Intelligent search engine system based on artificial intelligence multi-modal

Abstract

The application discloses an intelligent search engine system based on artificial intelligence multi-mode, belongs to the technical field of retrieval, and comprises a channel management module, a channel module and a retrieval module. The channel management module is used for managing various channel parties and determining the data sharing range of the various channel parties. The channel module is used for collecting data according to the data sharing range, obtaining channel data, performing safety processing on the channel data, setting corresponding use permission requirements for the channel data after safety processing, and sharing the channel data. The retrieval module is used for retrieving the user of the corresponding channel party, obtaining the retrieval keyword of the user, retrieving according to the retrieval keyword and the shared channel data, obtaining the retrieval result, and showing the retrieval result to the user. The application avoids the cumbersome process of screening useful content from a large amount of irrelevant information, and greatly improves the efficiency and accuracy of problem solving.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of retrieval, and specifically relates to an intelligent search engine system based on artificial intelligence multi-modal. BACKGROUND

[0002] The construction field involves a large number of professional terms and daily language of workers. The traditional text search engine is difficult to match the colloquial query and the professional term due to the lack of industry semantic understanding ability, resulting in low relevance of the search results. For example, the worker inputs "how to deal with the wall peeling", the traditional engine may return decoration data instead of the solution of "improper base treatment" in the construction specification, especially the returned search results are not analyzed in the background of the current construction project, which is easy to have a large difference, and some workers and employees cannot accurately identify, resulting in that in the current construction field, many workers and employees rarely use the search engine system for retrieval, and generally adopt oral inquiry, self-searching for data and the like to solve the problem, which is low in efficiency.

[0003] In view of the above problems, the application provides an intelligent search engine system based on artificial intelligence multi-modal. SUMMARY

[0004] In order to solve the problems existing in the above scheme, the application provides an intelligent search engine system based on artificial intelligence multi-modal.

[0005] The object of the application can be achieved by the following technical scheme:

[0006] The intelligent search engine system based on artificial intelligence multi-modal comprises a channel management module, a channel module and a search module.

[0007] The channel management module is used for managing each channel party and determining the data sharing range of each channel party.

[0008] Further, the data sharing range of each channel party comprises:

[0009] Each channel party is identified, and the data source range possessed by each channel party is acquired; a search question bank is established, and the search question bank is used for storing various construction search questions;

[0010] According to the search question bank, the material data required by each construction search question is identified; and the data sharing range corresponding to each channel party is determined according to the data source range of each channel party and the material data.

[0011] Further, the data source range of each channel party is adjusted, and the adjustment process comprises:

[0012] The data source ranges of each channel party are compared, the corresponding common range is determined, and the channel party label corresponding to each common range is marked.

[0013] The common range is divided into several unit ranges, each unit range is evaluated in priority to determine the priority of the channel party on the unit range, the unit range is removed from the data source range of the channel party with the non-highest priority, and the adjustment of each data source range is completed.

[0014] Further, the priority of each channel party on the corresponding unit range is determined, including:

[0015] Setting evaluation items and weight coefficients of each evaluation item;

[0016] According to each evaluation item, the unit range is evaluated to obtain the single item score of the unit range on the corresponding evaluation item;

[0017] According to the priority formula, the priority value of the corresponding channel party on the unit range is calculated, and the priority formula is:

[0018] ;

[0019] In the formula, YQ is the priority value of the corresponding channel party on the unit range; i represents the corresponding evaluation item, i=1, 2, …, n, n is the number of evaluation items; βi represents the weight coefficient of the corresponding evaluation item; DFi represents the single item score of the corresponding evaluation item;

[0020] According to the order from large to small of the priority value, the priority of each channel party on the unit range is determined.

[0021] Further, the data sharing range corresponding to each channel party is determined according to the data source range and the material data of each channel party, including:

[0022] Judging whether the corresponding material data is located in the data source range of the channel party;

[0023] When the material data is located in the data source range of the channel party, the material data is supplemented into the data sharing range of the channel party;

[0024] When the material data is not located in the data source range of the channel party, no corresponding processing is performed;

[0025] In this way, the data sharing range of each channel party is determined.

[0026] Further, the data sharing range corresponding to each channel party is determined according to the data source range and the material data of each channel party, including:

[0027] Identifying the channel party including the data source range of the material data;

[0028] When the channel party is one, the material data is supplemented into the data sharing range of the channel party;

[0029] When the channel party is more than one, the priority of each channel party on the corresponding material data is determined, and the material data is supplemented into the data sharing range of the channel party with the highest priority;

[0030] By analogy, the data sharing range of each channel party is determined.

[0031] The channel module is used for data collection according to the data sharing range, obtaining channel data, performing security processing on the channel data, and setting corresponding use permission requirements for the security-processed channel data, and sharing the channel data.

[0032] The search module is used for searching by the user of the corresponding channel party, obtaining the search keyword of the user, searching according to the search keyword and the shared channel data, obtaining the search result, and displaying the search result to the user.

[0033] Further, searching according to the search keyword and the shared channel data includes:

[0034] A permission calibration model is established, and the permission calibration model is used to analyze whether the corresponding user information meets the use permission requirements of the corresponding channel data;

[0035] According to the search keyword, the channel data to be used is determined, the corresponding channel data is marked as search material data, the permission use requirements corresponding to the search material data are identified, the search material data is filtered through the permission calibration model, and the search result is generated according to the remaining search material data.

[0036] Further, the expression of the permission calibration model is:

[0037] ;

[0038] In the formula, (s, BZ) is input data, s is user permission information, BZ is corresponding use permission requirement, s→BZ indicates that the user permission information meets the corresponding use permission requirement, the output data is the permission calibration value QP(s, BZ), and the permission calibration value is 1 or 0.

[0039] Further, filtering the search material data through the permission calibration model includes:

[0040] The use permission requirement of the corresponding search material data is identified, and the user permission information of the user is obtained according to the use permission requirement;

[0041] The user permission information and the use permission requirement of the corresponding search material data are integrated as input data, the input data is analyzed through a permission calibration model, and a weight calibration value of the corresponding search material data is obtained; search material data with a permission calibration value of 0 is removed.

[0042] Further, when the user does not meet the corresponding use permission requirement, the user can apply for separate permission from the corresponding channel party.

[0043] Compared with the prior art, the beneficial effects of the present application are:

[0044] The present application can deeply understand the semantic information in the construction field by deeply integrating natural language processing, knowledge graph and other artificial intelligence technologies, accurately identify the professional intent behind the colloquial query, and provide highly relevant search results for users. This not only avoids the cumbersome process of users screening useful content from a large amount of irrelevant information, but also greatly improves the efficiency and accuracy of problem solving, so that workers can quickly obtain solutions that meet construction specifications and actual needs; at the same time, various search materials applied are accurate data provided by various related parties, improving search accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0046] Figure 1 The present application is a principle block diagram. DETAILED DESCRIPTION

[0047] The technical solutions of the present application will be described below in conjunction with the embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0048] As shown in Figure 1 The intelligent search engine system based on artificial intelligence multi-modal includes a channel module, a channel module and a search module.

[0049] The channel management module is used for managing various channel parties, such as adding channel parties of the construction project, reducing channel parties of the construction project, opening corresponding permissions for corresponding channel parties, such as opening user accounts for employees, workers, etc. in the channel party, facilitating corresponding employees to log in, and determining the data sharing range of each channel party.

[0050] Channel parties such as construction parties, opening parties, contracting parties, supervision parties, design parties and the like; the channel parties have data within the data sharing range provided for the system, such as data related to construction specifications, construction schemes, supervision standards, progress requirements and the like.

[0051] In an embodiment, the specific channel party management mode, permission management and the like can be managed according to actual needs based on existing modes.

[0052] In an embodiment, the data sharing range of each channel party is determined, including:

[0053] Each channel party is identified, and the data source range possessed by each channel party is obtained, i.e., which data does the channel party have, such as the construction range managed by each construction personnel of the general construction contracting party, the construction scheme of each construction project, the detailed construction briefing, the construction specification and the like.

[0054] Based on big data, historical demand and the like, various construction search problems that can be had under the background of the construction project are determined, and are summarized into a search problem library; the search problem library is dynamically updated according to changes in the later period.

[0055] According to the search problem library, various construction search problems are identified, and the data needed by each construction search problem is identified and marked as material data; according to the data source range of each channel party and the material data, the data sharing range corresponding to each channel party is determined.

[0056] In an embodiment, the data source range of each channel party is adjusted to avoid a material data corresponding to the data source range of multiple channel parties; the adjustment process includes:

[0057] The data source ranges of each channel party are compared to determine the range intersection, which is marked as a common range, and each common range is marked with a channel party tag to indicate which data source range of which channel party is in the common range intersection;

[0058] The common range is segmented to form a plurality of unit ranges; that is, the common range is divided into different unit ranges according to whether the corresponding data is the same, such as the common range consisting of data A, B, C and D, which is divided into A, B, C and D respectively corresponding unit ranges; each unit range is evaluated in priority to determine the priority of each channel party in the unit range, and the unit range is removed from the data source range of the channel party with the highest priority, and so on, to complete the adjustment of each data source range.

[0059] In an embodiment, the priority of each channel party in the unit range is determined, including:

[0060] Set the corresponding evaluation items and the weight coefficients of each evaluation item, such as the accuracy, completeness, timeliness, cost and other evaluation items that have an impact on the priority;

[0061] According to the evaluation of each evaluation item on the unit range, the single-item score of the unit range in the corresponding evaluation item is obtained; the score range is unified (such as 0-100 or 1-5), so as to avoid the weight invalidation caused by the dimension difference of different indicators;

[0062] According to the priority formula, the priority value of the corresponding channel party on the corresponding unit range is calculated, and the priority formula is:

[0063] ;

[0064] In the formula, YQ is the priority value of the corresponding channel party on the unit range; i represents the corresponding evaluation item, i=1, 2, …, n, and n is the number of evaluation items; βi represents the weight coefficient of the corresponding evaluation item; and DFi represents the single-item score of the corresponding evaluation item.

[0065] In one embodiment, the priority formula can also be other existing formulas.

[0066] In one embodiment, according to the evaluation of each evaluation item on the unit range, the evaluation based on the existing evaluation method is carried out, such as:

[0067] Accuracy: Accuracy score=100-(error rate×100);

[0068] Completeness: formula: Completeness score=(actual field number / demand field number)×100;

[0069] Timeliness:

[0070] Delay time method: formula: Timeliness score=100-[(actual delay time / maximum tolerance time)×100];

[0071] Update frequency method:

[0072] The preset reference frequency (such as real-time, daily, weekly) is scored by gradient:

[0073] Real-time: 100 points;

[0074] Daily: 80 points;

[0075] Weekly: 60 points;

[0076] Monthly: 40 points.

[0077] Cost: formula: Cost score=(minimum cost / actual cost)×100.

[0078] In an embodiment, the priority of each channel party in the unit range is determined, and other existing manners can also be used for evaluation, such as determining each evaluation item according to the above embodiment, and then evaluating the priority according to the existing manner, as long as it is determined whether it is the highest priority.

[0079] In an embodiment, the data sharing range of each channel party is determined according to the data source range and the material data of each channel party, which can be determined based on existing manners, such as determining whether the material data is within the data source range of the channel party, if so, it is a component of the data sharing range, and so on, to determine the data sharing range of each channel party.

[0080] In an embodiment, the data sharing range of each channel party is determined according to the data source range and the material data of each channel party, which includes:

[0081] The data source range including the material data may appear a case that one material data corresponds to multiple data source ranges. The evaluation according to the priority evaluation manner (the manner of determining the priority of each channel party in the corresponding unit range in the above embodiment) needs to be performed on the channel party that needs to share the material data, so as to avoid multi-source sharing of the same data. Subsequently, the data sharing range is determined according to the above embodiment.

[0082] In an embodiment, the retrieval problem library can not be established, and each construction retrieval problem can be directly classified according to retrieval requirements to obtain various retrieval requirements, determine the data required to solve the retrieval requirement, determine the corresponding channel party according to the data source range of the corresponding channel party, and then aggregate the data range that needs to be shared by the channel party to mark as the data sharing range.

[0083] The channel module is used for data collection according to the data sharing range to obtain channel data, that is, the data in the data sharing range collected, and the channel data is safely processed. The channel party sets the corresponding use permission requirement for the channel data after the safe processing, such as indicating that only the user corresponding to the channel party is applied, and the current channel data is shared. Generally, it is shared to a unified database, such as a cloud database.

[0084] And the channel module can also be managed by the channel party according to the corresponding basic function, that is, the basic function of each channel party, such as opening the user permission of the employee, data uploading, information management, etc.

[0085] In one embodiment, the security processing of channel data and the sharing of current channel data need to be determined according to specific measures adopted, such as the combination of end-to-end encryption and homomorphic analysis, secure indexing and blind retrieval technology, access control and identity verification, federated learning and secure aggregation, blockchain and smart contract, data desensitization and anonymization, and other technical methods.

[0086] For example, the process of combining end-to-end encryption with homomorphic analysis:

[0087] The channel party uses symmetric encryption algorithms such as AES to encrypt data.

[0088] Upload encrypted data to the cloud through HTTPS.

[0089] The cloud uses KMS-managed keys for second-layer encrypted storage.

[0090] Subsequently, homomorphic encryption technology is used to combine encrypted data for direct retrieval analysis, and the corresponding retrieval results are obtained; the retrieval results are returned to the user in encrypted form, and after decryption, they are displayed to the user.

[0091] The retrieval module is used for users of the corresponding channel party to perform retrieval, obtain the user's retrieval keywords, perform retrieval according to the retrieval keywords and shared channel data, obtain retrieval results, and display the retrieval results to the user.

[0092] In one embodiment, displaying the retrieval results to the user refers to decrypted data that is decrypted according to the security processing measures in the early stage.

[0093] In one embodiment, performing retrieval according to the retrieval keywords and shared channel data means performing retrieval based on the retrieval keywords and the channel data shared by each channel party according to existing retrieval technology, and performing retrieval based on channel data with usage rights.

[0094] In one embodiment, performing retrieval according to the retrieval keywords and shared channel data includes:

[0095] Establish a permission calibration model, which is used to analyze whether the corresponding user information meets the usage permission requirements of the corresponding channel data;

[0096] According to the search keyword, the channel data to be used is determined, such as the user searching for the plastering construction method of building No. 2, and the corresponding plastering construction disclosure of building No. 2 is required. Based on the existing search technology, the required channel data is determined, and the corresponding channel data is marked as search material data. The permission use requirement corresponding to the search material data is identified, the search material data is screened through the permission calibration model, the search material data that does not meet the permission requirement is removed, and the search result is generated according to the remaining search material data according to the existing search technology. If there is no remaining search material data or the data is incomplete, it can be prompted that the permission is insufficient, and the user can apply for separate permission from the corresponding channel party in the future.

[0097] In one embodiment, the permission calibration model can be established based on existing machine learning, deep learning algorithms and other intelligent algorithms.

[0098] In one embodiment, the expression of the permission calibration model is:

[0099] ;

[0100] In the formula: (s, BZ) is the input data, s is the user permission information, BZ is the corresponding use permission requirement, s→BZ indicates that the user permission information meets the corresponding use permission requirement, and the corresponding historical data is used to set the training set for training; the output data is the permission calibration value QP(s, BZ), and the permission calibration value is 1 or 0.

[0101] The search material data is screened through the permission calibration model, including:

[0102] The use permission requirement of the corresponding search material data is identified, and the user permission information of the user is obtained according to the use permission requirement, that is, the user is data collected according to the data required by the use permission requirement, and the subsequent comparison and judgment are carried out;

[0103] The user permission information and the use permission requirement of the corresponding search material data are integrated into the input data, the input data is analyzed through the permission calibration model, and the weight calibration value of the corresponding search material data is obtained; the search material data with the permission calibration value of 0 is removed.

[0104] In one embodiment, according to the search keyword and the shared channel data, the search is carried out according to the no permission requirement, the initial search result is obtained, the use permission requirement of the channel data corresponding to the initial search result is identified, the initial search result is adjusted according to the use permission requirement, and the search result is obtained.

[0105] The above formulas are dimensionless values calculated by removing dimensions, the formulas are obtained by collecting a large amount of data to simulate software to obtain a formula closest to the actual situation, and the preset parameters and the preset threshold in the formula are set by a person skilled in the art according to the actual situation or obtained by a large amount of data simulation.

[0106] The above embodiments are only used to illustrate the technical method of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. An intelligent search engine system based on artificial intelligence and multimodal operation, characterized in that: It includes a channel management module, a channel module, and a search module; The channel management module is used to manage various channel partners, identify each channel partner, and obtain the data source range of each channel partner; it also establishes a retrieval question database to store various construction retrieval questions. The required material data for each construction search question is identified based on the search question database; the data sharing scope for each channel is determined based on the data source range and material data of each channel. The data source scope of each channel party will be adjusted. The adjustment process includes: Compare the data source ranges of each channel party, determine the corresponding common ranges, and label each common range with the corresponding channel party tag; The shared scope is divided into several unit scopes; each unit scope is prioritized and evaluated to determine the priority of the corresponding channel party in the unit scope; the unit scope is removed from the data source scope of the channel party whose priority is not the highest, thus completing the adjustment of each data source scope. The channel module is used to collect data according to the data sharing scope, obtain channel data, perform security processing on the channel data, set corresponding usage permission requirements for the security-processed channel data, and share the channel data. The retrieval module is used by users of the corresponding channel to conduct searches, obtain users' search keywords, perform searches based on search keywords and shared channel data, obtain search results, and display the search results to users.

2. The intelligent search engine system based on artificial intelligence and multimodal operation according to claim 1, characterized in that, Determine the priority of each channel within its respective unit scope, including: Set the evaluation items and the weight coefficients for each evaluation item; The unit range is evaluated based on each evaluation item to obtain the individual score of the unit range on the corresponding evaluation item; The priority value of the corresponding channel within the unit range is calculated according to the priority formula, which is: ； In the formula: YQ is the priority value of the corresponding channel in this unit range; i represents the corresponding evaluation item, i=1, 2, ..., n, where n is the number of evaluation items; βi represents the weight coefficient of the corresponding evaluation item; DFi represents the individual score of the corresponding evaluation item; The priority of each channel within the unit range is determined according to the priority value from largest to smallest.

3. The intelligent search engine system based on artificial intelligence and multimodal operation according to claim 1, characterized in that, The data sharing scope for each channel is determined based on the data source range and material data of each channel, including: Determine whether the corresponding material data falls within the data source range of the channel provider; When the material data is within the data source range of the channel provider, the material data is added to the data sharing range of the channel provider; When the material data is not within the data source range of the channel provider, no corresponding processing is performed; By analogy, the scope of data sharing among various channel partners can be determined.

4. The intelligent search engine system based on artificial intelligence and multimodal operation according to claim 1, characterized in that, The data sharing scope for each channel is determined based on the data source range and material data of each channel, including: Identify the channel providers that include the data source range of the material data; When there is only one channel, the material data will be added to the data sharing scope of that channel. When there is more than one channel, determine the priority of each channel on the corresponding material data, and supplement the material data into the data sharing scope of the channel with the highest priority. By analogy, the scope of data sharing among various channel partners can be determined.

5. The intelligent search engine system based on artificial intelligence and multimodal operation according to claim 1, characterized in that, Searches are conducted based on search keywords and shared channel data, including: Establish a permission calibration model, which is used to analyze whether the corresponding user information meets the usage permission requirements of the corresponding channel data; Based on the search keywords, determine the channel data to be used and mark the corresponding channel data as search material data; identify the permission requirements corresponding to the search material data, filter the search material data through the permission calibration model, and generate search results based on the remaining search material data.

6. The intelligent search engine system based on artificial intelligence and multimodal operation according to claim 5, characterized in that, The expression for the permission calibration model is: ； In the formula: (s, BZ) are the input data, s is the user permission information, BZ is the corresponding usage permission requirement, s→BZ means that the user permission information meets the corresponding usage permission requirement, and the output data is the permission calibration value QP(s, BZ), which is 1 or 0.

7. The intelligent search engine system based on artificial intelligence and multimodal operation according to claim 6, characterized in that, The retrieved material data is filtered using a permission calibration model, including: Identify the usage permission requirements for the corresponding search material data, and obtain the user's user permission information based on the usage permission requirements; User permission information and the usage permission requirements of corresponding search material data are integrated into input data. The input data is analyzed through a permission calibration model to obtain the weight calibration value of the corresponding search material data. Search material data with a permission calibration value of 0 are removed.

8. The intelligent search engine system based on artificial intelligence and multimodal operation according to claim 5, characterized in that, When a user does not meet the requirements for the corresponding usage permissions, the user can apply to the relevant channel provider to grant permissions separately.

Images