information to the user. Figure 3 As shown, Figure 3 A scenario diagram of a knowledge information recommendation method provided in this application includes: a knowledge tag system and a customer portrait tag. The knowledge tag system includes multiple knowledge tags. The customer portrait tag is a user portrait tag, such as Figure 4 As shown, Figure 4 A scenario diagram of another knowledge information recommendation method provided for this application includes: customer portrait tags, tag alignment system (i.e., a preset tag library), tag recommendations, knowledge recommendations, product recommendations, and activity recommendations.

[0140] In an embodiment of the present application, multiple portrait tags, knowledge tags, and knowledge information are obtained. Each sentence in the second knowledge text is analyzed and processed according to the dependency syntax method to obtain the grammatical relationship between the knowledge tag and the knowledge information in each sentence. According to the grammatical relationship between the knowledge tag and the knowledge information, a first corresponding relationship between the knowledge tag and the knowledge information is established and stored; wherein, multiple groups of knowledge tags and knowledge information generate a knowledge tag map. The knowledge tags in the knowledge tag map are extracted by a preset tag classification model, and the extracted multiple knowledge tags are converted into knowledge tag vectors, and multiple portrait tags are converted into portrait tag vectors. The portrait tag vector and the knowledge tag vector are embedded in the same matrix, and the matrix is ​​calculated using a preset cosine similarity method to obtain the semantic similarity score between each portrait tag vector and each knowledge tag vector. The association value between each knowledge tag vector and each portrait tag vector is determined. According to the correspondence between each portrait tag vector, each knowledge tag vector, and the semantic similarity, and the correspondence between each portrait tag vector, each knowledge tag vector, and the association value, the product of the semantic similarity score and the association value corresponding to each knowledge tag vector and each portrait tag vector is calculated and obtained. According to the product, determine and establish a second correspondence between the knowledge tag corresponding to each knowledge tag vector and the portrait tag corresponding to each portrait tag vector. Establish a label library based on the first correspondence and the second correspondence. Get the portrait tag of the current user. According to the preset correspondence between the portrait tag and the knowledge tag, search the label library for the knowledge tag that matches the portrait tag; wherein the label library includes the correspondence between the portrait tag and the knowledge tag, and the correspondence between the knowledge tag and the knowledge information. According to the correspondence between the knowledge tag and the knowledge information, determine the knowledge information to be recommended corresponding to the knowledge tag, and push the knowledge information to the user. Therefore, the present application breaks the semantic barrier between the library where the portrait tag is located and the library where the knowledge tag is located, and integrates the library where the portrait tag is located with the library where the knowledge tag is located, which can be flexibly applied to multiple business fields, so as to facilitate the recommendation of knowledge tags related to the portrait to users in a more targeted manner based on the user's portrait tag, and solves the technical problem that the independent label system has low targeting in recommending knowledge information to users.

[0141] Figure 5 A schematic diagram of the structure of a knowledge information recommendation device provided in an embodiment of the present application is shown in FIG. Figure 5 As shown, the device comprises:

[0142] The first acquisition unit 51 is used to acquire the portrait tag of the current user.

[0143] The search unit 52 is used to search for a knowledge tag matching the portrait tag in a preset tag library; wherein the tag library includes the correspondence between the portrait tag and the knowledge tag, and the correspondence between the knowledge tag and the knowledge information.

[0144] The determination unit 53 is used to determine the knowledge information to be recommended corresponding to the knowledge tag according to the corresponding relationship between the knowledge tag and the knowledge information.

[0145] The push unit 54 is used to push knowledge information to the user.

[0146] The device of this embodiment can execute the technical solution in the above method. Its specific implementation process and technical principles are the same and will not be repeated here.

[0147] Figure 6 A schematic diagram of the structure of another knowledge information recommendation device provided in an embodiment of the present application, in Figure 5 Based on the embodiment shown, Figure 6 As shown, the search unit 52 is specifically used for:

[0148] According to the correspondence between the preset portrait tags and knowledge tags, search for knowledge tags that match the portrait tags in the tag library.

[0149] In one example, the device further includes:

[0150] The second acquisition unit 61 is used to acquire multiple knowledge tags.

[0151] The third acquisition unit 62 is used to acquire multiple portrait tags and knowledge information.

[0152] The first establishing unit 63 is used to establish a first corresponding relationship between the knowledge tag and the knowledge information.

[0153] The second establishing unit 64 is used to establish a second corresponding relationship between the portrait label and the knowledge label.

[0154] The third establishing unit 65 is used to establish a label library according to the first corresponding relationship and the second corresponding relationship.

[0155] In one example, the knowledge tags include core knowledge tags and generalized knowledge tags; the second acquisition unit 61 includes:

[0156] The acquisition module 611 is used to acquire a plurality of first knowledge texts and second knowledge texts to be detected.

[0157] The processing module 612 is used to perform word segmentation and screening on the first knowledge text to obtain multiple text words.

[0158] The marking module 613 is used to mark the text words according to the preset marking type to obtain marked text words.

[0159] The training module 614 is used to train the neural network model according to the annotated text words to obtain a label classification model.

[0160] The detection module 615 is used to detect the second knowledge text through a label classification model to obtain multiple core knowledge labels.

[0161] In one example, the second obtaining unit 61 includes:

[0162] The core distribution determination module 616 is used to determine the core distribution data of the core knowledge tags in each second knowledge text.

[0163] The module 617 for determining text word distribution is used to randomly assign a core knowledge label to each text word and determine the text word distribution data of each text word; each core knowledge label corresponds to a core number.

[0164] The text word number update module 618 is used to determine and update the new core number corresponding to the text word using the Gibbs sampling formula, and update the new core number to the text word number of the text word until the Gibbs sampling converges.

[0165] The generalized knowledge label determination module 619 is used to determine the core knowledge label to which each text word belongs according to the text word number of the text word, and use the text word corresponding to the core knowledge label as the generalized knowledge label corresponding to the core knowledge label.

[0166] In one example, the first establishing unit 63 includes:

[0167] The analysis module 631 is used to analyze and process each sentence in the second knowledge text according to the dependency syntax method to obtain the grammatical relationship between the knowledge label and the knowledge information in each sentence.

[0168] The first establishing module 632 is used to establish and store a first corresponding relationship between knowledge tags and knowledge information according to the grammatical relationship between knowledge tags and knowledge information; wherein, multiple groups of knowledge tags and knowledge information generate a knowledge tag graph.

[0169] In one example, the second establishing unit 64 includes:

[0170] The conversion module 641 is used to extract knowledge tags in the knowledge tag graph through a preset tag classification model, convert the extracted multiple knowledge tags into knowledge tag vectors, and convert multiple portrait tags into portrait tag vectors.

[0171] The score calculation module 642 is used to embed the portrait label vector and the knowledge label vector into the same matrix, and calculate the matrix using a preset cosine similarity method to obtain the semantic similarity score between each portrait label vector and each knowledge label vector.

[0172] The association value determination module 643 is used to determine the association value between each knowledge label vector and each portrait label vector.

[0173] The product calculation module 644 is used to calculate and obtain the product of the semantic similarity score and the associated value corresponding to each knowledge label vector and each portrait label vector based on the correspondence between each portrait label vector, each knowledge label vector, and the semantic similarity score, as well as the correspondence between each portrait label vector, each knowledge label vector, and the associated value.

[0174] The second establishing module 645 is used to determine and establish a second corresponding relationship between the knowledge label corresponding to each knowledge label vector and the image label corresponding to each image label vector based on the product.

[0175] In one example, the determining association value module 643 includes:

[0176] The determination probability submodule 6431 is used to determine the first probability that the knowledge label vector and the portrait label vector both appear in multiple second knowledge texts at the same time, and the second probability that the knowledge label vector and the portrait label vector both appear separately in multiple second knowledge texts.

[0177] The association value determination submodule 6432 is used to determine a first expectation corresponding to the first probability and a second expectation corresponding to the second probability, and determine the association value between each knowledge label vector and each portrait label vector based on the first expectation and the second expectation.

[0178] The device of this embodiment can execute the technical solution in the above method. Its specific implementation process and technical principles are the same and will not be repeated here.

[0179] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the electronic device includes: a memory 71 and a processor 72.

[0180] The memory 71 stores a computer program that can be executed on the processor 72 .

[0181] The processor 72 is configured to execute the method provided in the above embodiment.

[0182] The electronic device further includes a receiver 73 and a transmitter 74. The receiver 73 is used to receive instructions and data sent by an external device, and the transmitter 74 is used to send instructions and data to the external device.

[0183] Figure 8 This is a block diagram of an electronic device provided in an embodiment of the present application. The electronic device may be a mobile phone, a computer, a digital broadcast terminal, a message transceiver, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0184] The device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0185] The processing component 802 generally controls the overall operation of the device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0186] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0187] The power supply component 806 provides power to the various components of the device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 800.

[0188] The multimedia component 808 includes a screen that provides an output interface between the device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0189] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the device 800 is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0190] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0191] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the device 800. For example, the sensor assembly 814 can detect the open / closed state of the device 800, the relative positioning of components, such as the display and keypad of the device 800, the sensor assembly 814 can also detect the position change of the device 800 or a component of the device 800, the presence or absence of user contact with the device 800, the orientation or acceleration / deceleration of the device 800 and the temperature change of the device 800. The sensor assembly 814 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 can also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 can also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor or a temperature sensor.

[0192] The communication component 816 is configured to facilitate wired or wireless communication between the device 800 and other devices. The device 800 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0193] In an exemplary embodiment, the apparatus 800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0194] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by the processor 820 of the device 800 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0195] An embodiment of the present application also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the method provided by the above embodiment.

[0196] An embodiment of the present application also provides a computer program product, which includes: a computer program, which is stored in a readable storage medium, and at least one processor of an electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device executes the solution provided by any of the above embodiments.

[0197] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0198] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A knowledge information recommendation method, characterized in that: include: Get the current user's portrait tag; Searching for a knowledge tag matching the portrait tag in a preset tag library; wherein the tag library includes a correspondence between the portrait tag and the knowledge tag, and a correspondence between the knowledge tag and the knowledge information; According to the correspondence between the knowledge tag and the knowledge information, determine the knowledge information to be recommended corresponding to the knowledge tag, and push the knowledge information to the user; The method further comprises: Obtain multiple portrait tags, knowledge tags and knowledge information; Establishing a first correspondence between the knowledge tag and the knowledge information, establishing a second correspondence between the image tag and the knowledge tag, and establishing a tag library according to the first correspondence and the second correspondence; The knowledge tags include core knowledge tags and generalized knowledge tags; obtaining multiple core knowledge tags includes: Acquire a plurality of first knowledge texts and second knowledge texts to be detected; Performing word segmentation and screening processing on the first knowledge text to obtain a plurality of text words; Annotating the text words according to a preset annotation type to obtain annotated text words; The neural network model is trained according to the annotated text words to obtain a label classification model; The second knowledge text is detected by a label classification model to obtain a plurality of core knowledge labels; Obtain multiple generalized knowledge tags, including: Determine the core distribution data of the core knowledge tags in each second knowledge text; Randomly assign a core knowledge label to each text word, and determine the text word distribution data of each text word; each of the core knowledge labels corresponds to a core number; Determine and update a new core number corresponding to the text word using a Gibbs sampling formula, and update the new core number to the text word number of the text word until Gibbs sampling converges; Determine the core knowledge label to which each text word belongs according to the text word number of the text word, and use the text word corresponding to the core knowledge label as the generalized knowledge label corresponding to the core knowledge label; Establishing a first corresponding relationship between the knowledge tag and the knowledge information includes: Analyze and process each sentence in the second knowledge text according to the dependency syntax method to obtain the grammatical relationship between the knowledge label and the knowledge information in each sentence; According to the grammatical relationship between the knowledge label and the knowledge information, a first corresponding relationship between the knowledge label and the knowledge information is established and stored; wherein a plurality of groups of knowledge labels and knowledge information generate a knowledge label graph; Establishing a second corresponding relationship between the portrait label and the knowledge label includes: Extracting knowledge tags in the knowledge tag graph through a preset tag classification model, converting the extracted multiple knowledge tags into knowledge tag vectors, and converting the multiple portrait tags into portrait tag vectors; The portrait label vector and the knowledge label vector are embedded in the same matrix, and the matrix is ​​calculated using a preset cosine similarity method to obtain a semantic similarity score between each portrait label vector and each knowledge label vector; Determine the association value between each knowledge label vector and each image label vector; According to the correspondence between each portrait label vector, each knowledge label vector, and the semantic similarity score, and the correspondence between each portrait label vector, each knowledge label vector, and the association value, the product of the semantic similarity score corresponding to each knowledge label vector and each portrait label vector and the association value is calculated and obtained; According to the product, determining and establishing a second correspondence between the knowledge label corresponding to each knowledge label vector and the image label corresponding to each image label vector; Determine the association value between each knowledge label vector and each image label vector, including: Determine a first probability that the knowledge label vector and the portrait label vector both appear in the plurality of second knowledge texts at the same time, and a second probability that the knowledge label vector and the portrait label vector both appear separately in the plurality of second knowledge texts; A first expectation corresponding to the first probability and a second expectation corresponding to the second probability are determined, and an association value between each knowledge label vector and each portrait label vector is determined based on the first expectation and the second expectation.

2. The method according to claim 1, characterized in that Search the preset tag library for knowledge tags that match the portrait tag, including: According to the correspondence between the preset portrait tags and knowledge tags, the knowledge tags matching the portrait tags are searched in the tag library.

3. A knowledge information recommendation device, characterized in that: include: A first acquisition unit, used to acquire a portrait tag of the current user; A search unit, used to search for a knowledge tag matching the portrait tag in a preset tag library; wherein the tag library includes a correspondence between the portrait tag and the knowledge tag, and a correspondence between the knowledge tag and the knowledge information; A determination unit, configured to determine the knowledge information to be recommended corresponding to the knowledge tag according to the correspondence between the knowledge tag and the knowledge information; A push unit, used for pushing the knowledge information to the user; The device also includes: A second acquisition unit, used to acquire a plurality of knowledge tags; A third acquisition unit is used to acquire multiple portrait tags and knowledge information; A first establishing unit, used to establish a first corresponding relationship between the knowledge tag and the knowledge information; A second establishing unit, used to establish a second corresponding relationship between the portrait label and the knowledge label; A third establishing unit, configured to establish a label library according to the first corresponding relationship and the second corresponding relationship; The knowledge tags include core knowledge tags and generalized knowledge tags; the second acquisition unit includes: An acquisition module, used for acquiring a plurality of first knowledge texts and second knowledge texts to be detected; A processing module, used for performing word segmentation and screening on the first knowledge text to obtain a plurality of text words; A marking module, used to mark the text words according to a preset marking type to obtain marked text words; A training module is used to train the neural network model according to the annotated text words to obtain a label classification model; A detection module, used to detect the second knowledge text through a label classification model to obtain a plurality of core knowledge labels; The second acquisition unit includes: A core distribution determination module, used to determine core distribution data of core knowledge tags in each second knowledge text; A text word distribution determination module is used to randomly assign a core knowledge label to each text word and determine the text word distribution data of each text word; each of the core knowledge labels corresponds to a core number; A text word number updating module is used to determine and update a new core number corresponding to the text word by using a Gibbs sampling formula, and update the new core number to the text word number of the text word until the Gibbs sampling converges; A generalized knowledge label determination module is used to determine the core knowledge label to which each text word belongs according to the text word number of the text word, and use the text word corresponding to the core knowledge label as the generalized knowledge label corresponding to the core knowledge label; The first establishing unit comprises: An analysis module, used to analyze and process each sentence in the second knowledge text according to a dependency syntax method to obtain a grammatical relationship between the knowledge label and the knowledge information in each sentence; A first establishing module is used to establish and store a first corresponding relationship between the knowledge label and the knowledge information according to a grammatical relationship between the knowledge label and the knowledge information; wherein a plurality of groups of knowledge labels and knowledge information generate a knowledge label graph; The second establishing unit comprises: A conversion module, used to extract knowledge tags in the knowledge tag graph through a preset tag classification model, convert the extracted multiple knowledge tags into knowledge tag vectors, and convert the multiple portrait tags into portrait tag vectors; A score calculation module, used for embedding the portrait label vector and the knowledge label vector into the same matrix, and calculating the matrix using a preset cosine similarity method to obtain a semantic similarity score between each portrait label vector and each knowledge label vector; A correlation value determination module, used to determine the correlation value between each knowledge label vector and each portrait label vector; A product calculation module is used to calculate and obtain the product of the semantic similarity score and the association value corresponding to each knowledge label vector and each portrait label vector according to the corresponding relationship between each portrait label vector, each knowledge label vector, and the semantic similarity score, and the corresponding relationship between each portrait label vector, each knowledge label vector, and the association value; A second establishing module, used for determining and establishing a second corresponding relationship between the knowledge label corresponding to each knowledge label vector and the image label corresponding to each image label vector according to the product; The module for determining the associated value comprises: A determination probability submodule, used to determine a first probability that the knowledge label vector and the portrait label vector both appear in a plurality of second knowledge texts at the same time, and a second probability that the knowledge label vector and the portrait label vector both appear separately in a plurality of second knowledge texts; The association value determination submodule is used to determine a first expectation corresponding to the first probability and a second expectation corresponding to the second probability, and determine the association value between each knowledge label vector and each portrait label vector based on the first expectation and the second expectation.

4. The device according to claim 3, characterized in that The searching unit is specifically used for: According to the correspondence between the preset portrait tags and knowledge tags, the knowledge tags matching the portrait tags are searched in the tag library.

5. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the method described in claim 1 or 2 when executing the computer program.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to claim 1 or 2 when executed by a processor.

7. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to claim 1 or 2 when being executed by a processor.

Images