Cultural relic collection retrieval system based on intelligent algorithm

Abstract

The invention is applicable to the technical field of cultural relic collection retrieval, and provides a cultural relic collection retrieval system based on an intelligent algorithm, which comprises a query end, a server and a database, and is characterized in that the server is respectively connected with the query end and the database; the query end comprises a user interface module, a query request generation module, a user interaction module and a local data storage module; the server comprises a request processing module, a data searching module, a data processing module, a cache management module and a result returning module; the database comprises a data storage module, a data extraction module, a data updating module and a data authority management module. The semantic query precision is improved at the query end through an intelligent algorithm, and the security of cultural relic information is guaranteed while the cultural relic retrieval efficiency is improved through data information interaction among the query end, the server and the database.

Description

Technical Field

[0001] The present invention relates to the technical field of cultural relics collection retrieval, and in particular to a cultural relics collection retrieval system based on intelligent algorithms. Background Art

[0002] The retrieval of cultural relics is done by using certain retrieval tools to screen and match various attribute information such as the name, age, category, excavation site, preservation status, etc. of the cultural relics in a database, archive system or other information repository containing numerous cultural relics records according to the conditions set by the user. At present, the retrieval tools and retrieval methods used for cultural relics retrieval are relatively limited. The traditional retrieval method is mainly based on simple text keyword matching, and the language for voice retrieval is single, which requires the setting of a multi-level retrieval process, making the retrieval efficiency very low. In addition, the retrieval system has a low degree of intelligence. When it cannot well understand the user's fuzzy query intention, it cannot accurately locate the relevant cultural relics, resulting in low accuracy of retrieval results.

[0003] In view of this, a cultural relics collection retrieval system based on intelligent algorithms is needed. Summary of the invention

[0004] The embodiment of the present application provides a cultural relic collection retrieval system based on an intelligent algorithm, which is used to solve the problem of low accuracy of cultural relic retrieval results.

[0005] The first aspect of the embodiment of the present application provides a cultural relic collection retrieval system based on an intelligent algorithm, comprising:

[0006] A query end, a server and a database, wherein the server is connected to the query end and the database respectively;

[0007] The query end includes a user interface module, a query request generation module, a user interaction module and a local data storage module; the user interface module is used to receive a query instruction input by a user and output a query result corresponding to the query instruction, the query request generation module is used to generate a query request according to the query instruction input by the user and a preset rule, the user interaction module is used to feedback the query process, and the local data storage module is used to store data related to the query;

[0008] The server includes a request processing module, a data search module, a data processing module, a cache management module and a result return module; the request processing module is used to parse and verify the received query request, the data search module is used to search for a query statement that meets the conditions according to the processing result of the request processing module, the data processing module is used to perform data preprocessing on the query statement, the cache management module is used to cache data related to the query, and the result return module is used to send the query result to the query end;

[0009] The database includes a data storage module, a data extraction module, a data update module and a data authority management module; the data storage module is used to store cultural relics collection information data, the data extraction module is used to extract corresponding cultural relics collection information according to the user's query instructions, the data update module is used to update the information of cultural relics collection in real time, and the data authority management module is used to control the user's access to the cultural relics collection.

[0010] Furthermore, the query request generation module includes a request construction submodule and a request sending submodule. The request construction submodule is used to generate a query request based on a query instruction input by a user, wherein the query instruction is a query condition determined by the user according to different query methods, and the query methods include text query methods, voice query methods, and video query methods; the request sending submodule is used to send the query request to the server.

[0011] Furthermore, the request building submodule is used to generate a query request according to the query instruction input by the user, including:

[0012] When the voice in the query instruction input by the user is a dialect, extracting a target acoustic feature vector corresponding to the dialect in the query instruction;

[0013] Determining the dialect cluster to which the speech in the query instruction belongs by calculating the distance between the target acoustic feature vector and the central feature vector of each clustered dialect cluster, wherein the dialect cluster is provided with a trained hidden Markov model;

[0014] Generate a speech sequence based on the speech in the query instruction, input the speech sequence into a trained hidden Markov model, and output an optimal hidden state sequence;

[0015] The optimal hidden state sequence is converted into text corresponding to the dialect in the query instruction according to a preset state-text mapping table, and a query request is generated based on the text.

[0016] Furthermore, the method of determining the dialect cluster to which the speech in the query instruction belongs by calculating the distance between the target acoustic feature vector and the central feature vector of each clustered dialect cluster, wherein the dialect cluster is provided with a trained hidden Markov model, comprises:

[0017]

[0018] C assigned =min[d m ]

[0019] Among them: Among them: d m is the distance between the target acoustic feature vector and the center feature vector of the mth cluster dialect cluster, L is the dimension of the target acoustic feature vector, xinput,k is the k-dimensional target acoustic vector, μ m,k is the k-dimensional central feature vector of the m-th dialect cluster, C assigned is the dialect cluster to which the voice in the query command belongs.

[0020] Furthermore, the generating of a speech sequence based on the speech in the query instruction, and inputting the speech sequence into a trained hidden Markov model to output an optimal hidden state sequence includes:

[0021] The probability P(O) of the speech sequence appearing under the model is calculated by the forward algorithm:

[0022]

[0023] Where: N is the number of states, α T,i is the probability of observing the target acoustic feature vector in state i at time step T;

[0024] The optimal hidden state sequence S = {S1, S2, ..., S T}:

[0025]

[0026] Where: S T is the last state of the optimal hidden state sequence, δ T,i is the probability of the optimal path ending in state i at time step T.

[0027] Furthermore, the data extraction module includes an index management submodule, a query execution submodule and a result analysis submodule. The index management submodule is used to update the index of the query field, the query execution submodule is used to query the corresponding cultural relics collection information according to the received query instruction, and the result analysis submodule is used to analyze the privacy information of the query results and send the analyzed query results to the server.

[0028] Furthermore, the query execution submodule is used to query the corresponding cultural relic collection information according to the received query instruction, including:

[0029] Determine a query attribute value based on the received query instruction, and match the query attribute value with the attribute value of each root node in a preset decision tree;

[0030] When there is a root node matching the query attribute value, the split point of the root node and the corresponding branch are determined; when there is no root node matching the query attribute value, the query attribute value corresponding to the query instruction is returned to be re-determined;

[0031] Traversing all branches of the root node, determining a branch node that matches the query attribute value, and traversing all leaf branches of the branch node;

[0032] When a leaf branch node that cannot be further matched is determined, the cultural relic collection information is determined based on the attribute value of the leaf node.

[0033] Furthermore, traversing all branches of the root node, determining a branch node that matches the query attribute value, and traversing all leaf branches of the branch node includes:

[0034] When the query attribute value matches the attribute value of the first branch node of the root node, the split point of the first branch node and the corresponding leaf branch are determined; when the query attribute value does not match the attribute value of the first branch node of the root node, the query attribute value is matched with the attribute values ​​of other branches of the root node except the first branch;

[0035] When the query attribute value does not match other branches of the root node, the cultural relic collection information is determined according to the attribute value of the leaf node of the first branch node;

[0036] When the query attribute value matches other branches of the root node, the leaf branches of other branch nodes are traversed until a leaf branch node that cannot be matched further is determined.

[0037] The result analysis submodule is used to analyze the privacy information of the query result and send the analyzed query result to the server, including:

[0038] When the query result contains primary privacy information, an error code prompt is fed back to the server;

[0039] When the query result contains secondary privacy information, the privacy information in the query result is hidden, and a ciphertext is generated based on the non-privacy information in the query result and sent to the server;

[0040] When the query result does not contain private information, the cultural relics collection information in the query result is encrypted, and the ciphertext is generated and sent to the server.

[0041] The result returning module is used to send the query result to the query end, including:

[0042] When the received query result is prompt information of an error code, the prompt information of the error code is sent to the query end;

[0043] When the received query result is a ciphertext, the ciphertext is decrypted according to a pre-configured decryption key, and the decrypted cultural relic collection information is sent to the query end through a secure transmission protocol.

[0044] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0045] The system of the present invention performs data interaction between the query end, the server and the database. When the user inputs a query instruction at the query end, the query request is generated by the intelligent algorithm after identification and screening and then sent to the server. After receiving the query request, the server performs further verification processing and transmits it to the database. The database extracts the corresponding cultural relic information according to the query request. The database finally sends the retrieved cultural relic information to the server, and the server returns it to the query end. The present invention can improve the accuracy of cultural relic query through intelligent algorithms while ensuring the security of cultural relic information.

[0046] Other advantages, objectives, and features of the present invention will be set forth in part in the following description, and in part will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a structural schematic diagram of a cultural relic collection retrieval system based on an intelligent algorithm in the present invention;

[0048] Figure 2 This is a flow chart of an embodiment of a query request generation module in the present invention;

[0049] Figure 3 It is a schematic diagram of an embodiment of the query execution submodule of the data extraction module in the present invention. DETAILED DESCRIPTION

[0050] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "corresponding to" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0051] Embodiment 1

[0052] The following will introduce the cultural relics collection retrieval system based on intelligent algorithms in this application from the perspective of data information interaction between the query end, server and database. Figure 1 , a cultural relic collection retrieval system based on intelligent algorithm includes:

[0053] Query terminal 1, server 2 and database 3, server 2 is connected to query terminal 1 and database 3 respectively;

[0054] The query terminal 1 includes a user interface module 101, a query request generation module 102, a user interaction module 103 and a local data storage module 104; the user interface module 101 is used to receive a query instruction input by a user and output a query result corresponding to the query instruction, the query request generation module 102 is used to generate a query request according to the query instruction input by the user and a preset rule, the user interaction module 103 is used to feedback the query process, and the local data storage module 104 is used to store data related to the query;

[0055] Server 2 includes a request processing module 201, a data search module 202, a data processing module 203, a cache management module 204 and a result return module 205; the request processing module 201 is used to parse and verify the received query request, the data search module 202 is used to find the query statement that meets the conditions according to the processing result of the request processing module, the data processing module 203 is used to perform data preprocessing on the query statement, the cache management module 204 is used to cache the data related to the query, and the result return module 205 is used to send the query result to the query end 1;

[0056] The database 3 includes a data storage module 301, a data extraction module 302, a data update module 303 and a data authority management module 304; the data storage module 301 is used to store cultural relics collection information data, the data extraction module 302 is used to extract the corresponding cultural relics collection information according to the user's query instructions, the data update module 303 is used to update the information of the cultural relics collection in real time, and the data authority management module 304 is used to control the user's access to the cultural relics collection.

[0057] Specifically, the user interface module 101 includes an input unit and an output unit. The input unit consists of a simple query box, an advanced query form and a semantic query input box. Users can enter simple information such as the name, number or keywords of the cultural relics for quick query, and can perform complex queries based on the set query conditions. They can also perform semantic queries by entering natural language. In addition, semantic queries also support dialect recognition queries. The output unit consists of a query result display area, a cultural relic detail display area, a cultural relic picture and video display area, a visualization component and a related cultural relic display area. The query result display area is used to display the basic information of the queried cultural relics in a list form, including the name, age, category and other main information of the cultural relic; the cultural relic detail display area is used to display various information of the cultural relic in detail, such as the historical background, production process, size, texture material, preservation status, appraisal report summary and other text content of the cultural relic; the cultural relic picture and video display area is used to display high-definition pictures, 3D models, and related videos of the cultural relics in appropriate positions on the details page or result list, and provides operation buttons such as zoom in, zoom out, rotate, pause / play, etc., so that users can view the details of the cultural relics; the visualization component displays the distribution of the unearthed sites of the cultural relics through a map, or uses charts to intuitively present the macro information of the cultural relics, to help users better understand the distribution and composition of the cultural relics; the related cultural relic display area is used to display other cultural relic information related to the current query cultural relic.

[0058] The query request generation module 102 includes a request construction submodule and a request sending submodule. The request construction submodule is used to generate a query request according to the query instruction input by the user. The query instruction is the query condition determined by the user according to different query methods. The query method includes a text query method, a voice query method and a video query method; the request sending submodule is used to send the query request to the server.

[0059] The user interaction module 103 includes a user account management submodule, a comment submodule and a sharing submodule, wherein the user account management submodule provides functions such as user registration, login, and password retrieval, supports users to log in through a third-party account, facilitates users to use the system, and provides different functional permissions according to the user's login status; the comment submodule allows users to comment on cultural relics information, including expressing their own opinions, raising questions, sharing relevant knowledge, etc., and supports users to reply to and like other users' comments; the sharing submodule provides a variety of sharing methods, and can automatically add system name, cultural relics highlights and other information when generating sharing content.

[0060] The local data storage module 104 includes a cache management submodule and a history storage submodule. The cache management submodule caches the retrieved cultural relic information. When the user queries the same or similar content again, part of the data can be directly read from the local cache to improve the query response speed. At the same time, the validity period and storage space of the cached data are managed, and expired or infrequently used cached data are regularly cleaned up. The history storage submodule records the user's query history, provides the user with personalized recommendation functions based on the user's history, and encrypts and stores the history to protect the user's privacy. The user can choose to delete part or all of the history.

[0061] Specifically, the request processing module 201 includes a protocol parsing submodule and a request verification submodule, wherein the protocol parsing submodule is used to parse the request protocol from the request generation module, verify the legality of the request protocol, and check whether the request complies with the standard protocol specification. If the request does not comply with the specification, such as format error, lack of necessary header information, etc., the request will be rejected and a corresponding error prompt will be returned to the query end; the request verification submodule verifies the content of the request, including checking whether the query conditions are reasonable, performing user authority verification, and determining whether the user who issued the request has sufficient authority to perform the corresponding query.

[0062] The data search module 202 includes a query construction submodule and a database interaction submodule, wherein the query construction submodule is used to construct a database query statement according to the query conditions parsed and verified by the request processing module; the database interaction submodule is used to establish a connection with the database and send a query statement that meets the data extraction module 302 in the database.

[0063] The data processing module 203 includes a data cleaning submodule, which is used to clean the cultural relic data information received and decrypted from the database and other related data, and feedback the cleaned data information to the module 205.

[0064] The cache management module 204 includes a cache checking submodule and a cache updating submodule. The cache checking submodule is used to check whether there is already cultural relic data that meets the request in the cache when processing a query request; the cache updating submodule updates the relevant data in the cache when the cultural relic data in the database changes, and manages the storage space of the cache at the same time, and reasonably allocates the cache space according to the frequency of use and importance of the cached data.

[0065] The result return module 205 includes a data conversion submodule and a response sending submodule. The result return module 205 is used to send the query result to the query end 1, including:

[0066] When the received query result is a prompt message of an error code, the prompt message of the error code is sent to the query end; when the received query result is a ciphertext, the ciphertext is decrypted according to a pre-configured decryption key, and the decrypted cultural relic collection information is sent to the query end through a secure transmission protocol.

[0067] Specifically, when the received query result is a prompt message of an error code, the data conversion submodule converts the prompt message of the error code into a format received and displayed by the query end; when the received query result is a ciphertext, the data conversion submodule decrypts it and converts the decrypted data into a format received and displayed by the query end. The response sending submodule is used to send the converted data information to the query end.

[0068] The data storage module 301 includes a cultural relic basic information storage unit, a cultural relic detailed description storage unit, a cultural relic image data storage unit, a cultural relic source and collection information storage unit, and a cultural relic identification and evaluation information storage unit, wherein the cultural relic basic information storage unit is used to store the name, number, age, category, texture, size, weight and other basic attribute information of the cultural relic; the cultural relic detailed description storage unit is used to store the detailed historical background, cultural connotation, production process, meaning and use of decorative patterns of the cultural relic; the cultural relic image data storage unit is used to store the image data such as pictures, videos, 3D models of the cultural relic; the cultural relic source and collection information storage unit records the location, time, excavation process related information of the cultural relic, and the current collection unit of the cultural relic, and the storage method combines the geographic location information system to record the geographic coordinates of the location of the location, so as to facilitate geographic space analysis; the cultural relic identification and evaluation information storage unit saves the identification results of the cultural relic, the identification agency and identification expert information, the evaluation value and other contents. By setting different storage units, cultural relic collections can be distinguished based on different types, and serve as the data basis for identifying the privacy information of cultural relic collections.

[0069] The data extraction module 302 includes an index management submodule, a query execution submodule and a result analysis submodule. The index management submodule is used to update the index of the query field. The query execution submodule is used to query the corresponding cultural relics collection information according to the received query instruction. The result analysis submodule is used to analyze the privacy information of the query results and send the analyzed query results to the server.

[0070] The data update module 303 includes a data entry submodule, a data modification and deletion submodule, and a data backup and recovery submodule, wherein the data entry submodule provides a data entry interface, allowing authorized personnel to enter the information of new cultural relics into the database; the data modification and deletion submodule is used to modify and delete the stored cultural relic information; the data backup and recovery submodule is used to regularly back up the cultural relic data in the database and use the backup data for recovery.

[0071] The data authority management module 304 includes a user authority management submodule and a data encryption and security protection submodule. The user authority management submodule is used to define the access rights of different user roles to the database. For example, ordinary users may only be able to perform query operations, administrators can perform all operations such as data entry, modification, and deletion, and cultural relics experts can update identification-related data, etc. User access to the database is controlled through user authentication and authorization mechanisms; the data encryption and security protection submodule is used to encrypt and store sensitive cultural relics information to prevent the database from external attacks.

[0072] In the retrieval system of the above embodiment, through data interaction between the query end, the server and the database, the security of the cultural relic information can be improved on the basis of improving the accuracy of cultural relic retrieval.

[0073] Embodiment 2

[0074] See also Figure 2 The request building submodule in the present invention is used to generate a query request according to the query instruction input by the user, including the following steps:

[0075] S21. When the voice in the query command input by the user is a dialect, extracting the target acoustic feature vector corresponding to the dialect in the query command;

[0076] In this embodiment, when the user inputs a voice in a query instruction, it is first determined whether the voice is a dialect. If it is a dialect, the Mel frequency cepstral coefficient is used to extract the acoustic features of the dialect. Specifically:

[0077] The speech signal is pre-emphasized using a first-order digital filter. The formula is: y(n) = x(n)-α1x(n-1), where y(n) is the pre-emphasized signal, x(n) is the original speech signal, and α1 is a constant, which is 0.97. The pre-emphasized signal is framed and each frame is fast Fourier transformed to obtain the spectrum information. The Mel frequency filter bank is calculated to map the spectrum information to the Mel frequency scale. The Mel frequency f mel The conversion formula with linear frequency f is: The output of the Mel frequency filter bank is logarithmically operated to obtain the logarithmic Mel spectrum. Finally, the logarithmic Mel spectrum is discrete cosine transformed, and the first L coefficients are taken as the MFCC feature vector, that is, each speech frame can be represented as an L-dimensional target acoustic feature vector.

[0078] S22. Determine the dialect cluster to which the voice in the query instruction belongs by calculating the distance between the target acoustic feature vector and the central feature vector of each clustered dialect cluster, and the dialect cluster is provided with a trained hidden Markov model;

[0079] In this embodiment, the K-Means clustering algorithm is used to perform clustering operations on the speech samples with good feature extraction. The number of cluster centers K is pre-set, and this number can be preliminarily estimated based on the approximate number of dialect classifications. For example, taking Guangxi as an example, based on the known major dialect categories such as Nanning dialect, Liuzhou dialect, Guilin dialect, and Zhuang language, an approximate value is first set, and then the optimal value is fine-tuned and determined through subsequent experiments and analysis.

[0080] The specific steps are as follows:

[0081] Random initialization: K cluster centers are randomly initialized. For each dimension of speech sample, the Euclidean distance between the MFCC feature vector and the dimension corresponding to the speech of each cluster center is calculated, and the vector is assigned to the cluster dialect cluster with the closest cluster center. Here, each cluster is explicitly defined as a cluster dialect cluster, which means that each cluster will gather speech samples with similar dialect pronunciation characteristics.

[0082] Cluster center update: recalculate the cluster center of each cluster dialect cluster. The new cluster center is the mean of all sample feature vectors in the cluster, that is, for the mth cluster dialect cluster C m , its new cluster center μ m,k The calculation formula is: Where |C m | is the clustering dialect cluster C m The number of samples, x i,k is the mth cluster dialect cluster C m The k-dimensional feature vector of the i-th training sample, k = 1, 2…L.

[0083] Repeat the above steps of assigning samples and updating cluster centers until the cluster centers no longer change significantly, and set a convergence threshold to judge, such as the change in the Euclidean distance of the cluster centers between two iterations is less than a certain value. At this point, each clustered dialect cluster represents one or several Guangxi dialect types with similar pronunciation characteristics.

[0084] Based on the above method, after extracting the target acoustic feature vector corresponding to the dialect in the query instruction, the distance with the central feature vector of each clustered dialect cluster is calculated to determine the dialect cluster to which the speech in the query instruction belongs.

[0085]

[0086] C assigned =min[d m ]

[0087] Where: d m is the distance between the target acoustic feature vector and the center feature vector of the mth cluster dialect cluster, l is the dimension of the target acoustic feature vector, x input,kis the k-dimensional target acoustic vector, μ m,k is the k-dimensional central feature vector of the m-th dialect cluster, C assigned is the dialect cluster to which the voice in the query command belongs.

[0088] This step clearly determines the specific dialect cluster to which the speech belongs, so that the hidden Markov model corresponding to the dialect cluster can be used for accurate recognition later.

[0089] S23. Generate a speech sequence based on the voice in the query command, and input the speech sequence into the trained hidden Markov model to output the optimal hidden state sequence;

[0090] Using the hidden Markov model corresponding to the dialect cluster to which it belongs, the forward algorithm calculates the probability of the input speech appearing under the model, and the Viterbi algorithm decodes it to obtain the optimal hidden state sequence, which is converted into the corresponding text content.

[0091] Forward algorithm: The input speech sequence is O = {o1, o2, ..., o T}, then the forward probability α t,i The calculation of is as follows:

[0092] initialization: where π i is the initial state distribution of state i, is the probability of observing symbol o1 in state i.

[0093] Recursive calculation: where a ij is the transition probability from state j to state i, is the symbol o observed in state i t The probability of , where t = 2,…,T.

[0094] Finally, the probability of the input speech sequence appearing in the last time step is:

[0095]

[0096] Where: N is the number of states, α T,i is the probability of observing the target acoustic feature vector in state i at time step T.

[0097] Viterbi Algorithm:

[0098] Let δ t,i is the probability of the optimal path ending in state i up to time step t, is the previous state on the optimal path ending in state i up to time step t. Then:

[0099] initialization:

[0100] Recursive calculation: Where t=2,…,T.

[0101] Finally, the last state of the optimal hidden state sequence is:

[0102]

[0103] Then by backtracking The entire optimal hidden state sequence S = {S1, S2, ..., S T}.

[0104] S24. Convert the optimal hidden state sequence into text corresponding to the dialect in the query instruction according to a preset state-to-text mapping table, and generate a query request based on the text.

[0105] Converting the optimal hidden state sequence into corresponding text content can be achieved through a pre-established state-text mapping table, and because it is for the recognition of a specific dialect cluster, this mapping table is also established based on the pronunciation characteristics of the dialect cluster and the corresponding text expression. Based on the recognized text content, combined with the grammatical rules and field requirements of the cultural relics collection retrieval system, a precise query request is constructed. For example, taking the Guangxi dialect as an example, if the recognized text is "I want to see bronze drums", the query request can be constructed as "Find cultural relics information whose names contain 'bronze drums'", and then the query request is sent to the system backend to retrieve matching cultural relics data. In this way, a complete process is realized from dialect voice input to accurate construction of query requests to retrieve relevant information about Guangxi cultural relics collections.

[0106] The above embodiment combines the clustering algorithm with the hidden Markov model to accurately recognize the dialect-based speech, effectively improve the efficiency of semantic query, and provide an effective data basis for subsequent retrieval of cultural relics collections.

[0107] Embodiment 3

[0108] See also Figure 3 , this implementation describes the query submodule and result analysis submodule in the data extraction module, as well as the result return module in detail as follows:

[0109] The query submodule in the data extraction module includes the following steps:

[0110] S31. Determine the query attribute value based on the received query instruction, and match the query attribute value with the attribute value of each root node in the preset decision tree;

[0111] S32. When there is a root node that matches the query attribute value, the split point of the root node and the corresponding branch are determined; when there is no root node that matches the query attribute value, the query attribute value corresponding to the query instruction is returned to be re-determined;

[0112] In this embodiment, the query attribute value is determined according to the received query instruction. For example, the query request may be "find Tang Dynasty pottery unearthed in Guilin", and the system will parse out the cultural relic attributes such as "excavation site = Guilin", "era = Tang Dynasty", and "type = pottery". The parsed attributes are matched with the node attributes in the decision tree to ensure consistency in format and semantics. For example, the "era" attribute in the decision tree may be in an encoded digital form, and the "Tang Dynasty" in the query request needs to be converted into the corresponding code.

[0113] First, determine whether there is a root node that matches the query attribute value. If so, determine the split point and corresponding branch of the root node; if not, return to redetermine the query attribute value. Specifically, the decision tree preset here is a CART decision tree constructed based on the received cultural relic data. The construction process of the CART decision tree is as follows:

[0114] 1. Obtain cultural relic records from the cultural relic database, including various cultural relic attributes, such as the name, age, excavation site, type, ethnic culture, and production process of the cultural relic. Clean the data and handle missing values ​​and outliers. Encode text attributes, such as converting cultural relic types such as "pottery" and "bronze" into digital codes.

[0115] 2. In the classification scenario of cultural relics query, for each node, calculate the Gini index of all possible attributes and split points within the attributes.

[0116] The calculation formula of the root node's Gini index Gini(D) is:

[0117]

[0118] Where: D is the data set, |y| is the category label, p k is the frequency of category k in database D.

[0119] The calculation formula of the Gini index Gini(D,A,a) of the branch after the root node splits is:

[0120]

[0121] Where: |D1| and |D2| are two subsets divided at the splitting point a of the root node |D| according to attribute A, Gini(D1) is the Gini index of |D1|, and Gini(D2) is the Gini index of |D2|.

[0122] Select the attribute and split point with the smallest Gini index as the splitting condition of the root node. For example, if the calculation finds that the attribute "artifact type" has the smallest Gini index at the split point of "pottery" and "non-pottery", then the node uses "artifact type" as the splitting attribute and "pottery" as the splitting point to divide the data set into two subsets: "pottery artifacts" and "non-pottery artifacts". Based on the selected splitting attribute and splitting point, the data set is divided into two sub-datasets, which serve as the training data sets of the left and right subtrees respectively. Repeat the above process of calculating the Gini index, selecting the splitting attribute and splitting point for the training data sets of the left and right subtrees until the stopping condition is met.

[0123] Starting from the root node of the decision tree, the attribute values ​​in the query request are compared with the splitting attributes and splitting points of the root node. For example, if the root node is split according to the type of artifact, the "type = pottery" attribute in the query request is used to determine which branch to search down.

[0124] S33. Traverse all branches of the root node, determine the branch node that matches the query attribute value, and traverse all leaf branches of the branch node;

[0125] S34. When it is determined that a leaf node cannot be matched further, the specific information of the cultural relic collection is determined based on the attribute value of the leaf node.

[0126] Step S33 also includes the following:

[0127] When the query attribute value matches the attribute value of the first branch node of the root node, the split point of the first branch node and the corresponding leaf branch are determined; when the query attribute value does not match the attribute value of the first branch node of the root node, the query attribute value is matched with the attribute values ​​of other branches of the root node except the first branch;

[0128] When the query attribute value does not match other branches of the root node, the cultural relic collection information is determined according to the attribute value of the leaf node of the first branch node;

[0129] When the query attribute value matches other branches of the root node, the leaf branches of other branch nodes are traversed until a leaf node that cannot be matched further is determined.

[0130] After determining the matching root node, traverse step by step along the root node. First, the first branch node here refers to any branch under the root node. When the attribute value of the branch matches the query attribute value, the split point of the branch node and the corresponding leaf branch are determined. Secondly, when the attribute value of the branch does not match the query attribute value, the query attribute value is matched with other branches at the same root node. When the query attribute value does not match other branches, all leaf branches of the branch node are traversed according to the determined split point. If the leaf branch of the branch node cannot find a matching attribute value, the cultural relic identifier or associated information of the branch node is the cultural relic collection information to be queried. If the leaf branch of the branch node can find a matching attribute value, continue to match along the matching leaf branch until no leaf node that continues to match can be found, then the current matching leaf branch node.

[0131] Until no more matching leaf nodes can be found. When the query attribute matches other branches, all leaf branches of the matching branches are traversed in the same way, and the above matching branches and determining the next branch process are repeated until no more matching leaf nodes can be found. At this time, the cultural relic identifier corresponding to the current leaf node is collected or the cultural relic collection information associated with the leaf node is directly obtained.

[0132] Specifically, the result analysis submodule in the data extraction module and the result return module interacting therewith include the following steps:

[0133] 1. When the query result contains primary privacy information, the error code prompt information is fed back to the server;

[0134] Specifically, first-level privacy information usually refers to information that is highly sensitive, involves core privacy, and is absolutely not allowed to be viewed in the current query scenario. When the queried cultural relic information contains first-level privacy information, an error code prompt is fed back to the server. The error code indicates a "cannot view" result prompt or a text message that can express "cannot view this part of information, involving first-level privacy", so that the server can accurately identify and convey the corresponding feedback to the query end.

[0135] 2. When the query results contain secondary private information, the private information in the query results is hidden, and a ciphertext is generated based on the non-private information in the query results and sent to the server;

[0136] Secondary privacy information is slightly less sensitive than primary privacy information, but still requires a certain degree of protection. For example, some historical details of the collection of cultural relics are private to a certain extent, but after proper processing, some content can be provided under certain conditions. When the retrieved cultural relic information contains secondary privacy information, part of the privacy information needs to be hidden. The remaining non-private information is encrypted using the AES encryption algorithm, and the ciphertext is generated and sent to the server. In addition, in the data structure sent to the server, it is possible to mark which parts of the hidden privacy information are for subsequent processing by the server.

[0137] 3. When the query results do not contain private information, the cultural relics collection information in the query results will be encrypted and sent to the server after generating ciphertext.

[0138] Cultural relic information does not contain the first-level and second-level privacy information defined above, and is basic cultural relic information or general research information that can be made public. When it is found that cultural relic information does not contain privacy information, the cultural relic information is also encrypted using an encryption algorithm, and the ciphertext is generated and sent to the server.

[0139] After receiving the query results, the server will parse the data. When receiving the feedback result of the error code, it can directly send information to the query end to inform the user that the relevant part of the information cannot be viewed. When receiving the ciphertext data, it is decrypted according to the corresponding decryption algorithm. After decryption, if the received data contains secondary privacy information and is marked with hidden information, the server needs to further process it according to specific business needs and authorization conditions.

[0140] The above-mentioned embodiments can realize taking corresponding reasonable processing methods according to the different levels of privacy information in the queried cultural relics information, and performing appropriate subsequent processing on the server side, so as to protect the privacy information of cultural relics while meeting the query needs of different users for cultural relics information under different authorization situations.

[0141] It is understandable that those skilled in the art can, under the guidance of the above embodiments, combine various implementation methods in the above embodiments to obtain technical solutions of multiple implementation methods.

[0142] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A cultural relic collection retrieval system based on intelligent algorithm, characterized in that: include: A query end, a server and a database, wherein the server is connected to the query end and the database respectively; The query end includes a user interface module, a query request generation module, a user interaction module and a local data storage module; The user interface module is used to receive the query instruction input by the user and output the query result corresponding to the query instruction, the query request generation module is used to generate the query request according to the query instruction input by the user and the preset rules, the user interaction module is used to feedback the query process, and the local data storage module is used to store the data related to the query; The server includes a request processing module, a data search module, a data processing module, a cache management module and a result return module; The request processing module is used to parse and verify the received query request, the data search module is used to search for query statements that meet the conditions according to the processing results of the request processing module, the data processing module is used to perform data preprocessing on the query statements, the cache management module is used to cache data related to the query, and the result return module is used to send the query results to the query end; The database includes a data storage module, a data extraction module, a data update module and a data authority management module; The data storage module is used to store cultural relics collection information data, the data extraction module is used to extract corresponding cultural relics collection information according to the user's query instruction, the data update module is used to update the information of cultural relics collection in real time, and the data authority management module is used to control the user's authority to access the cultural relics collection.

2. The cultural relics collection retrieval system based on intelligent algorithm according to claim 1 is characterized in that: The query request generation module includes a request construction submodule and a request sending submodule. The request construction submodule is used to generate a query request according to a query instruction input by a user. The query instruction is a query condition determined by the user according to different query methods. The query methods include text query methods, voice query methods and video query methods. The request sending submodule is used to send the query request to the server.

3. The cultural relics collection retrieval system based on intelligent algorithm according to claim 2 is characterized in that: The request building submodule is used to generate a query request according to the query instruction input by the user, including: When the voice in the query instruction input by the user is a dialect, extracting a target acoustic feature vector corresponding to the dialect in the query instruction; Determining the dialect cluster to which the speech in the query instruction belongs by calculating the distance between the target acoustic feature vector and the central feature vector of each clustered dialect cluster, wherein the dialect cluster is provided with a trained hidden Markov model; Generate a speech sequence based on the speech in the query instruction, input the speech sequence into a trained hidden Markov model, and output an optimal hidden state sequence; The optimal hidden state sequence is converted into text corresponding to the dialect in the query instruction according to a preset state-text mapping table, and a query request is generated based on the text.

4. The cultural relics collection retrieval system based on intelligent algorithm according to claim 3 is characterized in that: The method of determining the dialect cluster to which the speech in the query instruction belongs by calculating the distance between the target acoustic feature vector and the central feature vector of each clustered dialect cluster, wherein the dialect cluster is provided with a trained hidden Markov model, comprises: C assigned =min[d m ] Where: d m is the distance between the target acoustic feature vector and the center feature vector of the mth cluster dialect cluster, L is the dimension of the target acoustic feature vector, x input,k is the k-dimensional target acoustic vector, μ m,k is the k-dimensional central feature vector of the m-th dialect cluster, C assigned is the dialect cluster to which the voice in the query command belongs.

5. The cultural relics collection retrieval system based on intelligent algorithm according to claim 4 is characterized in that: The method of generating a speech sequence based on the speech in the query instruction, inputting the speech sequence into a trained hidden Markov model, and outputting an optimal hidden state sequence includes: The probability P(O) of the speech sequence appearing under the model is calculated by the forward algorithm: Where: N is the number of states, α T,i is the probability of observing the target acoustic feature vector in state i at time step T; The optimal hidden state sequence S = {S1, S2, ..., S T }: Where: S T is the last state of the optimal hidden state sequence, δ T,i is the probability of the optimal path ending in state i at time step T.

6. The cultural relics collection retrieval system based on intelligent algorithm according to claim 1 is characterized in that: The data extraction module includes an index management submodule, a query execution submodule and a result analysis submodule. The index management submodule is used to update the index of the query field, the query execution submodule is used to query the corresponding cultural relics collection information according to the received query instruction, and the result analysis submodule is used to analyze the privacy information of the query result and send the analyzed query result to the server.

7. The cultural relics collection retrieval system based on intelligent algorithm according to claim 6 is characterized in that: The query execution submodule is used to query the corresponding cultural relics collection information according to the received query instruction, including: Determine a query attribute value based on the received query instruction, and match the query attribute value with the attribute value of each root node in a preset decision tree; When there is a root node matching the query attribute value, the split point of the root node and the corresponding branch are determined; when there is no root node matching the query attribute value, the query attribute value corresponding to the query instruction is returned to be re-determined; Traversing all branches of the root node, determining a branch node that matches the query attribute value, and traversing all leaf branches of the branch node; When a leaf branch node that cannot be further matched is determined, the cultural relic collection information is determined based on the attribute value of the leaf node.

8. The cultural relics collection retrieval system based on intelligent algorithm according to claim 7 is characterized in that: The traversing all branches of the root node, determining a branch node matching the query attribute value, and traversing all leaf branches of the branch node includes: When the query attribute value matches the attribute value of the first branch node of the root node, the split point of the first branch node and the corresponding leaf branch are determined; when the query attribute value does not match the attribute value of the first branch node of the root node, the query attribute value is matched with the attribute values ​​of other branches of the root node except the first branch; When the query attribute value does not match other branches of the root node, the cultural relic collection information is determined according to the attribute value of the leaf node of the first branch node; When the query attribute value matches other branches of the root node, the leaf branches of other branch nodes are traversed until a leaf branch node that cannot be matched further is determined.

9. The cultural relics collection retrieval system based on intelligent algorithm according to claim 6 is characterized in that: The result analysis submodule is used to analyze the privacy information of the query result and send the analyzed query result to the server, including: When the query result contains primary privacy information, an error code prompt is fed back to the server; When the query result contains secondary privacy information, the privacy information in the query result is hidden, and a ciphertext is generated based on the non-privacy information in the query result and sent to the server; When the query result does not contain private information, the cultural relics collection information in the query result is encrypted, and the ciphertext is generated and sent to the server.

10. The cultural relics collection retrieval system based on intelligent algorithm according to claim 9 is characterized in that: The result returning module is used to send the query result to the query end, including: When the received query result is prompt information of an error code, the prompt information of the error code is sent to the query end; When the received query result is a ciphertext, the ciphertext is decrypted according to a pre-configured decryption key, and the decrypted cultural relic collection information is sent to the query end through a secure transmission protocol.

Images