Visual intelligent decision-making system and method based on digital platform

Abstract

The present application discloses a visual intelligent decision-making system and method based on a digital platform, which constructs a digital intelligent decision-making system by utilizing natural language processing technology and intelligent algorithms, and realizes in-depth comparison between bidding requirements and bidding documents by semantically understanding and matching text descriptions of bidding requirements and bidding documents submitted by multiple bidding companies, and intelligently judges and decides on the matching degree between bidding requirements and bidding documents, thereby improving the efficiency of the bidding process and helping companies find the most suitable partners.

Description

Technical Field

[0001] The present application relates to the field of digitalization, and more specifically, to a visual intelligent decision-making system and method based on a digital platform. Background Art

[0002] In the modern business environment, the bidding process is a complex and important task. Enterprises need to select the most suitable partner from a large number of bidding companies. However, due to the number and complexity of bidding documents, this process often requires a lot of time and resources. In addition, the traditional bidding decision-making process may be biased and inconsistent due to human factors.

[0003] In order to solve these problems, many companies have begun to seek automated solutions. However, existing systems usually rely on simple keyword matching or statistical methods, which cannot fully understand and compare the semantic content of bidding requirements and bidding documents, resulting in limited efficiency in decision-making results.

[0004] Therefore, we look forward to an optimized visual intelligent decision-making system and method based on a digital platform. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a visual intelligent decision-making system and method based on a digital platform, which uses natural language processing technology and intelligent algorithms to build a digital intelligent decision-making system, and through semantic understanding and matching of the text description of the bidding requirements and the bid documents submitted by multiple bidding companies, realizes in-depth comparison between the bidding requirements and the bid documents, intelligently judges and decides on the matching degree between the bidding requirements and the bid documents, improves the efficiency of the bidding process, and helps companies find the most suitable partners.

[0006] According to one aspect of the present application, a visual intelligent decision-making method based on a digital platform is provided, which includes:

[0007] Get the text description of the bidding requirements;

[0008] Extract the tender documents submitted by multiple bidding companies from the backend database;

[0009] Performing a joint semantic matching analysis on the text description of the bidding requirements and the bid documents submitted by the multiple bidding enterprises to obtain multiple bidding requirements-bidding text semantic matching feature vectors;

[0010] Based on the multiple bidding requirement-bidding text semantic matching feature vectors, the recommendation priorities of the multiple bidding enterprises are determined.

[0011] According to another aspect of the present application, a visual intelligent decision-making system based on a digital platform is provided, which includes:

[0012] Information collection module, used to obtain text descriptions of bidding requirements;

[0013] The tender document extraction module is used to extract the tender documents submitted by multiple bidding companies from the backend database;

[0014] A joint semantic matching analysis module, used for performing a joint semantic matching analysis on the text description of the bidding requirements and the bid documents submitted by the multiple bidding enterprises to obtain multiple bidding requirements-bidding text semantic matching feature vectors;

[0015] The priority recommendation module is used to determine the recommendation priorities of the multiple bidding enterprises based on the multiple bidding requirements-bidding text semantic matching feature vectors.

[0016] Compared with the existing technology, the present application provides a visual intelligent decision-making system and method based on a digital platform. It constructs a digital intelligent decision-making system by utilizing natural language processing technology and intelligent algorithms, and performs semantic understanding and matching of the text description of the bidding requirements and the bid documents submitted by multiple bidding companies. It realizes an in-depth comparison between the bidding requirements and the bid documents, intelligently judges and decides on the matching degree between the bidding requirements and the bid documents, improves the efficiency of the bidding process, and helps companies find the most suitable partners. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 It is a flowchart of a visual intelligent decision-making method based on a digital platform according to an embodiment of the present application;

[0019] Figure 2 A system architecture diagram of a visual intelligent decision-making method based on a digital platform according to an embodiment of the present application;

[0020] Figure 3 It is a flowchart of sub-step S3 of the visual intelligent decision-making method based on a digital platform according to an embodiment of the present application;

[0021] Figure 4 It is a flowchart of sub-step S4 of the visual intelligent decision-making method based on a digital platform according to an embodiment of the present application;

[0022] Figure 5 It is a block diagram of a visual intelligent decision-making system based on a digital platform according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the exemplary embodiments according to 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, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" 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.

[0025] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0026] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0027] Below, the exemplary embodiments according to 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, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0028] The traditional bidding decision-making process may be biased and inconsistent due to human factors. In order to solve these problems, many companies have begun to seek automated solutions. However, existing systems usually rely on simple keyword matching or statistical methods, which cannot fully understand and compare the semantic content of bidding requirements and bidding documents, resulting in limited efficiency of decision-making results. Therefore, an optimized visual intelligent decision-making system and method based on a digital platform is expected.

[0029] In the technical solution of the present application, a visual intelligent decision-making method based on a digital platform is proposed. Figure 1It is a flowchart of a visual intelligent decision-making method based on a digital platform according to an embodiment of the present application. Figure 2 : is a system architecture diagram of a visual intelligent decision-making method based on a digital platform according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the visual intelligent decision-making method based on a digital platform includes the following steps: S1, obtaining a text description of the bidding requirements; S2, extracting bid documents submitted by multiple bidding companies from a background database; S3, performing a joint semantic matching analysis on the text description of the bidding requirements and the bid documents submitted by the multiple bidding companies to obtain multiple bidding requirement-bidding text semantic matching feature vectors; S4, determining the recommendation priority of the multiple bidding companies based on the multiple bidding requirement-bidding text semantic matching feature vectors.

[0030] In particular, S1 obtains a text description of the bidding requirements. The text description of the bidding requirements is a text description of the detailed description and requirements of the bidding project issued by the bidding agency or enterprise. These text descriptions usually include the following: 1. Project background and purpose: describe the background information, purpose and significance of the bidding project to help the bidding enterprise better understand the project background and objectives. 2. Project scope and requirements: include detailed requirements such as the specific scope of the project, work content, technical requirements, quality standards, delivery requirements, etc., to guide the bidding enterprise to understand the project requirements. 3. Schedule and plan: include time-related information such as the project schedule, construction period requirements, milestone plan, etc., to help the bidding enterprise evaluate whether its own capabilities meet the project requirements. 4. Qualification requirements and conditions: describe the qualification requirements, conditions, certificates or qualifications that the bidding enterprise needs to meet to determine whether the bidding enterprise meets the bidding qualifications. 5. Evaluation criteria and scoring methods: describe the criteria and methods of bidding evaluation to help the bidding enterprise understand how to be evaluated and compared.

[0031] In particular, the S2 extracts the bid documents submitted by multiple bidding companies from the background database. It should be understood that extracting the bid documents submitted by multiple bidding companies from the background database can understand the proposal content of each bidding company. In an embodiment of the present application, extracting the bid documents submitted by multiple bidding companies from the background database can be achieved through the following steps: 1. Database connection: ensure that the system can access the bid document data stored in the background database. 2. Query data: use appropriate query statements to retrieve the bid documents submitted by multiple bidding companies from the database, and select the corresponding bid document data according to the project ID, company ID or other identifiers. 3. Data extraction: once the required bid document data is queried, the system can extract this data into memory or a temporary file for subsequent semantic analysis and processing.

[0032] In particular, the S3 performs a joint semantic matching analysis on the text description of the bidding requirements and the bid documents submitted by the multiple bidding enterprises to obtain multiple bidding requirements-bidding text semantic matching feature vectors. In particular, in a specific example of the present application, Figure 3 As shown, the S3 includes: S31, extracting text parts from the bid documents submitted by the multiple bidding companies to obtain multiple bid text descriptions; S32, semantically encoding the text description of the bidding requirements to obtain a bidding requirement semantic encoding feature vector; S33, semantically encoding each bid text description in the multiple bid text descriptions to obtain multiple bid text semantic encoding feature vectors; S34, semantically interactively matching the bidding requirement semantic encoding feature vector and the multiple bid text semantic encoding feature vectors to obtain the bidding requirement-bidding text semantic matching feature vector.

[0033] Specifically, the S31 extracts text parts from the bid documents submitted by the multiple bidding companies to obtain multiple bid text descriptions. Considering that there may be a large amount of redundant information in the bid documents submitted by the multiple bidding companies, these redundant information are not important for judging and evaluating the matching degree between the bidding requirements and the bid documents. For example, irrelevant attachments or materials, such as advertising pictures, product manuals, etc., are not directly related to the core content of the bidding requirements, and cannot be directly semantically analyzed and understood. For example, the general template content, the general content in these templates, such as specific symbols and logos, is not of substantive significance for understanding the semantic connotation of the proposal. Therefore, in the technical solution of the present application, the text parts are extracted from the bid documents submitted by the multiple bidding companies to obtain multiple bid text descriptions. In this way, the subsequent model can perform semantic analysis on the multiple bid text descriptions, understand the semantic content of the proposals of each company, and provide more information and basis for subsequent semantic matching.

[0034] Specifically, the S32 semantically encodes the text description of the tender requirements to obtain a semantically encoded feature vector of the tender requirements. That is, the text description of the tender requirements is semantically encoded to obtain a semantically encoded feature vector of the tender requirements. That is, the text description of the tender requirements that cannot be directly recognized and understood by the computer is converted into a vector representation that is easy to read and analyze by means of semantic encoding. More specifically, the text description of the tender requirements is semantically encoded to obtain a semantically encoded feature vector of the tender requirements, including: performing word segmentation processing on the text description of the tender requirements to convert the text description of the tender requirements into a word sequence composed of multiple words; using the embedding layer of the semantic encoder containing the embedding layer to map each word in the word sequence to a word embedding vector to obtain a sequence of word embedding vectors; using the converter of the semantic encoder containing the embedding layer to perform global context semantic encoding on the sequence of word embedding vectors based on the converter idea to obtain multiple global context semantic feature vectors; and, cascading the multiple global context semantic feature vectors to obtain the semantic encoding feature vector of the tender requirements. In a specific example, a converter of the semantic encoder including the embedding layer is used to perform global context semantic encoding based on the converter idea on the sequence of word embedding vectors to obtain multiple global context semantic feature vectors, including: arranging the sequence of word embedding vectors in one dimension to obtain a global feature vector; calculating the product between the global feature vector and the transposed vector of each word embedding vector in the sequence of word embedding vectors to obtain multiple self-attention association matrices; normalizing each of the multiple self-attention association matrices to obtain multiple standardized self-attention association matrices; passing each of the multiple standardized self-attention association matrices through a Softmax classification function to obtain multiple probability values; weighting each of the word embedding vectors in the sequence of word embedding vectors using each of the multiple probability values ​​as a weight to obtain the multiple context semantic feature vectors; and, cascading the multiple context semantic feature vectors to obtain the multiple global context semantic feature vectors.

[0035] Specifically, the S33 performs semantic encoding on each of the multiple tender text descriptions to obtain multiple tender text semantic encoding feature vectors. That is, each of the multiple tender text descriptions is semantically encoded to obtain multiple tender text semantic encoding feature vectors. It should be understood that semantic encoding can better capture the semantic relationship expressed by each of the tender text descriptions, rather than just based on the surface similarity of words, thereby guiding the model to more deeply understand the core text expression of the proposal. In this way, the semantic meaning of the tender requirement expressed by the text description of the tender requirement and the semantic meaning of the tender proposal expressed by each tender text description are deeply understood and portrayed by means of semantic encoding, so as to facilitate subsequent semantic matching and comparison. More specifically, semantic encoding is performed on each of the multiple bid text descriptions to obtain multiple bid text semantic encoding feature vectors, including: performing word segmentation on the bid text description to convert the bid text description into a word sequence composed of multiple words; using the embedding layer of the semantic encoder including the embedding layer to map each word in the word sequence to a word embedding vector to obtain a sequence of word embedding vectors; using the converter of the semantic encoder including the embedding layer to perform global context semantic encoding based on the converter idea on the sequence of word embedding vectors to obtain multiple global context semantic feature vectors; and, cascading the multiple global context semantic feature vectors to obtain the bid text semantic encoding feature vector. It is worth noting that only the process of semantic encoding of one bid text description is expanded here, and multiple semantic encodings are required to obtain the multiple bid text semantic encoding feature vectors.

[0036] Specifically, the S34 performs semantic interactive matching on the bidding demand semantic encoding feature vector and the multiple bidding document text semantic encoding feature vectors to obtain the bidding demand-bidding text semantic matching feature vector. That is, the interactive attention layer is used to construct a semantic matching module, so as to achieve a more in-depth and accurate semantic matching between bidding demand and bidding document text. Specifically, the interactive attention layer uses the attention mechanism to construct the similarity correlation between the bidding demand semantic encoding feature vector and each bidding document text semantic encoding feature vector, and assigns different weights according to this similarity correlation, so as to adjust and optimize the distribution of the associated features between the bidding demand semantic expression and the bidding document document semantic expression, so that when the bidding demand semantic expression is semantically interactively matched with the semantic expressions of each bidding document document, the feature information with high interactive matching correlation is highlighted, and the bidding document text semantic encoding feature vector corresponding to the bidding document with high relevance to the bidding demand is naturally manifested, so that these manifested information can be reflected in each bidding demand-bidding text semantic matching feature vector, and have an important impact on the subsequent recommendation decision. More specifically, the bidding requirement semantic encoding feature vector and each of the bid document text semantic encoding feature vectors in the multiple bid document text semantic encoding feature vectors are sequentially input into a semantic matching module based on an interactive attention layer to obtain the multiple bidding requirement-tender text semantic matching feature vectors, including: processing the bidding requirement semantic encoding feature vector and the multiple bid document text semantic encoding feature vectors using the following interactive attention matching formula to obtain the multiple bidding requirement-tender text semantic matching feature vectors; wherein the interactive attention matching formula is:

[0037] e i =a i T v

[0038]

[0039] Among them, a i is the i-th bidding document text semantic coding feature vector among the multiple bidding document text semantic coding feature vectors, v is the bidding requirement semantic coding feature vector, e i is the i-th similarity matrix, e k is the kth similarity matrix, l a is the number of semantically encoded feature vectors of the bid text, is the i-th bidding requirement-bidding text semantic matching feature vector among the multiple bidding requirement-bidding text semantic matching feature vectors, (·) T represents the transpose of a vector, and exp(·) represents the exponential operation of a matrix.

[0040] It is worth mentioning that in other specific examples of the present application, the text description of the tender requirements and the bid documents submitted by the multiple bidding companies can also be jointly semantically matched and analyzed in other ways to obtain multiple tender requirements-tender text semantic matching feature vectors, for example: input the text description of the tender requirements and the bid documents submitted by the bidding companies; convert them into corresponding text representations, and use a bag of words model, word embedding or a pre-trained language model to represent the text; select a suitable semantic matching model, such as Siamese network, BERT, RoBERTa, etc., to calculate the semantic similarity between the tender requirement text and the bid text; extract features from the tender requirement text description and each bid text file to obtain their semantic representations. This can be done by inputting the text into the semantic matching model and extracting the output of the middle layer of the model as the feature representation; for each tender requirement text description and the corresponding bid text file, calculate the degree of semantic matching between them. This can be achieved by calculating their cosine similarity, Euclidean distance or using the similarity score output by a neural network model; taking the degree of semantic matching between each tender requirement text and the corresponding bid text file as a feature, generating a tender requirement-tender text semantic matching feature vector; repeating the above steps for multiple tender requirements and corresponding bid text files, generating a semantic matching feature vector for each pair of tender requirement-tender text to obtain the multiple tender requirement-tender text semantic matching feature vectors.

[0041] In particular, the S4 determines the recommendation priorities of the multiple bidding enterprises based on the multiple bidding requirements-bidding text semantic matching feature vectors. In particular, in a specific example of the present application, Figure 4 As shown, the S4 includes: S41, performing feature distribution correction on the multiple bidding requirement-bidding text semantic matching feature vectors to obtain multiple corrected bidding requirement-bidding text semantic matching feature vectors; S42, respectively passing the multiple corrected bidding requirement-bidding text semantic matching feature vectors through a classifier-based recommendation decision maker to obtain multiple probability values; S43, determining the recommendation priority of the multiple bidding companies based on the sorting of the multiple probability values.

[0042] Specifically, the S41 performs feature distribution correction on the multiple bidding requirement-bidding text semantic matching feature vectors to obtain multiple corrected bidding requirement-bidding text semantic matching feature vectors. In the technical solution of the present application, the bidding requirement semantic coding feature vector expresses the coded text semantic features of the text description of the bidding requirement, and the multiple bid text semantic coding feature vectors respectively express the coded text semantic features of the multiple bid text descriptions. In this way, considering the difference in text source semantic distribution between the bidding requirement text description and the bid text description, when using the semantic matching module based on the interactive attention layer to perform semantic matching processing on the bidding requirement semantic coding feature vector and the bid text semantic coding feature vector, it is expected to improve the mapping effect of the bidding requirement semantic coding feature vector and the bid text semantic coding feature vector to the feature distribution domain of the bidding requirement-bidding text semantic matching feature vector as a semantic matching fusion. Therefore, the applicant of the present application performs semantic matching fusion correction on the bidding requirement semantic coding feature vector and the bid text semantic coding feature vector, which is specifically expressed as:

[0043]

[0044] Wherein, V1 is the semantic coding feature vector of the bidding requirement, and V2 is the semantic coding feature vector of the tender document text, μ1 and σ1 are the mean and standard deviation of the semantic coding feature vector of the bidding requirement, μ2 and σ2 are the mean and standard deviation of the semantic coding feature vector of the tender document text, respectively. represents the position-by-position square root of the eigenvector, and log is the logarithm with base 2, V c is the corrected eigenvector, Represents addition by position, ⊙ represents point multiplication by position, and log represents logarithmic operation with base 2. Here, in order to improve the mapping effect of the bidding requirement semantic coding feature vector and the bid document text semantic coding feature vector to the semantic matching fusion feature distribution domain in the feature semantic matching scenario, based on the limitations of the traditional weighted fusion method for inferring the semantic space evolution diffusion mode based on feature superposition, a low-order superposition fusion mode and a high-order superposition fusion mode of the combined space are adopted, and the statistical feature interaction relationship between the bidding requirement semantic coding feature vector and the bid document text semantic coding feature vector is used to simulate the evolution center and evolution trajectory, so as to reconstruct the semantic space evolution diffusion in the fusion scenario based on asynchronous evolution under the action of different evolution diffusion speed fields, and effectively improve the projection effect into the same high-dimensional feature space. In this way, the corrected feature vector V cBy performing point multiplication weighting with the bidding requirement-bidding text semantic matching feature vector, effective semantic matching based on semantic representation alignment of the bidding requirement semantic encoding feature vector and the bid text semantic encoding feature vector is achieved, thereby improving the expression effect of the bidding requirement-bidding text semantic matching feature vector and improving the accuracy of the classification result obtained through the classifier.

[0045] Specifically, in S42, the multiple corrected bidding requirements-bidding text semantic matching feature vectors are respectively passed through a classifier-based recommendation decision maker to obtain multiple probability values. More specifically, the corrected bidding requirements-bidding text semantic matching feature vectors are fully connected encoded using multiple fully connected layers of the classifier to obtain encoded classification feature vectors; and the encoded classification feature vectors are passed through the Softmax classification function of the classifier to obtain the classification result. A classifier refers to a machine learning model or algorithm that is used to classify input data into different categories or labels. A classifier is a part of supervised learning, which performs classification tasks by learning the mapping relationship from input data to output categories.

[0046] A fully connected layer is a common type of layer in neural networks. In a fully connected layer, each neuron is connected to all neurons in the previous layer, and each connection has a weight. This means that each neuron in the fully connected layer receives inputs from all neurons in the previous layer, and performs a weighted sum of these inputs through the weights, and then passes the result to the next layer.

[0047] The Softmax classification function is a commonly used activation function for multi-classification problems. It converts each element of the input vector into a probability value between 0 and 1, and the sum of these probability values ​​is equal to 1. The Softmax function is often used in the output layer of a neural network, especially for multi-classification problems, because it can map the network output to the probability distribution of each category. During the training process, the output of the Softmax function can be used to calculate the loss function and update the network parameters through the back propagation algorithm. It is worth noting that the output of the Softmax function does not change the relative size relationship between the elements, but only normalizes them. Therefore, the Softmax function does not change the characteristics of the input vector, but only converts it into a probability distribution form.

[0048] Specifically, the S43 determines the recommendation priority of the multiple bidding companies based on the sorting of the multiple probability values. That is, in a specific example of the present application, for each bidding demand, the bidding companies are sorted according to their probability values, and the companies are arranged from high to low according to the probability values. In this way, the recommendation priority can be determined, and the company with a higher probability value ranks higher; for each bidding demand, their recommendation priority is determined based on the sorted list of bidding companies. Companies with higher rankings are considered to be more in line with the bidding requirements and have a higher recommendation priority. It is worth mentioning that in the actual application scenario of the present application, the decision-making process is displayed in a visual way, thereby increasing the transparency of the decision-making and improving the trust of the company.

[0049] It is worth mentioning that in other specific examples of the present application, the recommendation priority of the multiple bidding enterprises can also be determined based on the semantic matching feature vectors of the multiple bidding requirements-bidding texts in other ways, for example: input the semantic matching feature vectors of the multiple bidding requirements-bidding texts; for each bidding requirement, use cosine similarity or other similarity measurement methods to calculate the similarity between its semantic matching feature vector and the feature vector of each bidding enterprise; for each bidding requirement, sort the bidding enterprises according to the calculated similarity values, and arrange them from high to low according to the similarity, so as to determine the recommendation priority, and the higher the similarity, the higher the ranking of the enterprise; for each bidding requirement, determine their recommendation priority according to the sorted bidding enterprise list. The top-ranked enterprises are considered to be more in line with the bidding requirements and have a higher recommendation priority; repeat the above steps, perform similarity calculation, sorting and recommendation priority determination on all bidding requirements to determine the recommendation priority of multiple bidding enterprises corresponding to each bidding requirement; finally output the recommendation priority list of multiple bidding enterprises corresponding to each bidding requirement. In this way, it can be used in the recommendation system to help decision makers better select suitable bidding enterprises.

[0050] In summary, according to the embodiment of the present application, a visual intelligent decision-making method based on a digital platform is explained, which constructs a digital intelligent decision-making system by utilizing natural language processing technology and intelligent algorithms, and performs semantic understanding and matching of the text description of the bidding requirements and the bid documents submitted by multiple bidding companies, thereby realizing an in-depth comparison between the bidding requirements and the bid documents, and intelligently judging and deciding the matching degree between the bidding requirements and the bid documents, thereby improving the efficiency of the bidding process and helping companies find the most suitable partners.

[0051] Furthermore, a visual intelligent decision-making system based on a digital platform is also provided.

[0052] Figure 5 FIG. 1 is a block diagram of a visual intelligent decision-making system based on a digital platform according to an embodiment of the present application. Figure 5 As shown, according to the visual intelligent decision-making system 300 based on a digital platform according to the embodiment of the present application, it includes: an information collection module 310, which is used to obtain a text description of the bidding requirements; a tender document extraction module 320, which is used to extract the tender documents submitted by multiple bidding companies from the background database; a joint semantic matching analysis module 330, which is used to perform a joint semantic matching analysis on the text description of the bidding requirements and the tender documents submitted by the multiple bidding companies to obtain multiple bidding requirements-bidding text semantic matching feature vectors; a priority recommendation module 340, which is used to determine the recommendation priority of the multiple bidding companies based on the multiple bidding requirements-bidding text semantic matching feature vectors.

[0053] As described above, the visual intelligent decision-making system 300 based on a digital platform according to an embodiment of the present application can be implemented in various wireless terminals, such as a server with a visual intelligent decision-making algorithm based on a digital platform. In one possible implementation, the visual intelligent decision-making system 300 based on a digital platform according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the visual intelligent decision-making system 300 based on a digital platform can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the visual intelligent decision-making system 300 based on a digital platform can also be one of the many hardware modules of the wireless terminal.

[0054] Alternatively, in another example, the digital platform-based visual intelligent decision-making system 300 and the wireless terminal may also be separate devices, and the digital platform-based visual intelligent decision-making system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0055] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A visual intelligent decision-making method based on a digital platform, characterized in that: include: Get the text description of the bidding requirements; Extract the tender documents submitted by multiple bidding companies from the backend database; Performing a joint semantic matching analysis on the text description of the bidding requirements and the bid documents submitted by the multiple bidding enterprises to obtain multiple bidding requirements-bidding text semantic matching feature vectors; Determining the recommendation priorities of the multiple bidding enterprises based on the multiple bidding requirements-bidding text semantic matching feature vectors; The text description of the bidding requirements and the bid documents submitted by the multiple bidding enterprises are subjected to joint semantic matching analysis to obtain multiple bidding requirements-bidding text semantic matching feature vectors, including: Extracting text parts from the bid documents submitted by the multiple bidding enterprises to obtain multiple bid text descriptions; Performing semantic coding on the text description of the bidding requirement to obtain a bidding requirement semantic coding feature vector; Semantically encoding each of the plurality of tender text descriptions to obtain a plurality of tender text semantic encoding feature vectors; Performing semantic interactive matching on the bidding requirement semantic coding feature vector and the plurality of bidding document text semantic coding feature vectors to obtain the bidding requirement-bidding document semantic matching feature vector; The step of performing semantic interactive matching on the bidding requirement semantic coding feature vector and the plurality of bidding document text semantic coding feature vectors to obtain the bidding requirement-bidding document semantic matching feature vector includes: Inputting the bidding requirement semantic encoding feature vector and each bidding document text semantic encoding feature vector in the multiple bidding document text semantic encoding feature vectors into a semantic matching module based on an interactive attention layer in sequence to obtain the multiple bidding requirement-bidding text semantic matching feature vectors; The bidding requirement semantic coding feature vector and each of the bidding document text semantic coding feature vectors in the multiple bidding document text semantic coding feature vectors are sequentially input into a semantic matching module based on an interactive attention layer to obtain the multiple bidding requirement-bidding text semantic matching feature vectors, including: The bidding requirement semantic encoding feature vector and the multiple bidding document text semantic encoding feature vectors are processed using the following interactive attention matching formula to obtain the multiple bidding requirement-bidding text semantic matching feature vectors; wherein the interactive attention matching formula is: yes i =a i T v Among them, a i is the i-th bidding document text semantic coding feature vector among the multiple bidding document text semantic coding feature vectors, v is the bidding requirement semantic coding feature vector, e i is the i-th similarity matrix, e k is the kth similarity matrix, l a is the number of semantically encoded feature vectors of the bid text, is the i-th bidding requirement-bidding text semantic matching feature vector among the multiple bidding requirement-bidding text semantic matching feature vectors, (·) T represents the transpose of a vector, and exp(·) represents the exponential operation of a matrix; The step of determining the recommendation priorities of the plurality of bidding enterprises based on the plurality of bidding requirements-bidding text semantic matching feature vectors includes: Performing feature distribution correction on the plurality of bidding requirement-bidding text semantic matching feature vectors to obtain a plurality of corrected bidding requirement-bidding text semantic matching feature vectors; Passing the plurality of corrected bidding requirements-bidding text semantic matching feature vectors through a classifier-based recommendation decision maker to obtain a plurality of probability values; Determining the recommendation priorities of the multiple bidding enterprises based on the ranking of the multiple probability values; Among them, the semantic matching fusion correction is performed on the semantic coding feature vector of the bidding requirements and the semantic coding feature vector of the bidding document text, which is specifically expressed as follows: Wherein, V1 is the semantic coding feature vector of the bidding requirement, and V2 is the semantic coding feature vector of the tender document text, μ1 and σ1 are the mean and standard deviation of the semantic coding feature vector of the bidding requirement, μ2 and σ2 are the mean and standard deviation of the semantic coding feature vector of the tender document text, respectively. represents the position-by-position square root of the eigenvector, and log is the logarithm with base 2, V c is the corrected eigenvector, represents positional addition, ⊙ represents positional multiplication, and log represents logarithmic operation with base 2; The corrected eigenvector V c The bidding requirement-bidding text semantic matching feature vector is weighted by point multiplication to obtain a corrected bidding requirement-bidding text semantic matching feature vector.

2. The visual intelligent decision-making method based on a digital platform according to claim 1 is characterized in that: Semantically encoding the text description of the bidding requirement to obtain a bidding requirement semantic encoding feature vector, including: Performing word segmentation processing on the text description of the bidding requirements to convert the text description of the bidding requirements into a word sequence consisting of multiple words; Using an embedding layer of a semantic encoder including an embedding layer, each word in the word sequence is mapped to a word embedding vector to obtain a sequence of word embedding vectors; Using the converter of the semantic encoder including the embedding layer, performing converter-based global context semantic encoding on the sequence of word embedding vectors to obtain multiple global context semantic feature vectors; and The multiple global context semantic feature vectors are cascaded to obtain the bidding requirement semantic encoding feature vector.

3. The visual intelligent decision-making method based on a digital platform according to claim 2 is characterized in that: Using the converter of the semantic encoder including the embedding layer to perform global context semantic encoding based on the converter idea on the sequence of word embedding vectors to obtain multiple global context semantic feature vectors, including: Arranging the sequence of word embedding vectors in one dimension to obtain a global feature vector; Calculating the product of the global feature vector and the transposed vector of each word embedding vector in the sequence of word embedding vectors to obtain a plurality of self-attention association matrices; Normalizing each of the plurality of self-attention association matrices to obtain a plurality of standardized self-attention association matrices; Passing each of the plurality of standardized self-attention association matrices through a Softmax classification function to obtain a plurality of probability values; Using each probability value in the plurality of probability values ​​as a weight, weighting each word embedding vector in the sequence of word embedding vectors to obtain the plurality of contextual semantic feature vectors; and The multiple context semantic feature vectors are cascaded to obtain the multiple global context semantic feature vectors.

4. The visual intelligent decision-making method based on a digital platform according to claim 3 is characterized in that: The plurality of corrected bidding requirements-bidding text semantic matching feature vectors are respectively passed through a classifier-based recommendation decision maker to obtain a plurality of probability values, including: Using multiple fully connected layers of the classifier to perform fully connected encoding on the corrected bidding requirement-bidding text semantic matching feature vector to obtain an encoded classification feature vector; and The encoded classification feature vector is passed through the Softmax classification function of the classifier to obtain a classification result.

5. A visual intelligent decision-making system based on a digital platform, characterized in that: include: Information collection module, used to obtain text descriptions of bidding requirements; The tender document extraction module is used to extract the tender documents submitted by multiple bidding companies from the backend database; A joint semantic matching analysis module, used for performing a joint semantic matching analysis on the text description of the bidding requirements and the bid documents submitted by the multiple bidding enterprises to obtain multiple bidding requirements-bidding text semantic matching feature vectors; A priority recommendation module, used to determine the recommendation priorities of the multiple bidding enterprises based on the multiple bidding requirements-bidding text semantic matching feature vectors; The text description of the bidding requirements and the bid documents submitted by the multiple bidding enterprises are subjected to joint semantic matching analysis to obtain multiple bidding requirements-bidding text semantic matching feature vectors, including: Extracting text parts from the bid documents submitted by the multiple bidding enterprises to obtain multiple bid text descriptions; Performing semantic coding on the text description of the bidding requirement to obtain a bidding requirement semantic coding feature vector; Semantically encoding each of the plurality of tender text descriptions to obtain a plurality of tender text semantic encoding feature vectors; Performing semantic interactive matching on the bidding requirement semantic coding feature vector and the plurality of bidding document text semantic coding feature vectors to obtain the bidding requirement-bidding document semantic matching feature vector; The step of performing semantic interactive matching on the bidding requirement semantic coding feature vector and the plurality of bidding document text semantic coding feature vectors to obtain the bidding requirement-bidding document semantic matching feature vector includes: Inputting the bidding requirement semantic encoding feature vector and each bidding document text semantic encoding feature vector in the multiple bidding document text semantic encoding feature vectors into a semantic matching module based on an interactive attention layer in sequence to obtain the multiple bidding requirement-bidding text semantic matching feature vectors; The bidding requirement semantic coding feature vector and each of the bidding document text semantic coding feature vectors in the multiple bidding document text semantic coding feature vectors are sequentially input into a semantic matching module based on an interactive attention layer to obtain the multiple bidding requirement-bidding text semantic matching feature vectors, including: The bidding requirement semantic encoding feature vector and the multiple bidding document text semantic encoding feature vectors are processed using the following interactive attention matching formula to obtain the multiple bidding requirement-bidding text semantic matching feature vectors; wherein the interactive attention matching formula is: yes i =a i T v Among them, a i is the i-th bidding document text semantic coding feature vector among the multiple bidding document text semantic coding feature vectors, v is the bidding requirement semantic coding feature vector, e i is the i-th similarity matrix, e k is the kth similarity matrix, l a is the number of semantically encoded feature vectors of the bid text, is the i-th bidding requirement-bidding text semantic matching feature vector among the multiple bidding requirement-bidding text semantic matching feature vectors, (·) T represents the transpose of a vector, and exp(·) represents the exponential operation of a matrix; The step of determining the recommendation priorities of the plurality of bidding enterprises based on the plurality of bidding requirements-bidding text semantic matching feature vectors includes: Performing feature distribution correction on the plurality of bidding requirement-bidding text semantic matching feature vectors to obtain a plurality of corrected bidding requirement-bidding text semantic matching feature vectors; Passing the plurality of corrected bidding requirements-bidding text semantic matching feature vectors through a classifier-based recommendation decision maker to obtain a plurality of probability values; Determining the recommendation priorities of the multiple bidding enterprises based on the ranking of the multiple probability values; Among them, the semantic matching fusion correction is performed on the semantic coding feature vector of the bidding requirements and the semantic coding feature vector of the bidding document text, which is specifically expressed as follows: Wherein, V1 is the semantic coding feature vector of the bidding requirement, and V2 is the semantic coding feature vector of the tender document text, μ1 and σ1 are the mean and standard deviation of the semantic coding feature vector of the bidding requirement, μ2 and σ2 are the mean and standard deviation of the semantic coding feature vector of the tender document text, respectively. represents the position-by-position square root of the eigenvector, and log is the logarithm with base 2, V c is the corrected eigenvector, represents positional addition, ⊙ represents positional multiplication, and log represents logarithmic operation with base 2; The corrected eigenvector V c The bidding requirement-bidding text semantic matching feature vector is weighted by point multiplication to obtain a corrected bidding requirement-bidding text semantic matching feature vector.

Images