Method and device for retrieving bidding data, equipment and storage medium

Abstract

This application provides a method, apparatus, device, and storage medium for retrieving bidding and tendering data, relating to the field of data retrieval technology. In this application, firstly, bidding and tendering data retrieval request information is obtained; secondly, a bidding and tendering data map pre-constructed according to the bidding and tendering data retrieval requirements is obtained; then, for each bidding and tendering data sub-map in the bidding and tendering data map, it is determined whether the sub-map matches the bidding and tendering data retrieval request information; finally, based on the text block corresponding to each matching bidding and tendering data sub-map, the bidding and tendering data retrieval data corresponding to the bidding and tendering data retrieval request information is determined. Based on the above, the relatively low reliability of bidding and tendering data retrieval in the prior art can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of data retrieval, and in particular to a bidding data retrieval method and device, equipment and storage medium. Background Art

[0002] The bidding and winning data refers to the explanatory data about the bidding and winning of various projects, including the bidding data (such as project content, bidding amount, etc.) and winning data (such as winning amount, etc.) of each project. In order to facilitate the user's bidding and winning data retrieval, a corresponding database is generally established to store the bidding and winning data, and then the words in the user's search request are simply compared with the words in the bidding and winning data, and then the bidding and winning data corresponding to the hit words are returned to complete the corresponding search. Such a simple comparison is prone to the problem that the reliability of bidding and winning data retrieval is relatively low. Summary of the invention

[0003] In view of this, the purpose of the present application is to provide a bidding winning data retrieval method and device, equipment and storage medium to improve the problem of relatively low reliability of bidding winning data retrieval in the prior art.

[0004] To achieve the above objectives, this application adopts the following technical solutions:

[0005] A bidding data retrieval method, comprising:

[0006] Obtain the bidding data retrieval request information;

[0007] Obtain the bidding and tendering data map that is pre-built for bidding and tendering data retrieval needs;

[0008] For each bidding and winning data sub-graph in the bidding and winning data graph, determine whether the bidding and winning data sub-graph matches the bidding and winning data retrieval request information, wherein the data entities and the relationship entities representing the relationship between the data entities included in each of the bidding and winning data sub-graphs are determined based on corresponding text blocks, and each of the text blocks is determined based on at least one bidding and winning project data or based on at least part of the bidding and winning project data;

[0009] Based on the text block corresponding to each tender winning bid data sub-graph matched by the tender winning bid data retrieval request information, the tender winning bid retrieval data corresponding to the tender winning bid data retrieval request information is determined, wherein the tender winning bid retrieval data serves as the retrieval result of the tender winning bid data retrieval request information.

[0010] In a preferred embodiment of the present application, in the above-mentioned bidding winning data retrieval method, the step of determining, for each bidding winning data sub-graph in the bidding winning data graph, whether the bidding winning data sub-graph matches the bidding winning data retrieval request information includes:

[0011] For each bidding winning data sub-graph in the bidding winning data graph, semantic mining processing is performed on the bidding winning data sub-graph, and bidding winning data features corresponding to the bidding winning data sub-graph are output, wherein each of the bidding winning data features includes entity sub-features corresponding to each data entity in the bidding winning data sub-graph and entity sub-features corresponding to each relationship entity;

[0012] Transmit and aggregate the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding and winning data features corresponding to the bidding and winning data sub-graph, and output the bidding and winning data aggregate features corresponding to the bidding and winning data sub-graph;

[0013] Performing semantic mining processing on the bidding data retrieval request information, and outputting retrieval request information features corresponding to the bidding data retrieval request information;

[0014] For each of the bidding winning data sub-graphs, based on the bidding winning data aggregation features corresponding to the bidding winning data sub-graph and the retrieval request information features corresponding to the bidding winning data retrieval request information, determine whether the bidding winning data sub-graph matches the bidding winning data retrieval request information.

[0015] In a preferred embodiment of the present application, in the above-mentioned bidding winning data retrieval method, the step of transferring and aggregating the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding winning data features corresponding to the bidding winning data sub-graph, and outputting the bidding winning data aggregate features corresponding to the bidding winning data sub-graph includes:

[0016] For each data entity in the bidding data sub-graph, the data entity is determined as a starting data entity, and wanders in the bidding data sub-graph to form an entity wandering link network corresponding to the data entity, wherein the length of the shortest wandering path between each other entity and the starting data entity in the entity wandering link network is used as the relationship deviation level of each other entity compared with the starting data entity;

[0017] For each of the entity wandering link networks, the entity sub-features corresponding to each entity are transferred along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity, so as to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network;

[0018] The entity transfer sub-features corresponding to each data entity in the bidding and winning data sub-graph are aggregated, and the bidding and winning data aggregate features corresponding to the bidding and winning data sub-graph are output.

[0019] In a preferred embodiment of the present application, in the above-mentioned bidding winning data retrieval method, for each of the entity walking link networks, the entity sub-features corresponding to each entity are transferred along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity walking link network, including:

[0020] For each other entity in the entity wandering link network whose relationship deviates from the highest level, based on the entity sub-feature corresponding to the other entity, determining the entity transfer sub-feature corresponding to the other entity;

[0021] For each other entity at each relationship deviation level other than the highest relationship deviation level in the entity wandering link network, cross-attention processing is performed on the entity sub-features corresponding to each other entity at the current relationship deviation level based on the entity transfer sub-features corresponding to each other entity at a higher relationship deviation level with a connection relationship, so as to obtain the entity transfer sub-features corresponding to each other entity at the current relationship deviation level, until the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network are obtained, wherein the starting data entity belongs to other entities at the lowest relationship deviation level.

[0022] In a preferred embodiment of the present application, in the above-mentioned bidding winning data retrieval method, the step of determining the entity transfer sub-feature corresponding to each other entity with the highest relationship deviation level in the entity wandering link network based on the entity sub-feature corresponding to the other entity includes:

[0023] For each bidding winning data sub-graph in the bidding winning data graph, in the bidding winning data graph, determine another bidding winning data sub-graph with the greatest correlation with the bidding winning data sub-graph as the corresponding associated bidding winning data sub-graph, and perform splicing and pooling processing on the entity sub-features corresponding to each entity in the associated bidding winning data sub-graph, and output the pooled splicing features corresponding to the bidding winning data sub-graph, wherein the correlation between the bidding winning data sub-graphs is determined based on the correlation between the corresponding text blocks;

[0024] For each other entity whose relationship deviates from the highest level in the entity wandering link network, cross-attention processing is performed on the entity sub-features corresponding to the other entity based on the pooled splicing features corresponding to the bidding data sub-graph where the other entity is located, and a weighted superposition operation is performed on the result of the cross-attention processing and the entity sub-features corresponding to the other entity to obtain the entity transfer sub-features corresponding to the other entity.

[0025] In a preferred embodiment of the present application, in the above-mentioned bidding winning data retrieval method, the step of aggregating the entity transfer sub-features corresponding to each data entity in the bidding winning data sub-graph and outputting the bidding winning data aggregate features corresponding to the bidding winning data sub-graph includes:

[0026] For each of the bidding data sub-graphs, the entity transfer sub-features corresponding to each data entity in the bidding data sub-graph are spliced, the entity transfer splicing features corresponding to the bidding data sub-graph are output, and each other bidding data sub-graph having an associated relationship with the bidding data sub-graph is determined as the corresponding associated bidding data sub-graph;

[0027] Based on the entity transfer splicing features corresponding to each of the associated bidding data sub-graphs, cross-attention processing is performed on the entity transfer splicing features corresponding to the bidding data sub-graphs, and each cross-attention feature corresponding to the bidding data sub-graphs is output;

[0028] Based on the entity transfer splicing features corresponding to the bidding winning data sub-graph and each corresponding cross-attention feature, the bidding winning data aggregation features corresponding to the bidding winning data sub-graph are determined.

[0029] In a preferred embodiment of the present application, in the above-mentioned bidding winning data retrieval method, the step of determining, for each of the bidding winning data sub-graphs, whether the bidding winning data sub-graph matches the bidding winning data retrieval request information based on the bidding winning data aggregation features corresponding to the bidding winning data sub-graph and the retrieval request information features corresponding to the bidding winning data retrieval request information, comprises:

[0030] For each of the bidding winning data sub-graphs, determine the feature similarity between the bidding winning data aggregation feature corresponding to the bidding winning data sub-graph and the retrieval request information feature corresponding to the bidding winning data retrieval request information, and use it as the request matching degree corresponding to the bidding winning data sub-graph;

[0031] For each of the tender winning bid data sub-graph, the request matching degree corresponding to the tender winning bid data sub-graph is compared with a predetermined reference matching degree, and when the request matching degree is greater than the reference matching degree, it is determined that the tender winning bid data sub-graph matches the tender winning bid data retrieval request information, and when the request matching degree is not greater than the reference matching degree, it is determined that the tender winning bid data sub-graph does not match the tender winning bid data retrieval request information.

[0032] The present application also provides a bidding data retrieval device, comprising:

[0033] The retrieval request acquisition module is used to obtain the bidding data retrieval request information;

[0034] The data graph acquisition module is used to obtain the bidding and winning data graph pre-built for the bidding and winning data retrieval needs;

[0035] A graph request matching module, for determining, for each bidding and winning data sub-graph in the bidding and winning data graph, whether the bidding and winning data sub-graph matches the bidding and winning data retrieval request information, wherein the data entities and the relationship entities representing the relationship between the data entities included in each of the bidding and winning data sub-graphs are determined based on corresponding text blocks, and each of the text blocks is determined based on at least one bidding and winning project data or based on at least part of the bidding and winning project data;

[0036] A retrieval data determination module is used to determine the bidding winning bid retrieval data corresponding to the bidding winning bid data retrieval request information based on the text block corresponding to each bidding winning bid data sub-graph matched by the bidding winning bid data retrieval request information, wherein the bidding winning bid retrieval data serves as the retrieval result of the bidding winning bid data retrieval request information.

[0037] Based on the above, the present application also provides an electronic device, including:

[0038] Memory for storing computer programs;

[0039] The processor connected to the memory is used to execute the computer program stored in the memory to implement the above-mentioned bidding winning data retrieval method.

[0040] On the basis of the above, the present application also provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is run, each step of the above-mentioned bidding winning data retrieval method is executed.

[0041] The bidding winning data retrieval method, device, equipment and storage medium provided in the present application first obtain the bidding winning data retrieval request information; secondly, obtain the bidding winning data map constructed in advance for the bidding winning data retrieval requirements; then, for each bidding winning data sub-map in the bidding winning data map, determine whether the bidding winning data sub-map matches the bidding winning data retrieval request information; finally, based on the text block corresponding to each bidding winning data sub-map matched by the bidding winning data retrieval request information, determine the bidding winning retrieval data corresponding to the bidding winning data retrieval request information. Based on the above content, since the bidding winning data map will be constructed based on the bidding winning project data, the correlation between different entities is fully characterized. Therefore, when performing matching analysis, on the basis of considering the semantics of the entity itself, the relationship between entities is also fully considered. Therefore, the reliability of the matching analysis can be improved, thereby ensuring the reliability of the determined bidding winning retrieval data, thereby improving the problem of relatively low reliability of bidding winning data retrieval in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings.

[0043] Figure 1 A structural block diagram of an electronic device provided in an embodiment of the present application.

[0044] Figure 2 A flowchart of a bidding winning data retrieval method provided in an embodiment of the present application.

[0045] Figure 3 A schematic diagram of the bidding data sub-graph provided in an embodiment of the present application.

[0046] Figure 4 A schematic diagram of a physical wandering link network provided in an embodiment of the present application.

[0047] Figure 5 A block diagram of a bidding winning data retrieval device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0049] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0050] like Figure 1 As shown, an embodiment of the present application provides an electronic device, wherein the electronic device may include a memory, a processor, and a bidding winning data retrieval device.

[0051] In detail, the memory and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, the memory and the processor can be electrically connected through one or more communication buses or signal lines. The bidding winning data retrieval device includes at least one software function module stored in the memory in the form of software or firmware. The processor is used to execute an executable computer program stored in the memory, for example, the software function modules and computer programs included in the bidding winning data retrieval device, so as to implement the bidding winning data retrieval method provided in the embodiment of the present application.

[0052] Optionally, the memory may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc.

[0053] Furthermore, the processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0054] Understandably, Figure 1 The structure shown is for illustration only, and the electronic device may also include Figure 1 More or fewer components as shown, or with Figure 1 The different configurations shown, for example, may also include a communication unit for exchanging information with other devices (such as terminal devices).

[0055] Combination Figure 2 The present application also provides a bidding winning data retrieval method applicable to the electronic device. The method steps defined in the process related to the bidding winning data retrieval method can be implemented by the electronic device. Figure 2 The process shown is described in detail.

[0056] Step S110, obtaining the bidding data retrieval request information.

[0057] In the embodiment of the present application, the electronic device can obtain the bidding data retrieval request information. Exemplarily, the retrieval user can send the bidding data retrieval request information to the electronic device through the corresponding terminal device (such as a mobile phone, computer, etc.), such as the latest winning project of xxx company.

[0058] Step S120, obtaining a bidding and winning data map pre-constructed for bidding and winning data retrieval needs.

[0059] In an embodiment of the present application, the electronic device can obtain a bidding and winning data map that is pre-constructed for bidding and winning data retrieval needs, for example, it can be obtained from a corresponding database.

[0060] Step S130, for each bidding winning data sub-graph in the bidding winning data graph, determine whether the bidding winning data sub-graph matches the bidding winning data retrieval request information.

[0061] In an embodiment of the present application, the electronic device can determine whether the bidding and winning data sub-graph matches the bidding and winning data retrieval request information for each bidding and winning data sub-graph in the bidding and winning data graph. Wherein, each of the bidding and winning data sub-graphs includes data entities (such as bidding units, winning units, project names, etc.) and relationship entities (such as bidding, winning bids, etc.) that characterize the relationship between data entities, which are determined based on corresponding text blocks, and each of the text blocks is determined based on at least one bidding and winning project data or based on at least part of the bidding and winning project data (such as a paragraph in the text, etc.).

[0062] Step S140, based on the text block corresponding to each tender winning data sub-graph matched by the tender winning data retrieval request information, determine the tender winning bid retrieval data corresponding to the tender winning bid data retrieval request information.

[0063] In an embodiment of the present application, the electronic device can determine the bidding winning bid retrieval data corresponding to the bidding winning bid data retrieval request information based on the text block corresponding to each bidding winning bid data sub-graph matched by the bidding winning bid data retrieval request information. The bidding winning bid retrieval data serves as the retrieval result of the bidding winning bid data retrieval request information. Exemplarily, the matched text block can be directly determined as the bidding winning bid retrieval data, or the matched text block can be used as auxiliary data of the bidding winning bid data retrieval request information and loaded into a corresponding neural network model (which can be selected according to actual needs) for corresponding analysis to obtain the required bidding winning bid retrieval data.

[0064] Based on the above content, since a bidding and winning data map will be constructed based on the bidding and winning project data, the correlation between different entities can be fully characterized. Therefore, when performing matching analysis, on the basis of considering the semantics of the entity itself, the relationship between entities is also fully considered. Therefore, the reliability of the matching analysis can be improved, thereby ensuring the reliability of the determined bidding and winning retrieval data, thereby improving the problem of relatively low reliability of bidding and winning data retrieval in the prior art.

[0065] It should be noted that for step S130, the specific method of determining whether the bidding winning data sub-graph matches the bidding winning data retrieval request information is not limited and can be selected and configured according to actual needs.

[0066] For example, in an alternative embodiment, in order to improve the reliability of matching determination so that the subsequently determined bidding winning retrieval data has a higher reliability, the above-mentioned step S130 may further include step S131, step S132, step S133 and step S134, and the specific content of each step is described as follows.

[0067] Step S131, for each bidding winning data sub-graph in the bidding winning data graph, perform semantic mining processing on the bidding winning data sub-graph, and output the bidding winning data features corresponding to the bidding winning data sub-graph.

[0068] In an embodiment of the present application, for each bidding winning data sub-graph in the bidding winning data graph, semantic mining processing can be performed on the bidding winning data sub-graph, and the bidding winning data features corresponding to the bidding winning data sub-graph are output. Each of the bidding winning data features includes entity sub-features corresponding to each data entity in the bidding winning data sub-graph and entity sub-features corresponding to each relationship entity. In other words, semantic mining can be performed on each entity (including data entities and relationship entities) to obtain corresponding semantic features, wherein, in an embodiment of the present application, the expression form of each feature can be a vector. Specifically, each entity can be embedded to obtain an embedded vector, which can be used as the corresponding entity sub-feature. The embedding process can be implemented through a word embedding model. For example, for "the bidding amount is 10 million", the corresponding embedding vector is:

[0069] Embedding vector for “bid” = [0.12, −0.34, 0.56, −0.11, 0.98, −0.67, 0.45, 0.32, −0.56, 0.14, ..., −0.88, 0.77, 0.23, −0.31];

[0070] Embedding vector of “amount” = [0.33, −0.44, 0.12, 0.65, −0.76, 0.22, −0.53, 0.74, 0.19, −0.23, ..., 0.67, 0.09, −0.43, 0.81];

[0071] Embedding vector for “for” = [−0.36, 0.44, 0.22, −0.71, 0.21, −0.79, 0.78, 0.41, −0.34, 0.18, ..., −0.32, 0.67, 0.65, −0.24];

[0072] Embedding vectors for “1000” = [−0.25, 0.54, 0.32, −0.61, 0.21, −0.89, 0.77, 0.41, −0.34, 0.98, ..., −0.12, 0.67, 0.45, −0.23];

[0073] The embedding vector of “wan” = [0.19, −0.34, 0.22, 0.56, −0.49, 0.61, 0.08, −0.76, 0.34, 0.17, ..., −0.12, 0.29, −0.56, 0.61].

[0074] Step S132, transferring and aggregating the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding winning data features corresponding to the bidding winning data sub-graph, and outputting the bidding winning data aggregation features corresponding to the bidding winning data sub-graph.

[0075] In the embodiment of the present application, after obtaining each entity sub-feature, the entity sub-feature corresponding to each data entity and the entity sub-feature corresponding to each relationship entity included in the bidding winning data sub-graph corresponding to the bidding winning data sub-graph can be transferred and aggregated to output the bidding winning data aggregated features corresponding to the bidding winning data sub-graph. In other words, for each bidding winning data sub-graph, the entity sub-feature can be transferred within the bidding winning data sub-graph, so that the associated information within the bidding winning data sub-graph can be fully mined, and the aggregation of important semantics can be achieved, so that the semantic representation ability of the obtained bidding winning data aggregated features is better.

[0076] Step S133, performing semantic mining processing on the bidding data retrieval request information, and outputting retrieval request information features corresponding to the bidding data retrieval request information.

[0077] In an embodiment of the present application, semantic mining processing can also be performed on the bidding data retrieval request information to output the retrieval request information features corresponding to the bidding data retrieval request information. For example, the text corresponding to the bidding data retrieval request information can be embedded to obtain an embedding vector, which is used as the corresponding retrieval request information feature.

[0078] Step S134, for each of the bidding winning data sub-graphs, based on the bidding winning data features corresponding to the bidding winning data sub-graphs and the retrieval request information features corresponding to the bidding winning data retrieval request information, determine whether the bidding winning data sub-graph matches the bidding winning data retrieval request information.

[0079] In an embodiment of the present application, after obtaining the aggregation features of the bidding and winning bid data and the features of the retrieval request information, it is possible to determine, for each of the bidding and winning bid data sub-graphs, whether the bidding and winning bid sub-graph matches the bidding and winning bid data retrieval request information based on the bidding and winning bid data features corresponding to the bidding and winning bid data sub-graph and the retrieval request information features corresponding to the bidding and winning bid data retrieval request information, that is, determine whether they match based on the relationship between the features.

[0080] It can be understood that in the above-mentioned step S132, the specific method of transferring and aggregating the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding and winning data features corresponding to the bidding and winning data sub-graph is not restricted and can be selected according to actual needs. For example, in an alternative embodiment, in order to make the reliability of transfer aggregation higher and the semantic representation accuracy of the obtained bidding and winning data aggregation features better, the above-mentioned step S132 can further include step S132a, step S132b and step S132c, and the specific content of each step is described as follows.

[0081] Step S132a, for each data entity in the bidding data sub-graph, determine the data entity as the starting data entity, and roam in the bidding data sub-graph to form an entity roaming link network corresponding to the data entity.

[0082] In an embodiment of the present application, for each data entity in the bidding and winning data sub-graph, the data entity can be determined as the starting data entity, and wanders in the bidding and winning data sub-graph to form an entity wandering link network corresponding to the data entity. Among them, in the entity wandering link network, the length of the shortest wandering path between each other entity and the starting data entity is used as the relationship deviation level of each other entity compared to the starting data entity. It should be noted that in order to facilitate the problem that the path of information transmission is too long and the data processing is too complicated, the highest level of the relationship deviation level can be limited, such as setting the highest level to 3. Among them. The lowest level is the starting data entity itself, which can be 0. Among them, for Figure 3 The data entity 1 in the bidding data sub-graph shown in the figure is used as the starting data entity to walk, and the entity walking link network formed can be as follows Figure 4 As shown, the relationship deviation level of data entity 1 is 0, the relationship deviation level of relationship entity 1, data entity 2, data entity 5 and data entity 6 is 1, the relationship deviation level of data entity 3, data entity 4, relationship entity 2, relationship entity 3 and relationship entity 4 is 2, and the relationship deviation level of data entity 8 and data entity 7 is 3.

[0083] Step S132b, for each of the entity wandering link networks, the entity sub-features corresponding to each entity are transferred along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity, to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network.

[0084] In the embodiment of the present application, after determining the entity walking link network corresponding to each data entity, for each of the entity walking link networks, the entity sub-features corresponding to each entity can be transferred along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity walking link network. For example, Figure 4 The entity wandering link network shown can be transmitted from data entity 7 and data entity 8 until it is transmitted to data entity 1, and the entity transmission sub-feature corresponding to data entity 1 is obtained to aggregate the semantic information of each entity at data entity 1.

[0085] Step S132c, aggregate the entity transfer sub-features corresponding to each data entity in the bidding and winning data sub-graph, and output the bidding and winning data aggregate features corresponding to the bidding and winning data sub-graph.

[0086] In an embodiment of the present application, after obtaining the entity transfer sub-feature corresponding to each data entity, the entity transfer sub-feature corresponding to each data entity in the bidding data sub-graph can be aggregated to output the bidding data aggregate feature corresponding to the bidding data sub-graph. It should be noted that in some other implementations, the object of transfer may not be limited to data entities, and the entity wandering link network of relationship entities may also be determined to obtain the entity transfer sub-feature corresponding to the relationship entity. Finally, the entity transfer sub-feature corresponding to each data entity in the bidding data sub-graph and the entity transfer sub-feature corresponding to each relationship entity can be aggregated to output the bidding data aggregate feature corresponding to the bidding data sub-graph.

[0087] It can be understood that, in the above-mentioned step S132b, the specific method of transmitting the entity sub-features corresponding to each entity is not limited and can be selected according to actual needs. For example, in an alternative embodiment, in order to obtain entity transmission sub-features with higher semantic representation reliability, the above-mentioned step S132b can further include steps b1 and b2, as described below.

[0088] Step b1, for each other entity with the highest relationship deviation level in the entity wandering link network, based on the entity sub-feature corresponding to the other entity, determine the entity transfer sub-feature corresponding to the other entity.

[0089] In the embodiment of the present application, for each other entity with the highest relationship deviation level in the entity walking link network, the entity transfer sub-feature corresponding to the other entity can be determined based on the entity sub-feature corresponding to the other entity. For example, Figure 4The entity wandering link network shown can first determine the entity transfer sub-features corresponding to data entity 7 and data entity 8 based on the entity sub-features corresponding to data entity 7 and data entity 8 respectively.

[0090] Step b2, for each other entity at each relationship deviation level other than the highest relationship deviation level in the entity wandering link network, cross-attention processing is performed on the entity sub-features corresponding to each other entity at the current relationship deviation level based on the entity transfer sub-features corresponding to each other entity at a higher relationship deviation level with a connection relationship, so as to obtain the entity transfer sub-features corresponding to each other entity at the current relationship deviation level, until the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network are obtained.

[0091] In an embodiment of the present application, for each other entity of each relationship deviation level other than the highest relationship deviation level in the entity wandering link network, based on the entity transfer sub-features corresponding to each other entity of a higher relationship deviation level with a connection relationship, the entity sub-features corresponding to each other entity of the current relationship deviation level are cross-attention processed to obtain the entity transfer sub-features corresponding to each other entity of the current relationship deviation level, until the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network are obtained. Among them, the starting data entity belongs to other entities with the lowest relationship deviation level. For example, for data entity 1, the entity sub-features corresponding to data entity 1 can be cross-attention processed (Cross-Attention) based on the entity transfer sub-features corresponding to relationship entity 1, data entity 2, data entity 3 and data entity 6, respectively, to obtain the entity transfer sub-features corresponding to data entity 1. It should be noted that when there are multiple other entities of a higher relationship deviation level with a connection relationship, after performing cross-attention processing to obtain multiple cross-attention processing results, the multiple cross-attention processing results can be averaged to obtain the corresponding entity transfer sub-features.

[0092] It can be understood that in the above step b1, the specific manner of performing feature transfer for each other entity whose relationship deviates from the highest level in the entity wandering link network is not limited and can be selected according to actual needs. For example, in an alternative implementation, in order to make the semantics of the obtained entity transfer sub-feature richer, the above step b1 may include:

[0093] First, for each bidding data sub-graph in the bidding data graph, another bidding data sub-graph with the greatest correlation with the bidding data sub-graph can be determined in the bidding data graph as the corresponding associated bidding data sub-graph, and the entity sub-features corresponding to each entity in the associated bidding data sub-graph are spliced ​​and pooled (i.e., spliced ​​first, then pooled, such as mean pooling, etc.), and the pooled spliced ​​features corresponding to the bidding data sub-graph are output, wherein the correlation between the bidding data sub-graphs is determined based on the correlation between the corresponding text blocks (the correlation between the text blocks can be determined based on the number of identical words or the proportion of the number, such as the greater the number, the greater the correlation);

[0094] Secondly, for each other entity with the highest relationship deviation level in the entity wandering link network, cross-attention processing can be performed on the entity sub-features corresponding to the other entity based on the pooled splicing features corresponding to the bidding data sub-graph where the other entity is located, and a weighted superposition operation can be performed on the result of the cross-attention processing and the entity sub-features corresponding to the other entity (which can be mean superposition or the weighting coefficient corresponding to the result of the cross-attention processing can be less than the weighting coefficient corresponding to the entity sub-feature) to obtain the entity transfer sub-features corresponding to the other entity.

[0095] It is understandable that in the above step S132c, the specific method of aggregating the entity transfer sub-features corresponding to each data entity in the bidding data sub-graph is not limited and can be selected according to actual needs. For example, in an alternative implementation, in order to ensure the reliability of feature aggregation, the above step S132c can further include the following contents:

[0096] First, for each of the bidding data sub-graphs, the entity transfer sub-features corresponding to each data entity in the bidding data sub-graph can be spliced, and the entity transfer splicing features corresponding to the bidding data sub-graph can be output, and each other bidding data sub-graph that has an association relationship with the bidding data sub-graph can be determined as the corresponding associated bidding data sub-graph; illustratively, whether there is an association relationship can be determined based on whether the corresponding text blocks have the same words, or whether there is an association relationship can be determined based on the feature similarity between the entity transfer splicing features;

[0097] Secondly, based on the entity transfer splicing features corresponding to each of the related bidding data sub-graphs, cross-attention processing can be performed on the entity transfer splicing features corresponding to the bidding data sub-graphs, and each cross-attention feature corresponding to the bidding data sub-graphs can be output. For example, based on the entity transfer splicing features corresponding to the related bidding data sub-graph 1, cross-attention processing is performed on the entity transfer splicing features corresponding to the bidding data sub-graphs to obtain cross-attention feature 1; based on the entity transfer splicing features corresponding to the related bidding data sub-graph 2, cross-attention processing is performed on the entity transfer splicing features corresponding to the bidding data sub-graphs to obtain cross-attention feature 2; based on the entity transfer splicing features corresponding to the related bidding data sub-graph 3, cross-attention processing is performed on the entity transfer splicing features corresponding to the bidding data sub-graphs to obtain cross-attention feature 3;

[0098] Then, based on the entity transfer splicing features corresponding to the bidding data sub-graph and each corresponding cross-attention feature, the bidding winning data aggregation features corresponding to the bidding data sub-graph can be determined; exemplarily, the mean of each cross-attention feature can be calculated first, and then the mean feature and the entity transfer splicing features corresponding to the bidding data sub-graph can be superimposed to obtain the bidding winning data aggregation features corresponding to the bidding data sub-graph.

[0099] It can be understood that, in the above step S134, the specific method of determining whether the bidding winning data sub-graph matches the bidding winning data retrieval request information is not limited and can be selected according to actual needs. For example, in an alternative implementation, the above step S134 can further include the following content:

[0100] First, for each of the bidding data sub-graphs, the feature similarity (such as cosine similarity, etc.) between the bidding data aggregation feature corresponding to the bidding data sub-graph and the retrieval request information feature corresponding to the bidding data retrieval request information can be determined, and used as the request matching degree corresponding to the bidding data sub-graph;

[0101] Secondly, for each of the tender winning bid data sub-graph, the request matching degree corresponding to the tender winning bid data sub-graph is compared with a predetermined reference matching degree (which can be configured according to actual needs, such as 0.5, 0.6, 0.7 and other values), and when the request matching degree is greater than the reference matching degree, it is determined that the tender winning bid data sub-graph matches the tender winning bid data retrieval request information, and when the request matching degree is not greater than the reference matching degree, it is determined that the tender winning bid data sub-graph does not match the tender winning bid data retrieval request information.

[0102] Combination Figure 5The embodiment of the present application also provides a bidding winning data retrieval device applicable to the above electronic device. The bidding winning data retrieval device may include a retrieval request acquisition module, a data graph acquisition module, a graph request matching module and a retrieval data determination module.

[0103] The retrieval request acquisition module is used to obtain the bidding data retrieval request information. In the embodiment of the present application, the retrieval request acquisition module can be used to execute Figure 2 As shown in step S110, the relevant contents of the search request acquisition module can refer to the above description of step S110.

[0104] The data map acquisition module is used to acquire the bidding data map pre-built for the bidding data retrieval requirements. In the embodiment of the present application, the data map acquisition module can be used to execute Figure 2 As shown in step S120, the relevant contents of the data map acquisition module can refer to the above description of step S120.

[0105] The graph request matching module is used to determine whether each bidding data sub-graph in the bidding data graph matches the bidding data retrieval request information, wherein each of the bidding data sub-graphs includes data entities and relationship entities representing the relationship between data entities based on corresponding text blocks, and each of the text blocks is determined based on at least one bidding project data or based on at least part of the bidding project data. In the embodiment of the present application, the graph request matching module can be used to perform Figure 2 As shown in step S130, the relevant contents of the graph request matching module can refer to the above description of step S130.

[0106] The retrieval data determination module is used to determine the bidding winning retrieval data corresponding to the bidding winning data retrieval request information based on the text block corresponding to each bidding winning data sub-graph matched by the bidding winning data retrieval request information, wherein the bidding winning retrieval data is used as the retrieval result of the bidding winning data retrieval request information. In the embodiment of the present application, the retrieval data determination module can be used to perform Figure 2 As shown in step S140, the relevant contents of the retrieval data determination module can refer to the above description of step S140.

[0107] In the embodiment of the present application, corresponding to the above-mentioned bidding winning data retrieval method applied to the electronic device, a computer-readable storage medium is also provided, in which a computer program is stored, and when the computer program is run, each step of the bidding winning data retrieval method is executed. Among them, each step executed when the aforementioned computer program is run will not be repeated here one by one, and reference can be made to the above-mentioned explanation of the bidding winning data retrieval method.

[0108] In summary, the bidding data retrieval method, device, equipment and storage medium provided by the present application first obtain the bidding data retrieval request information; secondly, obtain the bidding data graph constructed in advance for the bidding data retrieval requirements; then, for each bidding data sub-graph in the bidding data graph, determine whether the bidding data sub-graph matches the bidding data retrieval request information; finally, based on the text block corresponding to each bidding data sub-graph matched by the bidding data retrieval request information, determine the bidding retrieval data corresponding to the bidding data retrieval request information. Based on the above content, since the bidding data graph is constructed based on the bidding project data, the correlation between different entities is fully characterized. Therefore, when performing matching analysis, on the basis of considering the semantics of the entity itself, the relationship between entities is also fully considered. Therefore, the reliability of the matching analysis can be improved, thereby ensuring the reliability of the determined bidding retrieval data, thereby improving the problem of relatively low reliability of bidding data retrieval in the prior art.

[0109] In several embodiments provided in the embodiments of the present application, it should be understood that the disclosed device and method can also be implemented in other ways. The device and method embodiments described above are merely schematic, for example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the device, method and computer program product according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0110] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0111] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code. It should be noted that in this article, the term "include", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or device. Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0112] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A bidding data retrieval method, characterized in that: include: Obtain the bidding data retrieval request information; Obtain the bidding and tendering data map that is pre-built for bidding and tendering data retrieval needs; For each bidding and winning data sub-graph in the bidding and winning data graph, determine whether the bidding and winning data sub-graph matches the bidding and winning data retrieval request information, wherein the data entities and the relationship entities representing the relationship between the data entities included in each of the bidding and winning data sub-graphs are determined based on corresponding text blocks, and each of the text blocks is determined based on at least one bidding and winning project data or based on at least part of the bidding and winning project data; Based on the text blocks corresponding to each bidding winning data sub-graph matched by the bidding winning data retrieval request information, determining the bidding winning retrieval data corresponding to the bidding winning data retrieval request information, wherein the bidding winning retrieval data serves as the retrieval result of the bidding winning data retrieval request information; The step of determining, for each bidding winning data sub-graph in the bidding winning data graph, whether the bidding winning data sub-graph matches the bidding winning data search request information comprises: For each bidding winning data sub-graph in the bidding winning data graph, semantic mining processing is performed on the bidding winning data sub-graph, and bidding winning data features corresponding to the bidding winning data sub-graph are output, wherein each of the bidding winning data features includes entity sub-features corresponding to each data entity in the bidding winning data sub-graph and entity sub-features corresponding to each relationship entity; Transmit and aggregate the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding and winning data features corresponding to the bidding and winning data sub-graph, and output the bidding and winning data aggregate features corresponding to the bidding and winning data sub-graph; Performing semantic mining processing on the bidding data retrieval request information, and outputting retrieval request information features corresponding to the bidding data retrieval request information; For each of the bidding winning data sub-graphs, based on the bidding winning data aggregation features corresponding to the bidding winning data sub-graphs and the retrieval request information features corresponding to the bidding winning data retrieval request information, determining whether the bidding winning data sub-graph matches the bidding winning data retrieval request information; The step of transferring and aggregating the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding winning data features corresponding to the bidding winning data sub-graph, and outputting the bidding winning data aggregate features corresponding to the bidding winning data sub-graph comprises: For each data entity in the bidding data sub-graph, the data entity is determined as a starting data entity, and wanders in the bidding data sub-graph to form an entity wandering link network corresponding to the data entity, wherein the length of the shortest wandering path between each other entity and the starting data entity in the entity wandering link network is used as the relationship deviation level of each other entity compared with the starting data entity; For each of the entity wandering link networks, the entity sub-features corresponding to each entity are transferred along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity, so as to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network; Aggregate the entity transfer sub-features corresponding to each data entity in the bidding data sub-graph, and output the bidding data aggregation features corresponding to the bidding data sub-graph; The step of transferring the entity sub-features corresponding to each entity along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity for each entity wandering link network to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network includes: For each other entity in the entity wandering link network whose relationship deviates from the highest level, based on the entity sub-feature corresponding to the other entity, determining the entity transfer sub-feature corresponding to the other entity; For each other entity of each relation deviation level other than the highest relation deviation level in the entity wandering link network, cross-attention processing is performed on the entity sub-features corresponding to each other entity of the current relation deviation level based on the entity transfer sub-features corresponding to each other entity of a higher relation deviation level with a connection relationship, so as to obtain the entity transfer sub-features corresponding to each other entity of the current relation deviation level, until the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network are obtained, wherein the starting data entity belongs to the other entities with the lowest relation deviation level; The step of determining, for each other entity with the highest relationship deviation level in the entity wandering link network, the entity transfer sub-feature corresponding to the other entity based on the entity sub-feature corresponding to the other entity, comprises: For each bidding winning data sub-graph in the bidding winning data graph, in the bidding winning data graph, determine another bidding winning data sub-graph with the greatest correlation with the bidding winning data sub-graph as the corresponding associated bidding winning data sub-graph, and perform splicing and pooling processing on the entity sub-features corresponding to each entity in the associated bidding winning data sub-graph, and output the pooled splicing features corresponding to the bidding winning data sub-graph, wherein the correlation between the bidding winning data sub-graphs is determined based on the correlation between the corresponding text blocks; For each other entity whose relationship deviates from the highest level in the entity wandering link network, cross-attention processing is performed on the entity sub-features corresponding to the other entity based on the pooled splicing features corresponding to the bidding data sub-graph where the other entity is located, and a weighted superposition operation is performed on the result of the cross-attention processing and the entity sub-features corresponding to the other entity to obtain the entity transfer sub-features corresponding to the other entity.

2. The bidding winning data retrieval method according to claim 1 is characterized in that: The step of aggregating the entity transfer sub-features corresponding to each data entity in the bidding and winning data sub-graph and outputting the bidding and winning data aggregated features corresponding to the bidding and winning data sub-graph comprises: For each of the bidding data sub-graphs, the entity transfer sub-features corresponding to each data entity in the bidding data sub-graph are spliced, the entity transfer splicing features corresponding to the bidding data sub-graph are output, and each other bidding data sub-graph having an associated relationship with the bidding data sub-graph is determined as the corresponding associated bidding data sub-graph; Based on the entity transfer splicing features corresponding to each of the associated bidding data sub-graphs, cross-attention processing is performed on the entity transfer splicing features corresponding to the bidding data sub-graphs, and each cross-attention feature corresponding to the bidding data sub-graphs is output; Based on the entity transfer splicing features corresponding to the bidding winning data sub-graph and each corresponding cross-attention feature, the bidding winning data aggregation features corresponding to the bidding winning data sub-graph are determined.

3. The bidding winning data retrieval method according to any one of claims 1 to 2, characterized in that: The step of determining, for each of the bidding winning data sub-graphs, whether the bidding winning data sub-graph matches the bidding winning data retrieval request information based on the bidding winning data aggregation features corresponding to the bidding winning data sub-graph and the retrieval request information features corresponding to the bidding winning data retrieval request information, comprises: For each of the bidding winning data sub-graphs, determine the feature similarity between the bidding winning data aggregation feature corresponding to the bidding winning data sub-graph and the retrieval request information feature corresponding to the bidding winning data retrieval request information, and use it as the request matching degree corresponding to the bidding winning data sub-graph; For each of the tender winning bid data sub-graph, the request matching degree corresponding to the tender winning bid data sub-graph is compared with a predetermined reference matching degree, and when the request matching degree is greater than the reference matching degree, it is determined that the tender winning bid data sub-graph matches the tender winning bid data retrieval request information, and when the request matching degree is not greater than the reference matching degree, it is determined that the tender winning bid data sub-graph does not match the tender winning bid data retrieval request information.

4. A bidding data retrieval device, characterized in that: include: The retrieval request acquisition module is used to obtain the bidding data retrieval request information; The data graph acquisition module is used to obtain the bidding and winning data graph pre-built for the bidding and winning data retrieval needs; A graph request matching module, for determining, for each bidding and winning data sub-graph in the bidding and winning data graph, whether the bidding and winning data sub-graph matches the bidding and winning data retrieval request information, wherein the data entities and the relationship entities representing the relationship between the data entities included in each of the bidding and winning data sub-graphs are determined based on corresponding text blocks, and each of the text blocks is determined based on at least one bidding and winning project data or based on at least part of the bidding and winning project data; A retrieval data determination module is used to determine the bidding winning bid retrieval data corresponding to the bidding winning bid retrieval data retrieval request information based on the text block corresponding to each bidding winning bid data sub-graph matched by the bidding winning bid retrieval data retrieval request information, wherein the bidding winning bid retrieval data serves as the retrieval result of the bidding winning bid retrieval data retrieval request information; The step of determining, for each bidding winning data sub-graph in the bidding winning data graph, whether the bidding winning data sub-graph matches the bidding winning data search request information includes: For each bidding winning data sub-graph in the bidding winning data graph, semantic mining processing is performed on the bidding winning data sub-graph, and bidding winning data features corresponding to the bidding winning data sub-graph are output, wherein each of the bidding winning data features includes entity sub-features corresponding to each data entity in the bidding winning data sub-graph and entity sub-features corresponding to each relationship entity; Transmit and aggregate the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding and winning data features corresponding to the bidding and winning data sub-graph, and output the bidding and winning data aggregate features corresponding to the bidding and winning data sub-graph; Performing semantic mining processing on the bidding data retrieval request information, and outputting retrieval request information features corresponding to the bidding data retrieval request information; For each of the bidding winning data sub-graphs, based on the bidding winning data aggregation features corresponding to the bidding winning data sub-graphs and the retrieval request information features corresponding to the bidding winning data retrieval request information, determining whether the bidding winning data sub-graph matches the bidding winning data retrieval request information; The step of transferring and aggregating the entity sub-features corresponding to each data entity and the entity sub-features corresponding to each relationship entity included in the bidding winning data features corresponding to the bidding winning data sub-graph, and outputting the bidding winning data aggregate features corresponding to the bidding winning data sub-graph, includes: For each data entity in the bidding data sub-graph, the data entity is determined as a starting data entity, and wanders in the bidding data sub-graph to form an entity wandering link network corresponding to the data entity, wherein the length of the shortest wandering path between each other entity and the starting data entity in the entity wandering link network is used as the relationship deviation level of each other entity compared with the starting data entity; For each of the entity wandering link networks, the entity sub-features corresponding to each entity are transferred along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity, so as to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network; Aggregate the entity transfer sub-features corresponding to each data entity in the bidding data sub-graph, and output the bidding data aggregation features corresponding to the bidding data sub-graph; Among them, for each of the entity wandering link networks, the entity sub-features corresponding to each entity are transferred along the direction from other entities with the highest relationship deviation level to the corresponding starting data entity to obtain the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network, including: For each other entity in the entity wandering link network whose relationship deviates from the highest level, based on the entity sub-feature corresponding to the other entity, determining the entity transfer sub-feature corresponding to the other entity; For each other entity of each relation deviation level other than the highest relation deviation level in the entity wandering link network, cross-attention processing is performed on the entity sub-features corresponding to each other entity of the current relation deviation level based on the entity transfer sub-features corresponding to each other entity of a higher relation deviation level with a connection relationship, so as to obtain the entity transfer sub-features corresponding to each other entity of the current relation deviation level, until the entity transfer sub-features corresponding to the starting data entity corresponding to the entity wandering link network are obtained, wherein the starting data entity belongs to the other entities with the lowest relation deviation level; The step of determining, for each other entity in the entity wandering link network whose relationship deviates from the highest level, the entity transfer sub-feature corresponding to the other entity based on the entity sub-feature corresponding to the other entity, includes: For each bidding winning data sub-graph in the bidding winning data graph, in the bidding winning data graph, determine another bidding winning data sub-graph with the greatest correlation with the bidding winning data sub-graph as the corresponding associated bidding winning data sub-graph, and perform splicing and pooling processing on the entity sub-features corresponding to each entity in the associated bidding winning data sub-graph, and output the pooled splicing features corresponding to the bidding winning data sub-graph, wherein the correlation between the bidding winning data sub-graphs is determined based on the correlation between the corresponding text blocks; For each other entity whose relationship deviates from the highest level in the entity wandering link network, cross-attention processing is performed on the entity sub-features corresponding to the other entity based on the pooled splicing features corresponding to the bidding data sub-graph where the other entity is located, and a weighted superposition operation is performed on the result of the cross-attention processing and the entity sub-features corresponding to the other entity to obtain the entity transfer sub-features corresponding to the other entity.

5. An electronic device, characterized in that: include: Memory for storing computer programs; A processor connected to the memory is used to execute the computer program stored in the memory to implement the bidding winning data retrieval method described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed, executes the bidding winning data retrieval method described in any one of claims 1 to 3.

Images