An information determination method, apparatus, device, and computer-readable storage medium

Abstract

Embodiments of the present application disclose an information determination method, which comprises the following steps: after receiving query information of a target user for a target bidding system, obtaining historical behavior information and user attribute information of the target user; based on the query information, obtaining a set of bidding information by obtaining bidding information from a plurality of bidding applications through the target bidding system; determining key bidding information from the set of bidding information; and based on the query information, the historical behavior information and the user attribute information, determining target bidding information from the key bidding information. Thus, the operation of user query screening is effectively simplified, the query time is reduced, and the query efficiency is improved. Embodiments of the present application also disclose an information determination device, equipment and computer readable storage medium.

Description

Technical Field

[0001] This application relates to information determination technology in the field of computers, and more particularly to an information determination method, apparatus, device and computer-readable storage medium. Background Technology

[0002] Government and corporate bidding information is published on bidding websites, allowing users to find the information they need. Generally, the bidding information a user requires is available on multiple bidding websites, necessitating manual searching and compilation of the results. However, this manual search method is cumbersome and time-consuming. Summary of the Invention

[0003] To address the aforementioned technical problems, embodiments of this application aim to provide an information determination method, apparatus, device, and computer-readable storage medium, which solves the problems of cumbersome and time-consuming manual query methods in the prior art, thereby improving query efficiency.

[0004] The technical solution of this application is implemented as follows:

[0005] An information determination method, the method comprising:

[0006] After receiving the query information from the target user regarding the target bidding system, the historical behavior information and user attribute information of the target user are obtained;

[0007] Based on the query information, a set of bidding information is obtained by acquiring bidding information from multiple bidding applications through the target bidding system.

[0008] Identify key bidding information from the aforementioned set of bidding information;

[0009] Based on the query information, the historical behavior information, and the user attribute information, target bidding information is determined from the key bidding information.

[0010] In the above scheme, determining key bidding information from the bidding information set includes:

[0011] Based on the key bidding information determination model, the key bidding information is determined from the set of bidding information;

[0012] Accordingly, before determining the key bidding information from the bidding information set based on the key bidding information determination model, the process further includes:

[0013] Obtain initial bidding information;

[0014] Based on multiple target keywords and the weight of each target keyword, a tag is determined for the initial bidding information; wherein, the tag indicates whether the initial bidding information is key bidding information;

[0015] Based on the initial bidding information and the label of each initial bidding information, the bidding information to be trained is obtained;

[0016] The model is trained based on the bidding information to be trained, and the key bidding information determination model is obtained.

[0017] In the above scheme, determining the tags for the initial bidding information based on multiple target keywords and the weight of each target keyword includes:

[0018] For each initial bidding information, the initial bidding information is segmented into words to obtain a word segmentation set;

[0019] From the word segmentation set, obtain multiple target word segments that match the multiple target keywords;

[0020] Based on the frequency of occurrence of each target word and the weight of the target keyword corresponding to each target word, a first value is determined for each initial bidding information; wherein, the frequency of occurrence is the proportion of the number of each target word to the number of words included in the word set;

[0021] Based on the first value and the target threshold, a label is determined for each initial bidding information.

[0022] In the above scheme, determining the target bidding information from the key bidding information based on the query information, the historical behavior information, and the user attribute information includes:

[0023] Determine the first semantic feature information of the query information; wherein, the first semantic feature information represents the semantic features of the query information;

[0024] A graph structure for the target user is constructed based on the historical behavior information; wherein, the graph structure represents the behavioral trajectory of the target user;

[0025] The graph structure is analyzed to obtain the behavioral feature information and temporal feature information of the target user; wherein, the behavioral feature information represents the occurrence relationship of the target user's historical behaviors, and the temporal feature information represents the occurrence sequence of the historical behaviors;

[0026] Based on the first semantic feature information, the behavioral feature information, the time feature information, and the user attribute information, the target bidding information is determined from the key bidding information.

[0027] In the above scheme, the step of analyzing the graph structure to obtain the behavioral and temporal characteristic information of the target user includes:

[0028] By analyzing the graph structure, the order in which the historical behaviors occurred and the relationships between them can be obtained.

[0029] The behavioral characteristic information is determined based on the occurrence relationship;

[0030] The time feature information is determined based on the second semantic feature information of the historical bidding information corresponding to the order of occurrence and the historical behavior; wherein, the second semantic feature information represents the semantic features of the historical bidding information.

[0031] In the above scheme, determining the target bidding information from the key bidding information based on the first semantic feature information, the behavioral feature information, the time feature information, and the user attribute information includes:

[0032] Based on the attribute information of the key bidding information, the third semantic feature information of the key bidding information is determined;

[0033] The first semantic feature information, the behavioral feature information, the time feature information, and the user attribute information are concatenated to obtain the target feature information;

[0034] The target bidding information is obtained by processing the target feature information and the third semantic feature information using a bidding information recommendation model.

[0035] In the above scheme, the step of using a bidding information recommendation model to process the target feature information and the third semantic feature information to obtain the target bidding information includes:

[0036] The bidding information recommendation model is used to process the target feature information and the third semantic feature information to obtain the bidding information to be recommended and the second value of the bidding information to be recommended.

[0037] Based on the attribute information of the tender information to be recommended, the topic corresponding to the tender information to be recommended, and the topic corresponding to the query information, the second value is processed to obtain the target value;

[0038] Based on the target value, the target bidding information is determined from the bidding information to be recommended.

[0039] The method in the above scheme further includes:

[0040] Based on the topic determination model, the first topic corresponding to the query information is obtained;

[0041] Based on the topic corresponding to the target bidding information clicked by the target user, a second topic corresponding to the query information is obtained;

[0042] The topic library of the target bidding system is updated based on the first topic, the second topic, and the query keywords of the query information.

[0043] Based on the usage frequency of the first hot words included in each topic in the updated topic library, the target hot words are determined from the first hot words and displayed in the target bidding system.

[0044] In the above scheme, updating the topic database of the target bidding system based on the first topic, the second topic, and the query keywords of the query information includes:

[0045] Based on the first topic and the second topic, the query topic of the query information is determined;

[0046] If the query topic does not match any topic in the topic library, add the query topic to the topic library.

[0047] If the query keyword matches the second hot words included in the query topic in the topic library, update the usage count of the matched second hot words in the topic library;

[0048] If the query keyword does not match the second hot topic, the query keyword is added to the query topic in the topic library, and the usage count of the query keyword is updated.

[0049] An information determining device, the device comprising:

[0050] The acquisition unit is used to acquire the target user's historical behavior information and user attribute information after receiving the query information of the target user for the target bidding system;

[0051] The acquisition unit is further configured to obtain a set of bidding information from multiple bidding applications through the target bidding system based on the query information.

[0052] The processing unit is used to determine key bidding information from the bidding information set;

[0053] The processing unit is further configured to determine target bidding information from the key bidding information based on the query information, the historical behavior information, and the user attribute information.

[0054] An electronic device, the device comprising: a processor, a memory, and a communication bus;

[0055] The communication bus is used to realize the communication connection between the processor and the memory;

[0056] The processor is used to execute the information determination program in the memory to implement the steps of the information determination method described above.

[0057] A computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps of the information determination method described above.

[0058] The information determination method, apparatus, device, and computer-readable storage medium provided in the embodiments of this application can receive query information from a target user regarding a target bidding system, obtain the target user's historical behavior information and user attribute information, then obtain bidding information from multiple bidding applications through the target bidding system based on the query information to obtain a bidding information set, determine key bidding information from the bidding information set, and then determine target bidding information from the key bidding information based on the query information, historical behavior information, and user attribute information. In this way, after receiving query information, bidding information is automatically obtained from multiple bidding applications, and the bidding information is filtered based on the query information, historical behavior information, and user attribute information, which effectively simplifies the user's query and filtering operation, reduces query time, solves the problem that the manual query method in the background art is cumbersome and time-consuming, and improves query efficiency. Attached Figure Description

[0059] Figure 1 A flowchart illustrating an information determination method provided in an embodiment of this application;

[0060] Figure 2 This is a schematic diagram illustrating the process of correcting query information in an information determination method provided in an embodiment of this application.

[0061] Figure 3 A schematic diagram illustrating the process of obtaining a set of bidding information in an information determination method provided in this application embodiment;

[0062] Figure 4 A flowchart illustrating another information determination method provided in an embodiment of this application;

[0063] Figure 5 A flowchart illustrating the training of a key bidding information determination model in an information determination method provided in this application embodiment;

[0064] Figure 6 This application provides a flowchart illustrating the process of determining key bidding information in an information determination method.

[0065] Figure 7This is a flowchart illustrating the optimization of historical behavior information in an information determination method provided in an embodiment of this application.

[0066] Figure 8 This is a flowchart illustrating the construction of a graph structure in an information determination method provided in an embodiment of this application.

[0067] Figure 9 A flowchart illustrating yet another information determination method provided in an embodiment of this application;

[0068] Figure 10 A schematic diagram illustrating the process of extracting semantic features from different data types in bidding information in an information determination method provided in this application embodiment;

[0069] Figure 11 This is a flowchart illustrating the optimization of semantic features in an information determination method provided in an embodiment of this application.

[0070] Figure 12 This is a flowchart illustrating the extraction of behavioral feature information in an information determination method provided in an embodiment of this application.

[0071] Figure 13 This is a flowchart illustrating the process of determining target bidding information in an information determination method provided in an embodiment of this application.

[0072] Figure 14 A schematic diagram of a topic library in an information determination method provided in an embodiment of this application;

[0073] Figure 15 A flowchart illustrating the updating of a topic library in an information determination method provided in this application embodiment;

[0074] Figure 16 A flowchart illustrating an information determination method provided in another embodiment of this application;

[0075] Figure 17 This is a schematic diagram of the structure of an information determination device provided in an embodiment of this application;

[0076] Figure 18 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0077] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0078] It should be understood that the phrases "embodiments of this application" or "foreign embodiments" throughout the specification mean that a specific feature, structure, or characteristic related to an embodiment is included in at least one embodiment of this application. Therefore, "embodiments of this application" or "in the foreign embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0079] Unless otherwise specified, any step in the embodiments of this application performed by the electronic device may be executed by the processor of the electronic device. It is also worth noting that the embodiments of this application do not limit the order in which the electronic device performs the following steps. Furthermore, the methods used to process data in different embodiments may be the same or different methods. It should also be noted that any step in the embodiments of this application can be executed independently by the electronic device; that is, when the electronic device performs any step in the following embodiments, it may not depend on the execution of other steps.

[0080] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application.

[0081] This application provides an information determination method, which can be applied to electronic devices. (Refer to...) Figure 1 As shown, the method includes the following steps:

[0082] Step 101: After receiving the query information from the target user regarding the target bidding system, obtain the target user's historical behavior information and user attribute information.

[0083] In this embodiment, the target user is the user currently querying bidding information. The target bidding system is a system for querying bidding information, which can obtain bidding information related to the query from multiple bidding websites. The query information is the information the target user currently needs to query, including keywords related to the bidding information. Historical behavior information includes the target user's browsing history, favorites, and clicks before this query. User attribute information is the user's personal information, which may include the user's age, gender, industry, years of employment, company, and position.

[0084] In one feasible approach, the target bidding system has a front-end interface through which target users can input query information. Alternatively, the front-end interface displays currently popular topics and hot keywords, allowing target users to directly click on hot keywords related to their query needs to obtain the bidding information they require.

[0085] In this embodiment, a user account can be created for each user in the target bidding system. User attribute information can be obtained when creating the user account. Behavioral information generated by each user during each query can be recorded, so that the user's historical behavioral information can be obtained based on the user identifier. Alternatively, user attribute information and historical behavioral information can also be obtained from other bidding websites based on the user identifier. This embodiment does not limit the method of obtaining historical behavioral information and user attribute information.

[0086] In other embodiments of this application, after obtaining the query information, the query information can be corrected; the method for correcting the query information can be as follows: Figure 2 As shown, the query information is segmented using tools such as Jieba to obtain the target user's query term set. Each word in the query term set is traversed sequentially to determine whether it matches a pre-built dictionary. If it does not match, the word is considered incorrect. At this point, multiple similar characters are searched using pre-built similar character databases and pre-built similar sound character databases. Multiple candidate words are determined based on these similar characters. Then, each candidate word is used to replace the incorrect word in the query information to obtain multiple different candidate statements. The semantic information of each candidate statement is then extracted and scored. The candidate statement with the highest score is determined as the corrected query information. Step 102 is executed based on the corrected query information to reduce errors and improve the accuracy of the query.

[0087] Step 102: Based on the query information, obtain the bidding information set by acquiring bidding information from multiple bidding applications through the target bidding system.

[0088] In this embodiment of the application, the bidding information set includes multiple bidding information obtained by the target bidding system from multiple bidding applications; the bidding application is an application that includes bidding information from multiple different fields, and the bidding application can be a bidding website; compared with obtaining bidding information from a single bidding application, obtaining bidding information from multiple bidding applications provides more comprehensive and richer bidding information.

[0089] In one feasible approach, the method for obtaining the tender information set can be as follows: Figure 3As shown, the target bidding system can collect multiple bidding applications in advance. After receiving the query information, it generates a query task. The task management distributes the query task to multiple bidding applications. Each bidding application sends a query request to the middleware through its search engine. Then, the downloader downloads the query results of each bidding application. Finally, the bidding information downloaded from multiple bidding applications is summarized to obtain a bidding information set and stored in the database for subsequent querying or processing.

[0090] Step 103: Identify key bidding information from the bidding information set.

[0091] In this embodiment, the key bidding information refers to bidding information that is highly important and relevant to the relevant field, selected from the bidding information set. Since the bidding information set is obtained from multiple bidding applications, it may contain some bidding information of lower importance or bidding information whose topic is inconsistent with the query information. Therefore, key bidding information can be identified from the bidding information set to facilitate recommending more important and relevant bidding information to the target user, thereby improving query accuracy.

[0092] Step 104: Based on query information, historical behavior information, and user attribute information, determine the target bidding information from the key bidding information.

[0093] In this embodiment, the target bidding information is bidding information recommended to the target user, that is, bidding information with high relevance to the query information obtained through the target bidding application. Based on the query information, historical behavior information, and user attribute information, the key bidding information is further filtered to obtain target bidding information with stronger relevance to the target user and better meet the target user's query needs. In one feasible approach, after the target bidding information is determined, it can be displayed on the front-end page of the target bidding application to facilitate the target user to view and browse the relevant information of the target bidding information.

[0094] The information determination method provided in this application embodiment receives query information from a target user regarding a target bidding system, obtains the target user's historical behavior information and user attribute information, then obtains bidding information from multiple bidding applications through the target bidding system based on the query information to obtain a bidding information set, identifies key bidding information from the bidding information set, and then determines target bidding information from the key bidding information based on the query information, historical behavior information, and user attribute information. In this way, after receiving query information, bidding information is automatically obtained from multiple bidding applications, and the bidding information is filtered based on the query information, historical behavior information, and user attribute information, which effectively simplifies the user's query and filtering operation, reduces query time, solves the problem that the manual query method in the background technology is cumbersome and time-consuming, and improves query efficiency.

[0095] Based on the foregoing embodiments, this application provides an information determination method, referring to... Figure 4 As shown, the method includes the following steps:

[0096] Step 201: After receiving the query information from the target user regarding the target bidding system, the electronic device obtains the target user's historical behavior information and user attribute information.

[0097] Step 202: Based on the query information, the electronic device obtains a set of bidding information from multiple bidding applications through the target bidding system.

[0098] In the embodiments of this application, the key bidding information determination model can be trained through steps 203 to 206. Currently, other methods can also be used to train the key bidding information determination model, which is not limited here.

[0099] Step 203: Electronic equipment obtains initial bidding information.

[0100] In this embodiment of the application, the initial bidding information is used to train the key bidding information determination model. Bidding information can be obtained from multiple bidding websites, and the obtained bidding information is determined as the initial bidding information.

[0101] Step 204: The electronic device determines the tags for the initial bidding information based on multiple target keywords and the weight of each target keyword.

[0102] Among them, the label indicates whether the initial bidding information is key bidding information.

[0103] In this embodiment, key bidding information refers to bidding information with high relevance to this field, while non-key bidding information refers to bidding information with low relevance to this field. Target keywords can be pre-set, and can be set as popular and representative words in each field. Each target keyword will be assigned a corresponding weight. In one feasible approach, in the software bidding direction, multiple target keywords can be set as software, platform, system, digitalization, intelligence, etc. The target keywords can be set according to the actual business scenario, and this embodiment does not limit this. As shown in the table below, the target keywords in the table are hot words in the software industry, and the weight can represent the importance of each target keyword. The weight of the target keyword "software" is ω0, the weight of the target keyword "platform" is ω1, the weight of the target keyword "software" is ω2, the weight of the target keyword "digitalization" is ω3, and the weight of the target keyword "intelligence" is ω4. The weight corresponding to each target keyword is different, and the corresponding weight can be set according to the usage frequency or importance of each target keyword in this field.

[0104] Target keywords Weight Software [CDATA[ω0]] Platform [CDATA[ω1]] System [CDATA[ω2]] Digitization [omega] 3 Intelligence [CDATA[ω4]] ...... ......

[0105] In this embodiment, it is possible to determine whether the initial bidding information is key bidding information based on multiple target keywords and the weight of each target keyword. If a certain initial bidding information is key bidding information, it is labeled as key bidding information; if a certain initial bidding information is not key bidding information, it is labeled as non-key bidding information. In this way, the data processing efficiency and model training speed can be improved through automatic labeling.

[0106] Step 205: The electronic device obtains the bidding information to be trained based on the initial bidding information and the label of each initial bidding information.

[0107] In this embodiment of the application, each initial bidding information is used as sample text, and the label of each initial bidding information is used as sample marker to obtain the bidding information to be trained.

[0108] Step 206: The electronic device trains the model based on the bidding information to be trained, and obtains the key bidding information determination model.

[0109] In this embodiment of the application, the key bidding information determination model is used to determine key bidding information from the set of bidding information.

[0110] In one feasible approach, model training can be performed using a Bidirectional Encoder Representations from Transformers (BERT) model. The process of training a model based on BERT is as follows: Figure 5 As shown, the bidding information to be trained is converted into query (seq) statements, and the seq statements are then embedded, that is, the seq statements are converted into vectors (N... X Then, the Transformer model is called to extract sentence features. The Transformer model consists of a multi-head attention layer and a feed forward operation. The input and output of the multi-head attention are added together and regularized before being input into the feed forward. The input and output of the feed forward are then added together and regularized to obtain the output of the Transformer layer. The output of the Transformer layer is used as the input of the Softmax layer. The Softmax layer outputs the probability value of each category. The loss value is calculated by combining the label of the bidding information to be trained. The network weights are optimized based on the loss value until the key bidding information determination model is obtained.

[0111] Step 207: The electronic equipment determines key bidding information from the set of bidding information based on the key bidding information determination model.

[0112] In this embodiment of the application, the operation of determining key bidding information from the bidding information set based on the key bidding information determination model can be as follows: Figure 6 As shown, after converting any one of the bidding information from multiple bidding websites into a seq statement, the BERT model can determine whether this bidding information is a key bidding information. By sequentially identifying all the bidding information included in the bidding information set, the key bidding information in the bidding information set can be determined. The BERT model inserts a classification (CLS) symbol before the text and uses the output vector corresponding to the symbol as the semantic representation of the seq statement for text classification.

[0113] Step 208: The electronic device determines the first semantic feature information of the query information.

[0114] Among them, the first semantic feature information represents the semantic features of the query information.

[0115] In this embodiment of the application, a text convolutional neural network (TextCNN) can be used to extract the query information to obtain the semantic features of the query information (i.e., the first semantic feature information).

[0116] Step 209: The electronic device constructs a graph structure of the target user based on historical behavior information.

[0117] The graph structure represents the behavioral trajectory of the target user.

[0118] In other embodiments of this application, after obtaining the historical behavior information of the target user, the user behavior information can be preprocessed, and a graph structure can be constructed based on the processed historical behavior information. In one feasible approach, the historical behavior information can be cleaned and merged to reduce redundant data and noise data such as erroneous operation data in the user behavior information. The method for cleaning and merging the historical behavior information can be as follows: if the interval between two bidding information being clicked is less than or equal to a first time threshold, or the interval between clicks before and after the same bidding time is less than a second time threshold, then the bidding information clicked later is removed. Figure 7 As shown in part a, the user first clicked on tender information D, then clicked on tender information C after a 1-second interval, then clicked on tender information A after a 1-minute interval, then clicked on tender information B after a 2-minute interval, and then clicked on tender information B again after a 1-minute interval. The time interval between clicking on tender information D and tender information C is less than 2 seconds, which could be a user error; therefore, clicking on tender information C is considered invalid. Furthermore, the time interval between the two clicks on tender information B is less than 5 minutes, so these two actions can be combined, meaning that tender information B is considered to have been clicked once by this user. The processed user behavior information is as follows: Figure 7 As shown in part b of the document.

[0119] In this embodiment, the user will browse multiple bidding information items sequentially. A browsing sequence can be constructed based on the bidding information browsed by the user in that order, and a graph structure can be built based on the browsing sequence. The graph structure can be constructed in the following ways: Figure 8 As shown, user a's click sequence a consists of bidding information D, bidding information A, and bidding information B. Based on click sequence a, the following is obtained: Figure 8 In the graph structure a; user b's click sequence b consists of bidding information B, bidding information E, bidding information D, bidding information E, and bidding information F. Based on the click sequence b, we obtain... Figure 8 In the graph structure b; user c's click sequence c is bidding information E, bidding information C, bidding information B, bidding information A, and bidding information C. Based on the click sequence c, we obtain... Figure 8The graph structure is shown in section c. Furthermore, if the time interval between two operations by the same user is greater than 2 hours, it is considered that a new browsing sequence has been initiated, and a graph structure is constructed based on this new browsing sequence.

[0120] Step 210: The electronic device analyzes the graph structure to obtain the target user's behavioral and temporal characteristics.

[0121] Among them, behavioral feature information represents the relationship of the target user's historical behavior, and time feature information represents the order in which historical behavior occurred.

[0122] In this embodiment of the application, the graph structure constructed based on the target user's historical behavior information implicitly contains the relationship between historical behaviors and the order in which historical behaviors occur, that is, it implicitly contains the target user's behavioral habits. Therefore, by analyzing the graph structure, behavioral feature information representing the relationship between historical behaviors and temporal feature information representing the order in which historical behaviors occur can be obtained.

[0123] Step 211: The electronic device determines the target bidding information from the key bidding information based on the first semantic feature information, behavioral feature information, time feature information and user attribute information.

[0124] In this embodiment, the bidding information is further filtered based on the first semantic feature information, behavioral feature information, time feature information, and user attribute information. The resulting target bidding information is more relevant to the target user and better meets the target user's query needs.

[0125] It should be noted that the descriptions of the same steps and contents as in other embodiments in this embodiment can be found in the descriptions in other embodiments, and will not be repeated here.

[0126] The information determination method provided in this application automatically obtains bidding information from multiple bidding applications after receiving query information, and filters the bidding information based on query information, historical behavior information and user attribute information. This effectively simplifies the user's query and filtering operation, reduces query time, solves the problem that the manual query method in the background technology is cumbersome and time-consuming, and improves query efficiency.

[0127] Based on the foregoing embodiments, this application provides an information determination method, referring to... Figure 9 As shown, the method includes the following steps:

[0128] Step 301: After receiving the query information from the target user regarding the target bidding system, the electronic device obtains the target user's historical behavior information and user attribute information.

[0129] Step 302: Based on the query information, the electronic device obtains a set of bidding information from multiple bidding applications through the target bidding system.

[0130] Step 303: Electronic equipment obtains initial bidding information.

[0131] Step 304: For each initial bidding information, the electronic device performs word segmentation to obtain a word segmentation set.

[0132] In this embodiment of the application, the word segmentation set is a set of all the words obtained after segmenting each initial bidding information; in one feasible way, word segmentation tools such as jieba can be used to segment each initial bidding information to obtain the word segmentation set.

[0133] Step 305: The electronic device retrieves multiple target words that match multiple target keywords from the word segmentation set.

[0134] In this embodiment of the application, multiple target words are multiple words in the word segmentation set that match multiple target keywords; in one feasible way, multiple different target keywords can be pre-set for different fields, and then the words in the word segmentation set are matched with the multiple target keywords, and the words in the word segmentation set that match successfully are determined as target words.

[0135] Step 306: The electronic device determines the first value of each initial bidding information based on the frequency of occurrence of each target word and the weight of the target keyword corresponding to each target word.

[0136] The frequency of occurrence is the proportion of the number of each target word to the total number of words included in the word segmentation set.

[0137] In this embodiment, the first value is determined based on the frequency of occurrence of the target words included in each initial bidding information and the weight of the target keyword corresponding to each target word, and is used to characterize the importance of the initial bidding information. In one feasible approach, for a given initial bidding information, the first value can be obtained by multiplying the probability of occurrence of each target word among the multiple target words included in the bidding information by the weight of the target keyword corresponding to each target word, and then summing the results of multiplying each target word. The formula for calculating the first value can be: Among them, Score seq Let W be the first value, Wordi be the i-th initial tender information, and W be the first value. i Let be the weight corresponding to the i-th initial bidding information.

[0138] Step 307: The electronic device determines the label for each initial tender information based on the first value and the target threshold.

[0139] Among them, the label indicates whether the initial bidding information is key bidding information.

[0140] In this embodiment, the target threshold can be preset, and the target threshold can be set according to the actual business scenario. This embodiment does not limit this setting. In one feasible approach, if the first value of a certain initial bidding information is greater than or equal to the target threshold, then this initial bidding information can be identified as key bidding information; if the first value of a certain initial bidding information is less than the target threshold, then this initial bidding information can be identified as non-key bidding information.

[0141] In other embodiments of this application, since a certain initial bidding information has a very low relevance to the current field, but a certain keyword of this initial bidding information happens to hit a preset keyword, and the preset keyword has a high weight, the first value of this bidding information will be greater than the target threshold, thereby labeling this initial bidding information as key bidding information. Therefore, after automatically labeling each initial bidding information, we can also manually check whether the label corresponding to each initial bidding information is correct to prevent misjudgment and improve the accuracy of the label.

[0142] Step 308: The electronic device obtains the bidding information to be trained based on the initial bidding information and the label of each initial bidding information.

[0143] Step 309: The electronic device trains the model based on the bidding information to be trained, and obtains the key bidding information determination model.

[0144] Step 310: The electronic equipment determines key bidding information from the set of bidding information based on the key bidding information determination model.

[0145] Step 311: The electronic device determines the first semantic feature information of the query information.

[0146] Among them, the first semantic feature information represents the semantic features of the query information.

[0147] Step 312: The electronic device constructs a graph structure of the target user based on historical behavior information.

[0148] The graph structure represents the behavioral trajectory of the target user.

[0149] Step 313: The electronic device analyzes the graph structure to obtain the order of occurrence of historical behaviors and the relationship between them.

[0150] In the embodiments of this application, for each browsing sequence, the order of occurrence of historical behaviors can be known through the graph structure, and the relationship between historical behaviors can also be known. The relationship can be the click jump situation of the target user between multiple bidding information; in one feasible way, it can be the number of jumps of the target user between multiple bidding information.

[0151] Step 314: The electronic device determines behavioral characteristic information based on the occurrence relationship.

[0152] Among them, behavioral feature information represents the historical behavior relationships of the target user.

[0153] In this embodiment, the click-through rate of a target user between multiple bidding information items can be stored in matrix form. In one feasible approach, if the number of times bidding information E jumps to bidding information C is 2, the number of times bidding information C jumps to bidding information B is 10, the number of times bidding information B jumps to bidding information A is 20, and the number of times bidding information A jumps to bidding information C is 15, then the matrix shown in the table below can be obtained:

[0154] E C B A E 0 2 0 0 C 0 0 10 0 B 0 0 0 20 A 0 15 0 0

[0155] Step 315: The electronic device determines the time feature information based on the second semantic feature information of the historical bidding information corresponding to the occurrence sequence and historical behavior.

[0156] Among them, the second semantic feature information represents the semantic features of historical bidding information, and the time feature information represents the order in which historical behaviors occurred.

[0157] In this embodiment, historical bidding information refers to bidding information corresponding to the target user's historical behavior. Time-related information can be determined based on the chronological order of the target user's click history; in one feasible approach, if the target user's browsing sequence is... Figure 8 If the browsing sequence C is given, then the target user's click behavior nodes can be stored as [E, C, B, A]. Based on the chronological order of historical behaviors, the semantic feature information of the historical bidding information corresponding to each historical behavior is connected to obtain the temporal feature information as [F]. E ,F C ,F B ,F A ];

[0158] In one feasible approach, the second semantic feature information of historical bidding information can be obtained as follows: first, the attribute information of each historical bidding information is obtained, and then the second semantic feature information of each historical bidding information is determined based on the attribute information. The attribute information of each historical bidding information may include the title, user input, number of impressions, number of clicks, duration, and current status. The user input refers to the input when querying historical bidding information; the duration is the time interval from the publication date of each historical bidding information to the current time; the number of impressions is the number of times each historical bidding information is displayed on the search interface; and the number of clicks is the number of times each historical bidding information is clicked by the user. Figure 10 As shown, for text features such as "title name\user input", TextCNN can be used to extract the text features of the entire sentence; for the three continuous numerical features of "number of impressions", "number of clicks", and "duration", one-hot representation after intervalization is used; if the number of clicks is 50 as an interval, and the current number of clicks is 200, then the one-hot representation of the number of clicks can be represented as: [0,0,0,1,0,…,0]; "current status" is discrete, and can take the range of three states: "about to tender", "tendering in progress", and "candidate announcement". These three states are represented numerically as 1, 2, and 3 respectively. When a certain state is taken, the value of each element in the vector is the corresponding numerical value. If the tendering status is "candidate announcement", the current state can be represented as [2,2,2,2,2,…,2]; then, the second semantic feature information of historical tendering information is determined based on all the obtained features; the operation of determining the second semantic feature information of historical tendering information based on all the obtained features can be as follows: Figure 11 As shown, the features corresponding to the title, user input, number of impressions, number of clicks, duration, and current state are superimposed. The superimposed features are then further processed by a BERT network to extract semantic features containing relationships. Finally, average pooling is applied to output a one-dimensional vector, yielding the second semantic feature information (F) for each historical bidding information. i ).

[0159] Step 316: Based on the attribute information of key bidding information, the electronic device determines the third semantic feature information of key bidding information.

[0160] In this embodiment of the application, the attribute information of the key bidding information may include the title of the key bidding information, user input, number of times it is displayed, number of times it is clicked, duration and current status, etc.; wherein, the method of determining the third semantic feature information based on the attribute information of the key bidding information is similar to the method of determining the second semantic feature information in step 312, and will not be described again in this embodiment of the application.

[0161] Step 317: The electronic device concatenates the first semantic feature information, behavioral feature information, temporal feature information, and user attribute information to obtain the target feature information.

[0162] In this embodiment of the application, after obtaining the behavioral feature information and time feature information corresponding to each behavioral sequence, deep features of the behavioral sequence can be extracted using a deep learning model; wherein, the method for extracting deep features of the behavioral sequence using a deep learning model can be as follows: Figure 12 As shown, behavioral and temporal features are input into a self-attention network to extract behavioral representation features. The self-attention network includes a self-attention layer and multiple fully connected layers. Then, the outputs of the fully connected layers are concatenated into vectors, and the concatenated features are input into multiple feedforward layers to further extract features. Finally, the behavioral feature information of the behavioral sequence is output.

[0163] In this embodiment, the behavioral feature information corresponding to historical behavior information, the first semantic feature information corresponding to query information, and the user attribute information can be concatenated to obtain target feature information. Based on the target feature information, bidding information that is most relevant to the user and better meets the user's needs can be recommended to the target user.

[0164] Step 318: The electronic device uses a bidding information recommendation model to process the target feature information and the third semantic feature information to obtain the target bidding information.

[0165] Step 318 can be achieved through the following steps:

[0166] Step 318a: The electronic device uses a bidding information recommendation model to process the target feature information and the third semantic feature information to obtain the bidding information to be recommended and the second value of the bidding information to be recommended.

[0167] In this embodiment, the tender information to be recommended is the N most similar tender information determined from key tender information based on the tender information recommendation model; the second value can be the score of each tender information to be recommended; in one feasible way, the tender information recommendation model can be pre-trained, so that the tender information recommendation model can be used to process the target feature information and the third semantic feature information to obtain the tender information to be recommended.

[0168] In the embodiments of this application, such as Figure 13As shown, the target feature information, which is the result of concatenating the behavioral feature information corresponding to the historical behavior information, the first semantic feature information corresponding to the query information, and the user attribute information, can be input into the bidding information recommendation model. Then, based on the third semantic feature information of the key bidding information, the bidding information to be recommended (i.e., the N most similar bidding information) and the second value of each bidding information to be recommended can be obtained.

[0169] Step 318b: The electronic device processes the second value to obtain the target value based on the attribute information of the tender information to be recommended, the topic corresponding to the tender information to be recommended, and the topic corresponding to the query information.

[0170] In this embodiment, the target value is a value optimized from the second value based on the attribute information of the tender information to be recommended and the topic to which the tender information to be recommended belongs, which is the final score of each tender information to be recommended. The second value is optimized based on the attribute information of the tender information to be recommended and the topic to which the tender information to be recommended belongs. The tender information recommended for the target user is determined based on the optimized value. In this way, the recommended tender information has higher timeliness and value, greater relevance to the target user, and better meets the target user's query needs.

[0171] In one feasible approach, the second value and the theme of the query information can be optimized based on attributes such as the duration, number of clicks, and number of impressions of each tender information to be recommended, as well as the theme corresponding to each tender information to be recommended. The duration is negatively correlated with the target value; that is, the longer the tender information to be recommended is published, the lower the target value. The number of impressions is also negatively correlated with the target value; that is, the more times the tender information to be recommended is displayed, the lower the target value. The number of clicks is positively correlated with the target value; that is, the more times the tender information to be recommended is clicked, the higher the target value. The theme is also positively correlated with the target value; if the tender information to be recommended and the query information belong to the same theme, the target value is higher, and in this case, the target value can be 1; if the tender information to be recommended and the query information do not belong to the same theme, the target value is higher, and in this case, the target value can be 0. The formula for optimizing the second value to obtain the target value can be expressed as: in, The second value, The number of clicks. To display the number of times, This indicates whether the tender information to be recommended belongs to the same topic as the query information. δ is used to prevent the denominator from being 0, and its value is usually 10. -7The topics corresponding to the tender information to be recommended and the topics corresponding to the query information can both be obtained based on a pre-trained topic determination model. Alternatively, when the target user enters the query information, they can also click on the topic corresponding to the query information at the same time. The topic corresponding to each tender information can be determined in advance. In this way, the topics corresponding to the tender information to be recommended and the topics corresponding to the query information can be obtained. This application does not limit the method of obtaining the topic corresponding to the tender information to be recommended and the method of determining the topic corresponding to the query information.

[0172] Step 318c: The electronic device determines the target bidding information from the recommended bidding information based on the target value.

[0173] In this embodiment, target bidding information that meets the target conditions can be determined from the bidding information to be recommended based on the target value. The target conditions can be preset. The target conditions can be set to determine the bidding information to be recommended with a target value greater than the preset value as the target bidding information. Alternatively, the bidding information to be recommended can be sorted based on the target value, and the top j bidding information to be recommended can be determined as the target bidding information. j can be set according to the actual business scenario. j can be set to 50, in which case the top 50 bidding information to be recommended can be determined as the target bidding information.

[0174] Based on the foregoing embodiments, in other embodiments of this application, reference is made to... Figure 9 As shown, the information determination method may also include the following steps:

[0175] Step 319: The electronic device determines the first topic corresponding to the query information based on the topic determination model.

[0176] In this embodiment of the application, the first topic is the topic corresponding to the query information; the topic determination model can be used to determine the first topic corresponding to the query information, or it can be used to determine the topic corresponding to the bidding information; the topic determination model can be a pre-trained model, so that the topic corresponding to each bidding information can be obtained based on the topic determination model, and the query information can also be classified based on the topic determination model to obtain the first topic corresponding to the query information.

[0177] Step 320: The electronic device obtains the second topic corresponding to the query information based on the topic corresponding to the target bidding information clicked by the target user.

[0178] In this embodiment of the application, the second topic is multiple topics corresponding to multiple target bidding information clicked by the target user; wherein, each target bidding information corresponds to one topic, and if multiple target bidding information clicked by the target user do not belong to the same topic, then the second topic obtained is multiple topics.

[0179] Step 321: The electronic device updates the topic database of the target bidding system based on the first topic, the second topic, and the query keywords of the query information.

[0180] In this embodiment, the topic library includes multiple topics, multiple hot keywords corresponding to each topic, and the usage frequency of each hot keyword. Topics and hot keywords have a hierarchical relationship, and both topics and the hot keywords corresponding to each topic can be preset; for example... Figure 14 As shown, the topic library includes five topics: education, agriculture and commerce, culture and tourism, healthcare, and drones. Among them, the education topic includes three hot keywords: sports, safe campus, and intelligent monitoring, while the healthcare topic includes five hot keywords: emergency rescue platform, medical care and elderly care, traditional Chinese medicine, quarantine visitation, and tongue diagnosis.

[0181] In this embodiment of the application, the query keywords are keywords extracted from the query information, and the query keywords may include one or more keywords; in one feasible way, the query keywords can be obtained by extracting the keywords from the query information through BERT + Long Short-Term Memory (LSTM) + Conditional Random Field (CRF).

[0182] In this embodiment of the application, the topics and hot words in the topic library can be expanded and updated based on the topic corresponding to the query information, the topic corresponding to the target bidding information clicked by the target user, and the query keywords, so as to keep the topic library up-to-date.

[0183] Step 321 can be achieved through the following steps:

[0184] Step 321a: The electronic device determines the query topic for the query information based on the first topic and the second topic.

[0185] In this embodiment, the query topic is the topic corresponding to the query information, which is determined based on the first topic corresponding to the query information and the second topic corresponding to the target bidding information clicked by the target user. If the first topic and the second topic are different, it means that the topic corresponding to the query information determined by the topic determination model is different from the topic corresponding to the target bidding information clicked by the target user. In this case, the first topic and the second topic can be used together as the query topic of the query information. If the first topic and the second topic are the same, it means that the topic corresponding to the query information determined by the topic determination model is the same as the topic corresponding to the target bidding information clicked by the target user. In this case, this topic can be determined as the query topic of the query information.

[0186] Step 321b: If the electronic device does not match the topic in the topic library, it adds the topic to the topic library.

[0187] In this embodiment of the application, querying a topic and comparing it with topics in the topic library indicates that the current topic library does not include the query topic. In this case, the query topic needs to be added to the topic library to expand the topics in the topic library, thereby continuously enriching the topics in the topic library and keeping the topics in the topic library up-to-date.

[0188] Step 321c: When the query keyword matches the second hot topic included in the query topic in the topic library, the electronic device updates the usage count of the matched second hot topic in the topic library.

[0189] In this embodiment, the second hot keywords are multiple hot keywords included under the query topic in the topic library. If the query keyword successfully matches the second hot keywords included in the query topic in the topic library, it indicates that the query keyword is already in the topic library. At this point, it is only necessary to update the usage count of the second hot keywords in the topic library to continuously update the usage count of each hot keyword in the topic library. This facilitates subsequent updates of the topics and hot keywords presented to the user on the front-end interface based on the usage counts of topics and hot keywords under those topics in the topic library.

[0190] Step 321d: If the query keyword does not match the second most popular keyword, the electronic device adds the query keyword to the query topic in the topic library and updates the usage count of the query keyword.

[0191] In this embodiment of the application, if the query keyword does not match the second hot word, it means that the query keyword is not included under the query topic in the topic library. In this case, the query keyword needs to be added to each query topic in the topic library to continuously enrich the hot words under each topic in the topic library. This will facilitate the subsequent updating of the topics and hot words presented to the user on the front-end interface based on the usage frequency of the topics and hot words under the topics in the topic library.

[0192] It should be noted that the topics and hot keywords in the topic library can be enriched not only based on the themes and hot keywords corresponding to the query information, but also based on the themes and hot keywords corresponding to the bidding information clicked by the user. Figure 15 As shown, the system can categorize the query information of multiple users to obtain the topics and hot keywords corresponding to each user's query information, and update the topics and hot keywords in the topic library accordingly. It can also categorize the bidding information clicked by each user to obtain the topics and hot keywords corresponding to each user's clicked bidding information, and update the topics and hot keywords in the topic library accordingly, keeping the topics and hot keywords in the topic library up-to-date. This facilitates subsequent updates to the topics and hot keywords presented to users on the front-end interface based on the usage frequency of topics and hot keywords under the topics in the topic library.

[0193] Step 322: The electronic device determines the target hot words from the first hot words based on the number of times the first hot words are used in each topic in the updated topic library, and displays the target hot words in the target bidding system.

[0194] In this embodiment, the target hot words are the hot words used most frequently in each topic in the updated topic library, that is, the hot words with the highest usage frequency. The target hot words with the highest usage frequency under each topic in the updated topic library can be displayed on the front-end interface of the target bidding system for users to quickly search and understand the current popular hot topics.

[0195] In other embodiments of this application, the most frequently used target topics can be determined from the updated topic library and displayed on the front-end interface of the target bidding system so that users can understand the most popular fields at present; the target topics and target hot words under each target topic can also be displayed to facilitate users to quickly understand the most popular neighborhoods and hot topics at present and to perform queries more quickly.

[0196] like Figure 16 As shown in the embodiments of this application, the information determination method is based on the query information of the target user for the target bidding application. It obtains a set of bidding information by acquiring bidding information from multiple bidding applications through the target bidding application. Then, based on a key information determination model, it determines key bidding information from the set of bidding information. Next, based on user-related information such as user attribute information, historical behavior information, and query information, it further filters and sorts the key bidding information to obtain target bidding information. This target bidding information is then displayed on the front-end interface of the target bidding system for the target user to view and browse. Furthermore, it can determine the topics and hot words corresponding to the query information, and the topics and hot words corresponding to the target bidding information clicked by the target user on the front-end interface. Based on the topics and hot words corresponding to the query information and the clicked target bidding information, it updates the topics and hot words in the topic library. Finally, it determines the most frequently used target hot words and target topics from the topic library and displays them on the query page for the user to choose from.

[0197] It should be noted that the descriptions of the same steps and contents as in other embodiments in this embodiment can be found in the descriptions in other embodiments, and will not be repeated here.

[0198] The information determination method provided in this application automatically obtains bidding information from multiple bidding applications after receiving query information, and filters the bidding information based on query information, historical behavior information and user attribute information. This effectively simplifies the user's query and filtering operation, reduces query time, solves the problem that the manual query method in the background technology is cumbersome and time-consuming, and improves query efficiency.

[0199] Based on the foregoing embodiments, this application provides an information determining device, which can be applied to... Figure 1 , 4 In the information determination method provided in the embodiment corresponding to 9, refer to Figure 17 As shown, the information determining device 4 may include:

[0200] The acquisition unit 41 is used to acquire the target user's historical behavior information and user attribute information after receiving the query information of the target user for the target bidding system;

[0201] The acquisition unit 41 is also used to obtain a set of bidding information by acquiring bidding information from multiple bidding applications through the target bidding system based on the query information.

[0202] Processing unit 42 is used to determine key bidding information from the bidding information set;

[0203] The processing unit 42 is also used to determine target bidding information from key bidding information based on query information, historical behavior information and user attribute information.

[0204] In other embodiments of this application, the processing unit 42 is further configured to determine key bidding information from the bidding information set based on the key bidding information determination model;

[0205] Accordingly, the processing unit 42 is also used to implement the following steps:

[0206] Obtain initial bidding information;

[0207] Based on multiple target keywords and the weight of each target keyword, the tags for the initial bidding information are determined; whereby the tags indicate whether the initial bidding information is key bidding information.

[0208] Based on the initial bidding information and the labels of each initial bidding information, the bidding information to be trained is obtained;

[0209] The model is trained based on the bidding information to be trained, and a model for determining key bidding information is obtained.

[0210] In other embodiments of this application, the processing unit 42 is configured to perform the following steps:

[0211] For each initial bidding information, word segmentation is performed to obtain a word segmentation set;

[0212] From the word segmentation set, obtain multiple target word segments that match multiple target keywords;

[0213] Based on the frequency of occurrence of each target word and the weight of the target keyword corresponding to each target word, the first value of each initial bidding information is determined; where the frequency of occurrence is the proportion of the number of each target word to the number of words included in the word set;

[0214] Based on the first value and the target threshold, a label is determined for each initial bidding information.

[0215] In other embodiments of this application, the processing unit 42 is configured to perform the following steps:

[0216] Determine the first semantic feature information of the query information; wherein, the first semantic feature information represents the semantic features of the query information;

[0217] A graph structure for the target user is constructed based on historical behavioral information; the graph structure represents the behavioral trajectory of the target user.

[0218] By analyzing the graph structure, we can obtain the behavioral and temporal characteristics of the target user. The behavioral characteristics represent the historical relationships of the target user's behavior, while the temporal characteristics represent the order in which the historical behavior occurred.

[0219] Based on the first semantic feature information, behavioral feature information, time feature information, and user attribute information, target bidding information is determined from key bidding information.

[0220] In other embodiments of this application, the processing unit 42 is configured to perform the following steps:

[0221] By analyzing the graph structure, we can obtain the order in which historical actions occur and the relationships between them.

[0222] Determine behavioral characteristic information based on the occurrence relationship;

[0223] Based on the second semantic feature information of historical bidding information corresponding to the order of occurrence and historical behavior, time feature information is determined; wherein, the second semantic feature information represents the semantic features of the historical bidding information.

[0224] In other embodiments of this application, the processing unit 42 is configured to perform the following steps:

[0225] Based on the attribute information of key bidding information, the third semantic feature information of key bidding information is determined;

[0226] The target feature information is obtained by concatenating the first semantic feature information, behavioral feature information, temporal feature information, and user attribute information.

[0227] A bidding information recommendation model is used to process target feature information and third semantic feature information to obtain target bidding information.

[0228] In other embodiments of this application, the processing unit 42 is configured to perform the following steps:

[0229] A bidding information recommendation model is used to process the target feature information and the third semantic feature information to obtain the bidding information to be recommended and the second value of the bidding information to be recommended.

[0230] Based on the attribute information of the tender information to be recommended, the topic corresponding to the tender information to be recommended, and the topic corresponding to the query information, the second value is processed to obtain the target value;

[0231] Based on the target value, target bidding information is determined from the bidding information to be recommended.

[0232] In other embodiments of this application, the processing unit 42 is configured to perform the following steps:

[0233] Based on the topic determination model, the first topic corresponding to the query information is obtained;

[0234] Based on the topic corresponding to the target bidding information clicked by the target user, the second topic corresponding to the query information is obtained;

[0235] Update the topic database of the target bidding system based on the first topic, the second topic, and the query keywords of the query information;

[0236] Based on the usage frequency of the first hot keywords in each topic in the updated topic library, target hot keywords are determined from the first hot keywords and displayed in the target bidding system.

[0237] In other embodiments of this application, the processing unit 42 is configured to perform the following steps:

[0238] Based on the first topic and the second topic, determine the query topic for the information being queried;

[0239] If the queried topic does not match a topic in the topic library, add the queried topic to the topic library;

[0240] If the query keywords match the second most popular terms included in the query topics in the topic library, update the usage count of the matching second most popular terms in the topic library.

[0241] If the query keyword does not match the second most popular keyword, add the query keyword to the query topic in the topic library and update the usage count of the query keyword.

[0242] It should be noted that a detailed explanation of the steps performed by each unit can be found in [reference needed]. Figure 1 , 4 The information determination method provided in the embodiments corresponding to 9 will not be described again here.

[0243] The information determination device provided in this application embodiment automatically obtains bidding information from multiple bidding applications after receiving query information, and filters the bidding information based on query information, historical behavior information and user attribute information. This effectively simplifies the user's query and filtering operation, reduces query time, solves the problem that the manual query method in the background technology is cumbersome and time-consuming, and improves query efficiency.

[0244] Based on the foregoing embodiments, embodiments of this application provide an electronic device that can be applied to... Figure 1 , 4 In the information determination method provided in the embodiment corresponding to 9, refer to Figure 18 As shown, the electronic device 5 may include: a processor 51, a memory 52, and a communication bus 53, wherein:

[0245] Communication bus 53 is used to realize the communication connection between processor 51 and memory 52;

[0246] The processor 51 is used to execute the information determination program in the memory 52 to perform the following steps:

[0247] After receiving the query information from the target user regarding the target bidding system, obtain the target user's historical behavior information and user attribute information;

[0248] Based on the query information, a set of bidding information is obtained by acquiring bidding information from multiple bidding applications through the target bidding system;

[0249] Identify key bidding information from the collection of bidding information;

[0250] Based on query information, historical behavior information, and user attribute information, target bidding information is determined from key bidding information.

[0251] In other embodiments of this application, processor 51 is used to execute an information determination program in memory 52 to determine key bidding information from a set of bidding information, in order to implement the following steps:

[0252] Based on the key bidding information identification model, key bidding information is identified from the set of bidding information;

[0253] Accordingly, in other embodiments of this application, the processor 51 is used to execute the information determination program in the memory 52 based on the key bidding information determination model, determine key bidding information from the bidding information set, and may also implement the following steps:

[0254] Obtain initial bidding information;

[0255] Based on multiple target keywords and the weight of each target keyword, the tags for the initial bidding information are determined; whereby the tags indicate whether the initial bidding information is key bidding information.

[0256] Based on the initial bidding information and the labels of each initial bidding information, the bidding information to be trained is obtained;

[0257] The model is trained based on the bidding information to be trained, and a model for determining key bidding information is obtained.

[0258] In other embodiments of this application, processor 51 is used to execute an information determination program in memory 52 to determine the tags of initial bidding information based on multiple target keywords and the weight of each target keyword, in order to implement the following steps:

[0259] For each initial bidding information, word segmentation is performed to obtain a word segmentation set;

[0260] From the word segmentation set, obtain multiple target word segments that match multiple target keywords;

[0261] Based on the frequency of occurrence of each target word and the weight of the target keyword corresponding to each target word, the first value of each initial bidding information is determined; where the frequency of occurrence is the proportion of the number of each target word to the number of words included in the word set;

[0262] Based on the first value and the target threshold, a label is determined for each initial bidding information.

[0263] In other embodiments of this application, processor 51 is used to execute an information determination program in memory 52 to determine target bidding information from key bidding information based on query information, historical behavior information, and user attribute information, in order to implement the following steps:

[0264] Determine the first semantic feature information of the query information; wherein, the first semantic feature information represents the semantic features of the query information;

[0265] A graph structure for the target user is constructed based on historical behavioral information; the graph structure represents the behavioral trajectory of the target user.

[0266] By analyzing the graph structure, we can obtain the behavioral and temporal characteristics of the target user. The behavioral characteristics represent the historical relationships of the target user's behavior, while the temporal characteristics represent the order in which the historical behavior occurred.

[0267] Based on the first semantic feature information, behavioral feature information, time feature information, and user attribute information, target bidding information is determined from key bidding information.

[0268] In other embodiments of this application, the processor 51 is used to execute the information determination program in the memory 52 to analyze the graph structure and obtain the behavioral feature information and time feature information of the target user, so as to implement the following steps:

[0269] By analyzing the graph structure, we can obtain the order in which historical actions occur and the relationships between them.

[0270] Determine behavioral characteristic information based on the occurrence relationship;

[0271] Based on the second semantic feature information of historical bidding information corresponding to the order of occurrence and historical behavior, time feature information is determined; wherein, the second semantic feature information represents the semantic features of historical bidding information.

[0272] In other embodiments of this application, processor 51 is used to execute an information determination program in memory 52 to determine target bidding information from key bidding information based on first semantic feature information, behavioral feature information, time feature information, and user attribute information, in order to implement the following steps:

[0273] Based on the attribute information of key bidding information, the third semantic feature information of key bidding information is determined;

[0274] The target feature information is obtained by concatenating the first semantic feature information, behavioral feature information, temporal feature information, and user attribute information.

[0275] A bidding information recommendation model is used to process target feature information and third semantic feature information to obtain target bidding information.

[0276] In other embodiments of this application, the processor 51 is used to execute the information determination program in the memory 52, employing a bidding information recommendation model to process the target feature information and the third semantic feature information to obtain the target bidding information, thereby implementing the following steps:

[0277] A bidding information recommendation model is used to process the target feature information and the third semantic feature information to obtain the bidding information to be recommended and the second value of the bidding information to be recommended.

[0278] Based on the attribute information of the tender information to be recommended, the topic corresponding to the tender information to be recommended, and the topic corresponding to the query information, the second value is processed to obtain the target value;

[0279] Based on the target value, target bidding information is determined from the bidding information to be recommended.

[0280] In other embodiments of this application, the processor 51 may execute the information determination program in the memory 52 to perform the following steps:

[0281] Based on the topic determination model, the first topic corresponding to the query information is obtained;

[0282] Based on the topic corresponding to the target bidding information clicked by the target user, the second topic corresponding to the query information is obtained;

[0283] Update the topic database of the target bidding system based on the first topic, the second topic, and the query keywords of the query information;

[0284] Based on the usage frequency of the first hot keywords in each topic in the updated topic library, target hot keywords are determined from the first hot keywords and displayed in the target bidding system.

[0285] In other embodiments of this application, processor 51 is used to execute the information determination program in memory 52 to update the topic library of the target bidding system based on the first topic, the second topic, and query information, in order to implement the following steps:

[0286] Based on the first topic and the second topic, determine the query topic for the information being queried;

[0287] If the queried topic does not match a topic in the topic library, add the queried topic to the topic library;

[0288] If the query keywords match the second most popular terms included in the query topics in the topic library, update the usage count of the matching second most popular terms in the topic library.

[0289] If the query keyword does not match the second most popular keyword, add the query keyword to the query topic in the topic library and update the usage count of the query keyword.

[0290] It should be noted that a detailed description of the steps performed by the processor can be found in [reference needed]. Figure 1 , 4 The information determination method provided in the embodiments corresponding to 9 will not be described again here.

[0291] The electronic device provided in this application automatically obtains bidding information from multiple bidding applications after receiving query information, and filters the bidding information based on query information, historical behavior information and user attribute information. This effectively simplifies the user's query and filtering operation, reduces query time, solves the problem that the manual query method in the background technology is cumbersome and time-consuming, and improves query efficiency.

[0292] Based on the foregoing embodiments, embodiments of this application provide a computer-readable storage medium storing one or more programs, which can be executed by one or more processors to implement... Figure 1 , 4The steps of the information determination method provided in the embodiment corresponding to 9.

[0293] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0294] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0295] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes ​ The function specified in one or more boxes.

[0296] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. ​ One or more processes and / or boxes ​ The steps of the function specified in one or more boxes.

[0297] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A method for determining information, characterized in that, The method includes: After receiving the query information from the target user regarding the target bidding system, the historical behavior information and user attribute information of the target user are obtained; Based on the query information, a set of bidding information is obtained by acquiring bidding information from multiple bidding applications through the target bidding system. Identify key bidding information from the aforementioned set of bidding information; Determine the first semantic feature information of the query information; A graph structure for the target user is constructed based on the preprocessed historical behavior information; wherein the graph structure represents the behavioral trajectory of the target user; the preprocessing includes cleaning and merging. By analyzing the graph structure, the order of occurrence of historical behaviors and the relationships between them can be obtained. Behavioral characteristic information is determined based on the aforementioned occurrence relationship; Based on the second semantic feature information of the historical bidding information corresponding to the order of occurrence and the historical behavior, time feature information is determined; wherein, the behavioral feature information represents the occurrence relationship of the target user's historical behavior, and the time feature information represents the order of occurrence of the historical behavior; the second semantic feature information is determined based on the attribute information of each historical bidding information; the attribute information of the historical bidding information includes duration and current status; the duration is the time interval from the publication date of each historical bidding information to the current time point; the current status includes about to be tendered, currently being tendered, or candidate announcement; Based on the attribute information of key bidding information, the third semantic feature information of the key bidding information is determined; the attribute information of the key bidding information includes the duration and the current status. The first semantic feature information, the behavioral feature information, the time feature information, and the user attribute information are concatenated to obtain the target feature information; A bidding information recommendation model is used to process the target feature information and the third semantic feature information to obtain the target bidding information.

2. The method according to claim 1, characterized in that, The step of determining key bidding information from the bidding information set includes: Based on the key bidding information determination model, the key bidding information is determined from the set of bidding information; Accordingly, before determining the key bidding information from the bidding information set based on the key bidding information determination model, the process further includes: Obtain initial bidding information; Based on multiple target keywords and the weight of each target keyword, a tag is determined for the initial bidding information; wherein, the tag indicates whether the initial bidding information is key bidding information; Based on the initial bidding information and the label of each initial bidding information, the bidding information to be trained is obtained; The model is trained based on the bidding information to be trained, and the key bidding information determination model is obtained.

3. The method according to claim 2, characterized in that, The process of determining the tags for the initial bidding information based on multiple target keywords and the weight of each target keyword includes: For each initial bidding information, the initial bidding information is segmented into words to obtain a word segmentation set; From the word segmentation set, obtain multiple target word segments that match the multiple target keywords; Based on the frequency of occurrence of each target word and the weight of the target keyword corresponding to each target word, a first value is determined for each initial bidding information; wherein, the frequency of occurrence is the proportion of the number of each target word to the number of words included in the word set; Based on the first value and the target threshold, a label is determined for each initial bidding information.

4. The method according to claim 1, characterized in that, The method employs a bidding information recommendation model to process the target feature information and the third semantic feature information to obtain the target bidding information, including: The bidding information recommendation model is used to process the target feature information and the third semantic feature information to obtain the bidding information to be recommended and the second value of the bidding information to be recommended. Based on the attribute information of the tender information to be recommended, the topic corresponding to the tender information to be recommended, and the topic corresponding to the query information, the second value is processed to obtain the target value; Based on the target value, the target bidding information is determined from the bidding information to be recommended.

5. The method according to claim 1, characterized in that, The method further includes: Based on the topic determination model, the first topic corresponding to the query information is obtained; Based on the topic corresponding to the target bidding information clicked by the target user, a second topic corresponding to the query information is obtained; The topic library of the target bidding system is updated based on the first topic, the second topic, and the query keywords of the query information. Based on the usage frequency of the first hot words included in each topic in the updated topic library, the target hot words are determined from the first hot words and displayed in the target bidding system.

6. The method according to claim 5, characterized in that, The step of updating the topic database of the target bidding system based on the first topic, the second topic, and the query keywords of the query information includes: Based on the first topic and the second topic, the query topic of the query information is determined; If the query topic does not match any topic in the topic library, add the query topic to the topic library. If the query keyword matches the second hot words included in the query topic in the topic library, update the usage count of the matched second hot words in the topic library; If the query keyword does not match the second hot topic, the query keyword is added to the query topic in the topic library, and the usage count of the query keyword is updated.

7. An information determining device, characterized in that, The device includes: The acquisition unit is used to acquire the target user's historical behavior information and user attribute information after receiving the query information of the target user for the target bidding system; The acquisition unit is further configured to obtain a set of bidding information from multiple bidding applications through the target bidding system based on the query information. The processing unit is used to determine key bidding information from the bidding information set; The processing unit is further configured to: determine the first semantic feature information of the query information; construct a graph structure of the target user based on the preprocessed historical behavior information; wherein the graph structure represents the behavioral trajectory of the target user; the preprocessing includes cleaning and merging; analyze the graph structure to obtain the order of occurrence of historical behaviors and the relationship between historical behaviors; determine behavioral feature information based on the relationship between occurrences; and determine time feature information based on the order of occurrence and the second semantic feature information of the historical bidding information corresponding to the historical behaviors; wherein the behavioral feature information represents the relationship between the historical behaviors of the target user, and the time feature information represents the order of occurrence of the historical behaviors; the second semantic feature information is based on each of the historical behaviors... The attribute information of historical bidding information is determined; the attribute information of the historical bidding information includes duration and current status; the duration is the time interval from the publication date of each historical bidding information to the current time point; the current status includes upcoming bidding, bidding in progress, or candidate announcement; based on the attribute information of key bidding information, the third semantic feature information of the key bidding information is determined; the attribute information of the key bidding information includes the duration and the current status; the first semantic feature information, the behavioral feature information, the time feature information, and the user attribute information are concatenated to obtain target feature information; a bidding information recommendation model is used to process the target feature information and the third semantic feature information to obtain target bidding information.

8. An electronic device, characterized in that, The electronic device includes: a processor, a memory, and a communication bus; The communication bus is used to realize the communication connection between the processor and the memory; The processor is used to execute the information determination program in the memory to implement the steps of the information determination method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of the information determination method as described in any one of claims 1 to 6.

Images