Artificial Intelligence-Based Intelligent Management Cloud Platform and Management Method

Abstract

The present invention discloses an intelligent management cloud platform and a management method based on artificial intelligence, which relates to the technical field of intelligent management. It performs semantic encoding on the police situation query input to obtain a police situation query input semantic encoding feature vector, extracts the text part from the first alternative police situation data to obtain the first alternative police situation text part, performs semantic encoding based on word granularity on the first alternative police situation text part to obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors, performs cross-granularity semantic matching on the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query - first alternative police situation semantic matching feature, and based on the police situation query - first alternative police situation semantic matching feature, determines whether the matching degree of the first alternative police situation data exceeds a predetermined threshold and determines whether to return the first alternative police situation data, which can improve the accuracy and efficiency of police situation query.

Description

Technical Field

[0001] This application relates to the field of intelligent management technology, and more specifically, to an intelligent management cloud platform and management method based on artificial intelligence. Background Art

[0002] With the continuous development of artificial intelligence technology, intelligent management cloud platforms have been widely used in various fields. Among them, intelligent security is an important field, which is related to public safety and social stability. Intelligent security refers to the use of artificial intelligence technology to improve the efficiency and effectiveness of security systems, such as realizing functions such as intelligent monitoring, intelligent alarm, and intelligent early warning through technologies such as face recognition, license plate recognition, and behavior analysis.

[0003] The police cloud alarm system is an application of intelligent security. By connecting alarm devices to the cloud platform, real-time monitoring, processing, and response to alarm information are realized. In the police cloud alarm system, police situation query is an important function. Users can query the alternative police situation data that matches it by inputting relevant police situation information. However, the traditional police situation query method is usually based on keyword matching, and there are problems such as inaccurate semantic understanding and inaccurate result matching.

[0004] Therefore, an optimized intelligent management cloud platform and management method based on artificial intelligence are expected to solve the above technical problems. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. This application provides an intelligent management cloud platform and management method based on artificial intelligence, which can screen out the police situation data that semantically matches the police situation query input from the alternative police situation data according to the user's police situation query input, and return it to the user, thereby improving the accuracy and efficiency of police situation query and providing better services and support for users.

[0006] According to one aspect of this application, an intelligent management cloud platform based on artificial intelligence is provided, which includes:

[0007] A police situation query input module for obtaining a police situation query input;

[0008] A police situation query input semantic encoding module for semantically encoding the police situation query input to obtain a police situation query input semantic encoding feature vector;

[0009] A first alternative police situation data acquisition module for obtaining first alternative police situation data;

[0010] A first alternative police situation text part extraction module for extracting the text part from the first alternative police situation data to obtain a first alternative police situation text part;

[0011] The first alternative police situation word granularity semantic encoding module is used to perform word granularity-based semantic encoding on the first alternative police situation text part to obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors;

[0012] The police situation query input - first alternative police situation semantic matching module is used to perform cross-granularity semantic matching on the police situation query input semantic encoding feature vector and the sequence of the first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature;

[0013] The first alternative police situation data matching detection and return module is used to determine whether the matching degree of the first alternative police situation data exceeds a predetermined threshold based on the police situation query input - first alternative police situation semantic matching feature, and determine whether to return the first alternative police situation data.

[0014] In the above intelligent management cloud platform based on artificial intelligence, the first alternative police situation word granularity semantic encoding module is used for:

[0015] After performing word segmentation on the first alternative police situation text part, use a semantic encoder including a word embedding layer to obtain a sequence of the first alternative police situation word granularity semantic encoding feature vectors.

[0016] In the above intelligent management cloud platform based on artificial intelligence, the police situation query input - first alternative police situation semantic matching module is used for:

[0017] Use a cross-granularity semantic matching module to calculate the semantic matching result between the police situation query input semantic encoding feature vector and the sequence of the first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature vector as the police situation query input - first alternative police situation semantic matching feature.

[0018] In the above intelligent management cloud platform based on artificial intelligence, the police situation query input - first alternative police situation semantic matching module is used for:

[0019] Use the cross-granularity semantic matching module to calculate the semantic matching result between the police situation query input semantic encoding feature vector and the sequence of the first alternative police situation word granularity semantic encoding feature vectors with the following semantic matching formula to obtain the police situation query input - first alternative police situation semantic matching feature vector;

[0020] Among them, the semantic matching formula is:

[0021]

[0022] Among them, v represents the police situation query input semantic encoding feature vector, A represents 1×N wmatrix, N w equals the scale of the semantic encoding feature vector of the police situation query input, and B is a 1×N h matrix, N h equals the number of the first alternative police situation word granularity semantic encoding feature vectors in the sequence of the first alternative police situation word granularity semantic encoding feature vectors, σ is the Sigmoid function, s is the weight coefficient, M w and M h represent the convolution operation of a 1×1 convolution kernel, h i represents the i-th first alternative police situation word granularity semantic encoding feature vector in the sequence of the first alternative police situation word granularity semantic encoding feature vectors, N represents the scale of each first alternative police situation word granularity semantic encoding feature vector in the sequence of the first alternative police situation word granularity semantic encoding feature vectors, and v′ represents the police situation query input - first alternative police situation semantic matching feature vector.

[0023] In the above-mentioned intelligent management cloud platform based on artificial intelligence, the first alternative police situation data matching detection return module includes:

[0024] A classification unit, configured to pass the police situation query input - first alternative police situation semantic matching feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the matching degree of the first alternative police situation data exceeds a predetermined threshold;

[0025] A police situation data return unit, configured to determine whether to return the first alternative police situation data based on the classification result.

[0026] In the above-mentioned intelligent management cloud platform based on artificial intelligence, it further includes: a training module, configured to train the semantic encoder including a word embedding layer, the cross-granularity semantic matching module, and the classifier.

[0027] In the above-mentioned intelligent management cloud platform based on artificial intelligence, the training module includes:

[0028] A training police situation query input acquisition unit, configured to acquire a training police situation query input;

[0029] A training police situation query input semantic encoding unit, configured to perform semantic encoding on the training police situation query input to obtain a training police situation query input semantic encoding feature vector;

[0030] A training first alternative police situation data acquisition unit, configured to acquire training first alternative police situation data;

[0031] A training first alternative police situation text part extraction unit, configured to extract a text part from the training first alternative police situation data to obtain a training first alternative police situation text part;

[0032] Train the first alternative police situation word granularity semantic encoding unit to perform word segmentation on the training first alternative police situation text part and then obtain a sequence of training first alternative police situation word granularity semantic encoding feature vectors through a semantic encoder including a word embedding layer;

[0033] Train the police situation query input - first alternative police situation semantic matching unit to use a cross-granularity semantic matching module to calculate the semantic matching result between the training police situation query input semantic encoding feature vector and the sequence of the training first alternative police situation word granularity semantic encoding feature vectors to obtain a training police situation query input - first alternative police situation semantic matching feature vector;

[0034] The optimization unit is used to optimize the training police situation query input - first alternative police situation semantic matching feature vector to obtain an optimized training police situation query input - first alternative police situation semantic matching feature vector;

[0035] The classification loss unit is used to pass the optimized training police situation query input - first alternative police situation semantic matching feature vector through a classifier to obtain a classification loss function value;

[0036] The iterative training unit is used to train the semantic encoder including the word embedding layer, the cross-granularity semantic matching module and the classifier based on the classification loss function value.

[0037] According to another aspect of the present application, an intelligent management method based on artificial intelligence is provided, which includes:

[0038] Obtain a police situation query input;

[0039] Perform semantic encoding on the police situation query input to obtain a police situation query input semantic encoding feature vector;

[0040] Obtain the first alternative police situation data;

[0041] Extract the text part from the first alternative police situation data to obtain the first alternative police situation text part;

[0042] Perform word granularity-based semantic encoding on the first alternative police situation text part to obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors;

[0043] Perform cross-granularity semantic matching on the police situation query input semantic encoding feature vector and the sequence of the first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature;

[0044] Based on the police situation query input - first alternative police situation semantic matching feature, determine whether the matching degree of the first alternative police situation data exceeds a predetermined threshold and determine whether to return the first alternative police situation data.

[0045] In the above artificial intelligence-based intelligent management method, performing semantic encoding on the first alternative police situation text part at the word granularity to obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors, including:

[0046] Performing word segmentation on the first alternative police situation text part and then passing it through a semantic encoder including a word embedding layer to obtain the sequence of first alternative police situation word granularity semantic encoding feature vectors.

[0047] In the above artificial intelligence-based intelligent management method, performing cross-granularity semantic matching on the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature, including:

[0048] Using a cross-granularity semantic matching module to calculate the semantic matching result between the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature vector as the police situation query input - first alternative police situation semantic matching feature.

[0049] Compared with the prior art, the artificial intelligence-based intelligent management cloud platform and management method provided by the present application first perform semantic encoding on the police situation query input to obtain a police situation query input semantic encoding feature vector, then extract a text part from the first alternative police situation data to obtain a first alternative police situation text part, then perform semantic encoding on the first alternative police situation text part at the word granularity to obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors, then perform cross-granularity semantic matching on the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature, and finally, based on the police situation query input - first alternative police situation semantic matching feature, determine whether the matching degree of the first alternative police situation data exceeds a predetermined threshold and determine whether to return the first alternative police situation data, which can improve the accuracy and efficiency of police situation query. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. The following drawings are not deliberately drawn to scale in actual size, and the focus is on showing the gist of the present application.

[0051] Figure 1 It is a block diagram schematic diagram of an artificial intelligence-based intelligent management cloud platform according to an embodiment of the present application.

[0052] Figure 2Flowchart of an intelligent management method based on artificial intelligence according to an embodiment of the present application.

[0053] Figure 3 Schematic diagram of the system architecture of an intelligent management method based on artificial intelligence according to an embodiment of the present application.

[0054] Figure 4 Application scenario diagram of an intelligent management cloud platform based on artificial intelligence according to an embodiment of the present application. Detailed implementation manners

[0055] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts also belong to the scope of protection of the present application.

[0056] As shown in the present application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include plural. Generally speaking, the terms "including" and "comprising" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0057] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0058] In the present application, flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the operations in front or below do not necessarily need to be executed precisely in sequence. On the contrary, according to needs, various steps can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several steps can be removed from these processes.

[0059] Next, exemplary embodiments of the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0060] In view of the above technical problems, the technical concept of this application is as follows: by collecting the user's police situation query input, introducing data processing and semantic understanding algorithms based on artificial intelligence and machine learning in the backend to perform semantic analysis on the police situation query input, and performing semantic association matching measurement with the alternative police situation data, so as to determine whether the alternative police situation data meets the matching requirements, and thus determine whether to return the alternative police situation data. In this way, according to the user's police situation query input, the police situation data semantically matching it can be screened out from the alternative police situation data and returned to the user, thereby improving the accuracy and efficiency of the police situation query, so as to provide better services and support for users.

[0061] Figure 1 It is a block diagram schematic of an intelligent management cloud platform based on artificial intelligence according to an embodiment of the present application. As Figure 1 shown, the intelligent management cloud platform 100 based on artificial intelligence according to an embodiment of the present application includes: a police situation query input module 110, configured to obtain a police situation query input; a police situation query input semantic encoding module 120, configured to perform semantic encoding on the police situation query input to obtain a police situation query input semantic encoding feature vector; a first alternative police situation data acquisition module 130, configured to acquire first alternative police situation data; a first alternative police situation text part extraction module 140, configured to extract a text part from the first alternative police situation data to obtain a first alternative police situation text part; a first alternative police situation word-level semantic encoding module 150, configured to perform word-level semantic encoding on the first alternative police situation text part to obtain a sequence of first alternative police situation word-level semantic encoding feature vectors; a police situation query input - first alternative police situation semantic matching module 160, configured to perform cross-level semantic matching on the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word-level semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature; and a first alternative police situation data matching detection and return module 170, configured to determine whether the matching degree of the first alternative police situation data exceeds a predetermined threshold based on the police situation query input - first alternative police situation semantic matching feature, and determine whether to return the first alternative police situation data.

[0062] It should be understood that the police situation query input module 110 is used to obtain the police situation query information input by the user. The police situation query input semantic encoding module 120 performs semantic encoding on the police situation query input and converts it into a semantic encoding feature vector for subsequent processing and analysis. The first alternative police situation data acquisition module 130 can obtain the first alternative police situation data that is relevant to the user's query. The first alternative police situation text part extraction module 140 extracts the text part from the first alternative police situation data to obtain information related to the text. The first alternative police situation word granularity semantic encoding module 150 performs word granularity-based semantic encoding on the first alternative police situation text part and converts the text into a sequence of word granularity semantic encoding feature vectors. The police situation query input - first alternative police situation semantic matching module 160 performs cross-granularity semantic matching on the semantic encoding feature vector of the police situation query input and the sequence of the first alternative police situation word granularity semantic encoding feature vectors to obtain the police situation query input - first alternative police situation semantic matching feature. The first alternative police situation data matching detection return module 170 determines the matching degree of the first alternative police situation data based on the police situation query input - first alternative police situation semantic matching feature and determines whether to return the first alternative police situation data. These modules together constitute the core function of the intelligent management cloud platform based on artificial intelligence. Through the processing and matching of the police situation query input and the alternative police situation data, the intelligent police situation management and query function is realized.

[0063] Specifically, in the technical solution of this application, first, the police situation query input is obtained. It should be understood that during the police situation query process, the user usually needs to use natural language to describe the police situation information they want to query, such as "a theft case that occurred in a certain place". There are semantic features regarding the user's needs and preferences in these police situation query inputs. In order to capture the semantic information of the police situation query input, in the technical solution of this application, it is necessary to perform semantic encoding on the police situation query input to extract the semantic encoding feature information in the police situation query input, thereby obtaining the police situation query input semantic encoding feature vector.

[0064] Then, during the police situation query process, the alternative police situation data usually contains detailed description information, such as the time, place, type, etc. of the case occurrence. Therefore, after obtaining the first alternative police situation data, in order to perform semantic understanding and subsequent semantic matching on it to determine whether to return the first alternative police situation data, in the technical solution of this application, the text part is further extracted from the first alternative police situation data to obtain the first alternative police situation text part to obtain text information related to the police situation. These text information may include case descriptions, witness testimonies, investigation reports, etc. Extracting the text part is to focus on the information related to the police situation for subsequent processing.

[0065] Subsequently, after performing word segmentation on the first alternative police situation text part, it is encoded through a semantic encoder including a word embedding layer to extract the semantic association feature information based on word granularity in the first alternative police situation text part, thereby obtaining a sequence of first alternative police situation word granularity semantic encoding feature vectors.

[0066] Correspondingly, the first alternative police situation word granularity semantic encoding module 150 is used to: after performing word segmentation on the first alternative police situation text part, obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors through a semantic encoder including a word embedding layer.

[0067] It is worth mentioning that the word embedding layer (Word Embedding Layer) is a technique in natural language processing used to map words in text to a low-dimensional real vector space. It can represent discrete words as continuous real-valued vectors, enabling the semantic relationships between words to be preserved and expressed in the vector space. The role of the word embedding layer is to represent words as dense vectors by learning the semantic relationships between words, where similar words are closer in the vector space and dissimilar words are farther apart. This representation method helps capture the semantic information of words and provides better feature representations for subsequent semantic analysis and model training. In the first alternative police situation word granularity semantic encoding module 150, the word embedding layer is used to convert each word in the first alternative police situation text part into a corresponding word embedding vector. Through a semantic encoder including a word embedding layer, the text sequence after word segmentation can be converted into a sequence of word granularity semantic encoding feature vectors, where each feature vector represents the representation of a word in the vector space. This can better capture the semantic information of the police situation text and provide more meaningful inputs for subsequent semantic matching and analysis.

[0068] It should be understood that the sequence of the semantic encoding feature vectors of the police situation query input and the sequence of the semantic encoding feature vectors of the first alternative police situation word granularity respectively represent the semantic encoding features of the police situation query input and the context semantic association features based on word granularity in the first alternative police situation. In police situation queries, in order to perform more accurate queries and feedback, it is necessary to match and associate the semantics between the police situation query input and the alternative police situation data. Based on this, in the technical solution of this application, a cross-granularity semantic matching module is further used to calculate the semantic matching result between the semantic encoding feature vector of the police situation query input and the sequence of the semantic encoding feature vectors of the first alternative police situation word granularity to obtain a police situation query input - first alternative police situation semantic matching feature vector. It should be understood that by using the cross-granularity semantic matching module to perform semantic association and matching between the two, the semantic matching degree between the police situation query input and the first alternative police situation can be evaluated more comprehensively. The obtained semantic matching result can provide a basis for subsequent decisions, such as determining whether to return the first alternative police situation data as the query result.

[0069] Correspondingly, the police situation query input - first alternative police situation semantic matching module 160 is used to: use a cross-granularity semantic matching module to calculate the semantic matching result between the semantic encoding feature vector of the police situation query input and the sequence of the semantic encoding feature vectors of the first alternative police situation word granularity to obtain a police situation query input - first alternative police situation semantic matching feature vector as the police situation query input - first alternative police situation semantic matching feature.

[0070] It is worth mentioning that the cross-granularity semantic matching module refers to considering semantic information of different granularities (such as word granularity, sentence granularity) simultaneously during the semantic matching process to obtain a more comprehensive semantic matching result. It can match the semantic encoding feature vector of the police situation query input and the sequence of the semantic encoding feature vectors of the first alternative police situation word granularity, thereby calculating the semantic similarity or matching degree between them. The role of the cross-granularity semantic matching module is to more accurately measure the semantic similarity between the police situation query input and the first alternative police situation by comprehensively considering semantic information of different granularities. It can capture more semantic associations, improving the accuracy and robustness of the matching. In the police situation query input - first alternative police situation semantic matching module 160, the cross-granularity semantic matching module is used to calculate the semantic matching result between the semantic encoding feature vector of the police situation query input and the sequence of the semantic encoding feature vectors of the first alternative police situation word granularity. By comparing their semantic similarities, a police situation query input - first alternative police situation semantic matching feature vector can be obtained, and this feature vector represents the semantic matching degree between the police situation query input and the first alternative police situation. This feature vector can be used to evaluate the matching degree between different alternative police situation data and the query input, and provide a basis for subsequent matching detection and result return.

[0071] Specifically, in one example, the police situation query input - first alternative police situation semantic matching module 160 is used to: use the cross - granularity semantic matching module to calculate the semantic matching result between the semantic encoding feature vector of the police situation query input and the sequence of the first alternative police situation word - granularity semantic encoding feature vectors with the following semantic matching formula to obtain the police situation query input - first alternative police situation semantic matching feature vector; where the semantic matching formula is:

[0072]

[0073] Where, v represents the semantic encoding feature vector of the police situation query input, A represents a 1×N w matrix, N w is equal to the scale of the semantic encoding feature vector of the police situation query input, B is a 1×N h matrix, N h is equal to the number of the first alternative police situation word - granularity semantic encoding feature vectors in the sequence of the first alternative police situation word - granularity semantic encoding feature vectors, σ is the Sigmoid function, s is the weight coefficient, M w and M h represent the convolution operation of a 1×1 convolution kernel, h i represents the i - th first alternative police situation word - granularity semantic encoding feature vector in the sequence of the first alternative police situation word - granularity semantic encoding feature vectors, N represents the scale of each first alternative police situation word - granularity semantic encoding feature vector in the sequence of the first alternative police situation word - granularity semantic encoding feature vectors, and v′ represents the police situation query input - first alternative police situation semantic matching feature vector.

[0074] Subsequently, the police situation query input - first alternative police situation semantic matching feature vector is then passed through a decoder - based matching degree measurer to obtain the matching degree. That is to say, the semantic matching feature information between the semantic encoding feature of the police situation query input and the word - granularity - based semantic association feature of the first alternative police situation data is used for decoding regression, so as to return the matching degree, and based on the classification result, it is determined whether to return the first alternative police situation data. In this way, according to the user's police situation query input, the police situation data that is semantically matched with it can be screened out from the alternative police situation data and returned to the user, thereby improving the accuracy and efficiency of the police situation query, so as to provide better services and support for users.

[0075] Accordingly, the first alternative police situation data matching detection return module 170 includes: a classification unit for passing the police situation query input - first alternative police situation semantic matching feature vector through a classifier to obtain a classification result, where the classification result is used to indicate whether the matching degree of the first alternative police situation data exceeds a predetermined threshold; and a police situation data return unit for determining whether to return the first alternative police situation data based on the classification result.

[0076] Accordingly, in one example, the classification unit is configured to: use the classifier to process the police situation query input - first alternative police situation semantic matching feature vector with the following classification formula to obtain the classification result; where the classification formula is: O = softmax{(W c ,B c )|X}, where X represents the police situation query input - first alternative police situation semantic matching feature vector, W c is the weight matrix, B c represents the bias vector, softmax represents the normalized exponential function, and O represents the classification result.

[0077] In the above technical solution, the sequence of the first alternative police situation word granularity semantic encoding feature vectors represents the text semantic feature based on word segmentation of the first alternative police situation data, while the police situation query input semantic encoding feature vector represents the text semantic feature of the full text granularity of the police situation query input. That is, the text semantic feature contents of the sequence of the first alternative police situation word granularity semantic encoding feature vectors and the police situation query input semantic encoding feature vector have semantic expression variability under different calculation dimensions of the word text sequence based on the first alternative police situation data and the full text of the police situation query input. Therefore, when using the cross-granularity semantic matching module to calculate the semantic matching result between the police situation query input semantic encoding feature vector and the sequence of the first alternative police situation word granularity semantic encoding feature vectors, it is expected to compensate for the category extreme deviation caused by the imbalance of the correspondence ratio of the sequence of the first alternative police situation word granularity semantic encoding feature vectors and the police situation query input semantic encoding feature vector under the sequence fine-grained distribution, so as to improve the accuracy of the classification result.

[0078] In a preferred embodiment, passing the police situation query input - first alternative police situation semantic matching feature vector through a classifier to obtain a classification result includes:

[0079] Cascading the sequence of the first alternative police situation word granularity semantic encoding feature vectors into a first alternative police situation word granularity semantic encoding cascaded feature vector, and converting the police situation query input semantic encoding feature vector into a police situation query input semantic encoding interpolation feature vector of the same length as the first alternative police situation word granularity semantic encoding cascaded feature vector through linear interpolation;

[0080] Calculate the first norm and the second norm of the mean vector of the first alternative police situation word granularity semantic coding concatenated feature vector and the police situation query input semantic coding feature vector;

[0081] Calculate the weighted sum of the reciprocal of the square root of the second norm and the first norm, and perform a dot product with the dot addition sum vector of the first alternative police situation word granularity semantic coding concatenated feature vector and the police situation query input semantic coding feature vector to obtain a first police situation query input - first alternative police situation semantic matching correction sub - vector;

[0082] Perform a dot product of the dot product vector of the first alternative police situation word granularity semantic coding concatenated feature vector and the police situation query input semantic coding interpolation feature vector with the square root of the length of the police situation query input - first alternative police situation semantic matching feature vector to obtain a second police situation query input - first alternative police situation semantic matching correction sub - vector;

[0083] Calculate the weighted sum of the first police situation query input - first alternative police situation semantic matching correction sub - vector and the second police situation query input - first alternative police situation semantic matching correction sub - vector to obtain a police situation query input - first alternative police situation semantic matching correction vector;

[0084] Perform a dot product of the police situation query input - first alternative police situation semantic matching correction vector and the police situation query input - first alternative police situation semantic matching feature vector to obtain a corrected police situation query input - first alternative police situation semantic matching feature vector;

[0085] Input the corrected police situation query input - first alternative police situation semantic matching feature vector into the classifier to obtain the classification result.

[0086] Wherein, the police situation query input - first alternative police situation semantic matching correction vector is expressed as:

[0087]

[0088] Wherein, V 1 and V 2 are respectively the first alternative police situation word granularity semantic coding concatenated feature vector and the police situation query input semantic coding feature vector, V μ is the mean vector of the first alternative police situation word granularity semantic coding concatenated feature vector and the police situation query input semantic coding feature vector, ‖·‖ 1 is the first norm of the feature vector, ‖·‖ 2 is the second norm of the feature vector, Addition is performed by position points, multiplication is performed by position points with ⊙, L is the length of the first alternative police situation query input - first alternative police situation semantic matching feature vector, and α and β are the weighted sum weights as hyperparameters, and V′ is the first alternative police situation query input - first alternative police situation semantic matching correction vector.

[0089] Thus, based on the mean vector norm between the sequence of the first alternative police situation word granularity semantic encoding feature vectors and the police situation query input semantic encoding feature vector for the constrained representation of the structured foreground and background distinction of the hypersurface body of the sequence of the first alternative police situation word granularity semantic encoding feature vectors and the police situation query input semantic encoding feature vector, the feature-level key correspondence between the sequence of the first alternative police situation word granularity semantic encoding feature vectors and the police situation query input semantic encoding feature vector is modeled, and the global association relationship between the corresponding features is adjusted. Thus, positive fine-grained correspondence suggestions are made through the imbalance ratio control between the corresponding feature values of the sequence of the first alternative police situation word granularity semantic encoding feature vectors and the police situation query input semantic encoding feature vector, so as to avoid the extreme imbalance of categories between the sequence of the first alternative police situation word granularity semantic encoding feature vectors and the police situation query input semantic encoding feature vector in a focused manner. In this way, by correcting the police situation query input - first alternative police situation semantic matching feature vector with the police situation query input - first alternative police situation semantic matching correction vector, the accuracy of the classification result of the police situation query input - first alternative police situation semantic matching feature vector input based on the classifier can be improved.

[0090] In a specific embodiment of the present application, the artificial intelligence-based intelligent management cloud platform further includes: a training module for training the semantic encoder including a word embedding layer, the cross-granularity semantic matching module, and the classifier. The training module includes: a training police situation query input acquisition unit for acquiring a training police situation query input; a training police situation query input semantic encoding unit for semantically encoding the training police situation query input to obtain a training police situation query input semantic encoding feature vector; a training first alternative police situation data acquisition unit for acquiring training first alternative police situation data; a training first alternative police situation text part extraction unit for extracting a text part from the training first alternative police situation data to obtain a training first alternative police situation text part; a training first alternative police situation word granularity semantic encoding unit for performing word segmentation on the training first alternative police situation text part and then passing it through a semantic encoder including a word embedding layer to obtain a sequence of training first alternative police situation word granularity semantic encoding feature vectors; a training police situation query input - first alternative police situation semantic matching unit for using the cross-granularity semantic matching module to calculate a semantic matching result between the training police situation query input semantic encoding feature vector and the sequence of training first alternative police situation word granularity semantic encoding feature vectors to obtain a training police situation query input - first alternative police situation semantic matching feature vector; an optimization unit for optimizing the training police situation query input - first alternative police situation semantic matching feature vector to obtain an optimized training police situation query input - first alternative police situation semantic matching feature vector; a classification loss unit for passing the optimized training police situation query input - first alternative police situation semantic matching feature vector through the classifier to obtain a classification loss function value; and an iterative training unit for training the semantic encoder including a word embedding layer, the cross-granularity semantic matching module, and the classifier based on the classification loss function value.

[0091] In a specific embodiment of the present application, optimizing the training police situation query input - first alternative police situation semantic matching feature vector to obtain an optimized training police situation query input - first alternative police situation semantic matching feature vector includes: determining a first training police situation query input - first alternative police situation semantic matching feature value and a second training police situation query input - first alternative police situation semantic matching feature value at any two different positions of the training police situation query input - first alternative police situation semantic matching feature vector; determining a compliance probability value that exceeds a predetermined threshold and represents the matching degree of the training first alternative police situation data and a non - compliance probability value that does not exceed the predetermined threshold and represents the matching degree of the training first alternative police situation data obtained by the classifier for the training police situation query input - first alternative police situation semantic matching feature vector; dividing the compliance probability value by the non - compliance probability value to obtain a training police situation query input - first alternative police situation semantic matching relative distribution value, and multiplying the compliance probability value by the non - compliance probability value to obtain a training police situation query input - first alternative police situation semantic matching co - distribution value; multiplying the results obtained by subtracting the training police situation query input - first alternative police situation semantic matching relative distribution value from the first training police situation query input - first alternative police situation semantic matching feature value and the second training police situation query input - first alternative police situation semantic matching feature value respectively to obtain a training police situation query input - first alternative police situation semantic matching relative joint representation value; adding the first training police situation query input - first alternative police situation semantic matching feature value and the second training police situation query input - first alternative police situation semantic matching feature value and then dividing by the training police situation query input - first alternative police situation semantic matching co - distribution value to obtain a training police situation query input - first alternative police situation semantic matching system response representation value; adding the training police situation query input - first alternative police situation semantic matching relative joint representation value and the training police situation query input - first alternative police situation semantic matching system response representation value, and using the result as the matrix value at the corresponding coordinates of different positions of the first training police situation query input - first alternative police situation semantic matching feature value and the second training police situation query input - first alternative police situation semantic matching feature value to obtain a training police situation query input - first alternative police situation semantic matching correction matrix; calculating the inner product of each row vector of the training police situation query input - first alternative police situation semantic matching correction matrix and the training police situation query input - first alternative police situation semantic matching feature vector to obtain the optimized training police situation query input - first alternative police situation semantic matching feature vector.

[0092] In the above preferred embodiment, taking the eigenvalue of the training police situation query input - first alternative police situation semantic matching feature vector as a unit, for the relative class probability distribution form and the collaborative class probability distribution form of the compliance class probability and the non - compliance class probability obtained by the classifier for the training police situation query input - first alternative police situation semantic matching feature vector as a feature set, the eigenvalue pairs of the training police situation query input - first alternative police situation semantic matching feature vector are respectively subjected to relative - based joint representation and collaboration - based response representation, so as to avoid the class induction bias caused by the discretization of the corresponding local distributions among the eigenvalues of the training police situation query input - first alternative police situation semantic matching feature vector, and to establish a robust understanding paradigm for class recognition of the training police situation query input - first alternative police situation semantic matching feature vector as a whole feature set, thereby improving the iterative effect of the training police situation query input - first alternative police situation semantic matching feature vector as a feature set in the classification process. That is, it improves the speed of classification training and the accuracy of classification results when the training police situation query input - first alternative police situation semantic matching feature vector is classified by the sealed quality inspector. In this way, it can more accurately determine whether the alternative police situation data meets the matching requirements, so as to determine whether to return the alternative police situation data, thereby improving the accuracy and efficiency of police situation query.

[0093] In summary, the intelligent management cloud platform 100 based on artificial intelligence according to the embodiments of the present application is elucidated, which can improve the accuracy and efficiency of police situation query.

[0094] As described above, the intelligent management cloud platform 100 based on artificial intelligence according to the embodiments of the present application can be implemented in various terminal devices, such as a server with an artificial - intelligence - based intelligent management algorithm according to the embodiments of the present application. In one example, the intelligent management cloud platform 100 based on artificial intelligence according to the embodiments of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent management cloud platform 100 based on artificial intelligence according to the embodiments of the present application can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the intelligent management cloud platform 100 based on artificial intelligence according to the embodiments of the present application can also be one of the many hardware modules of the terminal device.

[0095] Alternatively, in another example, the intelligent management cloud platform 100 based on artificial intelligence according to the embodiments of the present application and the terminal device can also be separate devices, and the intelligent management cloud platform 100 based on artificial intelligence can be connected to the terminal device through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0096] Figure 2It is a flowchart of an intelligent management method based on artificial intelligence according to an embodiment of the present application. Figure 3 It is a schematic diagram of the system architecture of an intelligent management method based on artificial intelligence according to an embodiment of the present application. As Figure 2 and Figure 3 shown, the intelligent management method based on artificial intelligence according to an embodiment of the present application includes: S110, obtaining a police situation query input; S120, performing semantic encoding on the police situation query input to obtain a police situation query input semantic encoding feature vector; S130, obtaining first alternative police situation data; S140, extracting a text part from the first alternative police situation data to obtain a first alternative police situation text part; S150, performing semantic encoding based on word granularity on the first alternative police situation text part to obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors; S160, performing cross-granularity semantic matching on the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature; and S170, based on the police situation query input - first alternative police situation semantic matching feature, determining whether the matching degree of the first alternative police situation data exceeds a predetermined threshold and determining whether to return the first alternative police situation data.

[0097] In a specific example, in the above intelligent management method based on artificial intelligence, performing semantic encoding based on word granularity on the first alternative police situation text part to obtain a sequence of first alternative police situation word granularity semantic encoding feature vectors includes: performing word segmentation on the first alternative police situation text part and then passing it through a semantic encoder including a word embedding layer to obtain the sequence of first alternative police situation word granularity semantic encoding feature vectors.

[0098] In a specific example, in the above intelligent management method based on artificial intelligence, performing cross-granularity semantic matching on the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature includes: using a cross-granularity semantic matching module to calculate the semantic matching result between the police situation query input semantic encoding feature vector and the sequence of first alternative police situation word granularity semantic encoding feature vectors to obtain a police situation query input - first alternative police situation semantic matching feature vector as the police situation query input - first alternative police situation semantic matching feature.

[0099] Here, those skilled in the art can understand that the specific operations of each step in the above intelligent management method based on artificial intelligence have been described in detail in the description of the intelligent management cloud platform 100 based on artificial intelligence above Figure 1 and therefore, the repeated description thereof will be omitted.

[0100] Figure 4 The figure is an application scenario diagram of an intelligent management cloud platform based on artificial intelligence according to an embodiment of the present application. As Figure 4 shown, in this application scenario, first, obtain an alarm query input (for example, Figure 4 D1 as shown in Figure 4 ) and first alternative alarm data (for example, Figure 4 D2 as shown in

[0101] ). Then, input the alarm query input and the first alternative alarm data into a server (for example,

[0102] S as shown in

[0103] ) deployed with an intelligent management algorithm based on artificial intelligence. Among them, the server can use the intelligent management algorithm based on artificial intelligence to process the alarm query input and the first alternative alarm data to obtain a matching degree, and then, based on the classification result, determine whether to return the first alternative alarm data.

[0104] According to another aspect of the present application, a non-volatile computer-readable storage medium is also provided, on which computer-readable instructions are stored, and when the instructions are executed by a computer, the method described above can be executed. The program part in the technology can be regarded as a "product" or "article" in the form of executable code and / or related data, which is participated in or implemented by a computer-readable medium. Tangible, permanent storage media can include any memory or storage used by a computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives, or any similar device capable of providing storage functions for software. The present application uses specific terms to describe the embodiments of the present application. Such as "the first / second embodiment", "an embodiment", and / or "some embodiments" mean a certain feature, structure, or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that the "an embodiment" or "one embodiment" or "an alternative embodiment" mentioned twice or more at different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of the present application can be combined appropriately.In addition, those skilled in the art can understand that various aspects of the present application can be illustrated and described by several patentable types or situations, including any new and useful process, machine, product, or combination of substances, or any new and useful improvement thereof. Accordingly, various aspects of the present application can be executed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software can all be referred to as "data blocks", "modules", "engines", "units", "components", or "systems". In addition, various aspects of the present application may be embodied as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0105] Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in a general dictionary should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and should not be interpreted in an idealized or overly formal sense, unless expressly defined as such herein.

[0106] The foregoing is illustrative of the present application and should not be construed as limiting thereof. Although several exemplary embodiments of the present application have been described, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application. Accordingly, all such modifications are intended to be included within the scope of the present application as defined by the claims. It should be understood that the foregoing is illustrative of the present application and should not be considered limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present application is defined by the claims and their equivalents.

Claims

1. An intelligent management cloud platform based on artificial intelligence, characterized by: include: The police situation query input module is used to obtain the police situation query input; A semantic coding module for a police query input, used for semantically coding the police query input to obtain a semantic coding feature vector of the police query input; A first candidate alarm data acquisition module, used to acquire first candidate alarm data; A first candidate warning text part extraction module, used for extracting a text part from the first candidate warning data to obtain a first candidate warning text part; A first candidate warning word granularity semantic encoding module, used for performing word granularity-based semantic encoding on the first candidate warning text portion to obtain a sequence of first candidate warning word granularity semantic encoding feature vectors; A warning query input-first candidate warning semantic matching module, used for performing cross-granularity semantic matching on the sequence of the warning query input semantic encoding feature vector and the first candidate warning word granularity semantic encoding feature vector to obtain a warning query input-first candidate warning semantic matching feature; A first candidate alarm data matching detection and return module, used for passing the alarm query input-first candidate alarm semantic matching feature through a classifier to obtain a classification result, determining whether the matching degree of the first candidate alarm data exceeds a predetermined threshold, and determining whether to return the first candidate alarm data; The alarm query input-first candidate alarm semantic matching feature is passed through a classifier to obtain a classification result, including: Cascading the sequence of the first candidate warning word granularity semantic coding feature vectors into a first candidate warning word granularity semantic coding cascade feature vector, and converting the warning query input semantic coding feature vector into a warning query input semantic coding interpolation feature vector of the same length as the first candidate warning word granularity semantic coding cascade feature vector through linear interpolation; Calculate the first norm and the second norm of the mean vector of the semantic encoding cascade feature vector of the first candidate warning word granularity and the warning query input semantic encoding feature vector; Calculate the weighted sum of the inverse of the square root of the second norm and the first norm, and perform dot multiplication with the dot-added sum vector of the first candidate warning word granularity semantic encoding cascade feature vector and the warning query input semantic encoding feature vector to obtain a first warning query input-first candidate warning semantic matching correction subvector; The dot product vector of the first candidate warning word granularity semantic encoding cascade feature vector and the warning query input semantic encoding interpolation feature vector is dotted with the square root of the length of the warning query input-first candidate warning semantic matching feature vector to obtain a second warning query input-first candidate warning semantic matching correction subvector; Calculating a weighted sum of the first alarm query input-first candidate alarm semantic matching syndrome vector and the second alarm query input-first candidate alarm semantic matching syndrome vector to obtain an alarm query input-first candidate alarm semantic matching correction vector; Performing a dot multiplication of the police situation query input-first candidate police situation semantic matching correction vector and the police situation query input-first candidate police situation semantic matching feature vector to obtain a corrected police situation query input-first candidate police situation semantic matching feature vector; The corrected alarm query input-first candidate alarm semantic matching feature vector is input into the classifier to obtain the classification result.

2. The intelligent management cloud platform based on artificial intelligence according to claim 1, characterized in that: The first candidate warning word granularity semantic encoding module is used to: After word segmentation processing is performed on the first candidate warning text portion, a semantic encoder including a word embedding layer is passed through to obtain a sequence of granular semantic encoding feature vectors of the first candidate warning word.

3. The intelligent management cloud platform based on artificial intelligence according to claim 2 is characterized in that: The alarm query input-first candidate alarm semantic matching module is used to: A cross-granularity semantic matching module is used to calculate the semantic matching result between the warning query input semantic encoding feature vector and the sequence of the first candidate warning word granularity semantic encoding feature vector to obtain the warning query input-first candidate warning semantic matching feature vector as the warning query input-first candidate warning semantic matching feature.

4. The intelligent management cloud platform based on artificial intelligence according to claim 3 is characterized in that: The alarm query input-first candidate alarm semantic matching module is used to: Using the cross-granularity semantic matching module to calculate the semantic matching result between the semantic encoding feature vector of the alarm query input and the sequence of the first candidate alarm word granularity semantic encoding feature vectors using the following semantic matching formula to obtain the alarm query input-first candidate alarm semantic matching feature vector; Wherein, the semantic matching formula is: ； ； in, represents the semantic encoding feature vector of the police query input, Indicates 1× The matrix of is equal to the scale of the semantic encoding feature vector of the police query input, Yes 1× The matrix of is equal to the number of the first candidate warning word granularity semantic coding feature vectors in the sequence of the first candidate warning word granularity semantic coding feature vectors, is the Sigmoid function, is the weight coefficient, and represents the convolution operation of a 1×1 convolution kernel, The first candidate warning word in the sequence of semantic encoding feature vectors of granularity The first candidate warning word granular semantic encoding feature vector, represents the scale of each first candidate warning word granularity semantic encoding feature vector in the sequence of the first candidate warning word granularity semantic encoding feature vectors, Represents the warning query input-first candidate warning semantic matching feature vector.

5. The intelligent management cloud platform based on artificial intelligence according to claim 4 is characterized in that: The first candidate alarm data matching detection return module includes: A classification unit, used for passing the alarm query input-first candidate alarm semantic matching feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the matching degree of the first candidate alarm data exceeds a predetermined threshold; The warning data returning unit is used to determine whether to return the first candidate warning data based on the classification result.

6. The intelligent management cloud platform based on artificial intelligence according to claim 5 is characterized in that: Also includes: A training module is used to train the semantic encoder including the word embedding layer, the cross-granularity semantic matching module and the classifier.

7. The intelligent management cloud platform based on artificial intelligence according to claim 6 is characterized in that: The training module comprises: A training alarm query input acquisition unit, used to acquire training alarm query input; A training alarm query input semantic encoding unit, used for semantically encoding the training alarm query input to obtain a training alarm query input semantic encoding feature vector; A training first candidate alarm situation data acquisition unit, used to acquire training first candidate alarm situation data; A training first candidate warning situation text part extraction unit, used for extracting a text part from the training first candidate warning situation data to obtain a training first candidate warning situation text part; A training first candidate warning word granularity semantic encoding unit is used to perform word segmentation processing on the training first candidate warning text part and then pass it through a semantic encoder including a word embedding layer to obtain a sequence of training first candidate warning word granularity semantic encoding feature vectors; A training alarm query input-first candidate alarm semantic matching unit, used to use a cross-granularity semantic matching module to calculate a semantic matching result between the training alarm query input semantic encoding feature vector and the sequence of the training first candidate alarm word granularity semantic encoding feature vector to obtain a training alarm query input-first candidate alarm semantic matching feature vector; An optimization unit, used for optimizing the training police situation query input-first candidate police situation semantic matching feature vector to obtain an optimized training police situation query input-first candidate police situation semantic matching feature vector; A classification loss unit, used for passing the optimized training alarm query input-first candidate alarm semantic matching feature vector through a classifier to obtain a classification loss function value; An iterative training unit is used to train the semantic encoder including the word embedding layer, the cross-granularity semantic matching module and the classifier based on the classification loss function value.

8. An intelligent management method based on artificial intelligence, used in the intelligent management cloud platform based on artificial intelligence as claimed in claim 1, characterized in that: The steps include: Get police query input; Performing semantic encoding on the police situation query input to obtain a semantic encoding feature vector of the police situation query input; Obtaining the first candidate alarm data; Extracting a text portion from the first candidate warning situation data to obtain a first candidate warning situation text portion; Performing word-granularity-based semantic coding on the first candidate warning text portion to obtain a sequence of first candidate warning word-granularity semantic coding feature vectors; Performing cross-granularity semantic matching on the sequence of the semantic encoding feature vector of the police situation query input and the granularity semantic encoding feature vector of the first candidate police situation word to obtain a police situation query input-first candidate police situation semantic matching feature; Based on the warning query input-first candidate warning semantic matching feature, it is determined whether the matching degree of the first candidate warning data exceeds a predetermined threshold, and it is determined whether to return the first candidate warning data.

9. The intelligent management method based on artificial intelligence according to claim 8, characterized in that: The first candidate warning text portion is semantically encoded based on word granularity to obtain a sequence of first candidate warning word granularity semantic encoding feature vectors, including: After word segmentation processing is performed on the first candidate warning text portion, a semantic encoder including a word embedding layer is passed through to obtain a sequence of granular semantic encoding feature vectors of the first candidate warning word.

10. The intelligent management method based on artificial intelligence according to claim 9, characterized in that: Performing cross-granularity semantic matching on the sequence of the semantic encoding feature vector of the alarm query input and the granularity semantic encoding feature vector of the first candidate alarm word to obtain the alarm query input-first candidate alarm semantic matching feature, including: A cross-granularity semantic matching module is used to calculate the semantic matching result between the warning query input semantic encoding feature vector and the sequence of the first candidate warning word granularity semantic encoding feature vector to obtain the warning query input-first candidate warning semantic matching feature vector as the warning query input-first candidate warning semantic matching feature.

Images