Search result generation method and device
A technology for search results and search information, applied in the field of search result generation methods and devices, capable of solving problems such as unsupported provision, dedicated system does not support the provision of candidate contacts, and users cannot find solutions smoothly, etc.
Pending Publication Date: 2022-07-29
AGRICULTURAL BANK OF CHINA
0 Cites 0 Cited by
AI-Extracted Technical Summary
Problems solved by technology
[0004] However, whether it is the search function or the user help function, for some users who cannot accurately describe their own help needs, such as colloquial expressions or typos, empty results will be retur...
Method used
In summary, this embodiment integrates different dimension information such as page content, page title, page function description, historical question and answer information, and relevant function responsible persons, and utilizes each dimension information to generate a solution at the same time, which is convenient for users to understand the system functions, scenarios, and one-stop solutions to problems. In addition, this embodiment utilizes the knowledge map information to mine the global behavioral elements of the system, and matches the dimension information corresponding to high-correlation information nodes in the map as candidate answers, so that users can In the absence of valid input, valuable information can also be obtained based on implicit data such as one's own behavior. In addition, the results of knowledge map matching are interpretable, which better meets users' understanding needs for solution attribution.
[0089] In this embodiment, according to the different types of search information input by the user, different candidate data collections are screened, so that this embodiment can use more scenarios, and when the search information is the purpose type, it can be filtered out to include operation shortcut...
Abstract
The invention provides a search result generation method and device.The method comprises the steps that search information input by a target user is obtained; screening a data set related to the search information from a preset full data set as a candidate data set; matching the search information with historical related information, and taking a user corresponding to the matched historical related information as a candidate contact person; determining an associated data set corresponding to the search information according to the search information and a pre-established knowledge graph and standard question theme corpus; and determining a search result corresponding to the search information according to the candidate data set, the candidate contact person and the associated data set. According to the application, the search result determined based on the candidate data set, the candidate contact person and the associated data set can be returned only by inputting the search information once, and the search result is more comprehensive and better meets the use requirements of the user.
Application Domain
Text database queryingSpecial data processing applications +1
Technology Topic
Data miningData science +3
Image
Examples
- Experimental program(1)
Example Embodiment
[0050] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.
[0051] In view of the problems existing in the prior art, the inventor of the present application has conducted in-depth research and finally proposed a method for generating search results. Next, the method for generating search results provided by the present application will be described in detail through the following embodiments.
[0052] see figure 1 , showing a schematic flowchart of the method for generating search results provided by the embodiments of the present application, and the method for generating search results may include:
[0053] Step S101 , acquiring search information input by a target user.
[0054] Here, the search information is the information entered by the target user when he seeks to solve the difficulties (use requirements) encountered in the current use of the special system, for example, "how to use a certain function", "if a certain problem is encountered in the system, how to Solution", "How to use basic user information", "How to generate transaction comprehensive query report" and other search information.
[0055] In this step, when the target user encounters the above-mentioned usage requirements in the special system, the search information can be entered in the search bar of the special system.
[0056] Step S102 , screening a data set related to the search information from a preset full data set as a candidate data set.
[0057] In this step, a corresponding full data set is preset for each dedicated system, for example, a full page set, a stock solution data set included in the stock solution corpus, and the like. In this step, when the target user inputs the search information, a data set related to the search information can be screened from the preset full data set, so that the target user can solve the current use requirement through the screened data set.
[0058] It should be noted that the stock solution data contained in the stock solution corpus refers to the data related to the solutions currently existing in the dedicated system. Here, the solutions are used to solve the usage requirements of users.
[0059] Optionally, the stock solution data includes a solution identifier (such as a solution ID), a solution label (used to mark the standard scenario corresponding to the solution), solution author, solution scenario description, solution content, solution Data such as the title, the user who asked the question, etc.
[0060] It should be noted that, if the search information input by the target user is valid input, the candidate data set screened in this step contains several candidate data, and if the search information input by the target user is invalid input, the candidate data screened out in this step Data collection is empty.
[0061] Step S103: Match the search information with the historical related information, and use the user corresponding to the matched historical related information as a candidate contact.
[0062] It can be understood that there may be situations where the target user cannot understand the candidate data set or the candidate data set is not enough to meet the target user's usage needs. In order to further help the target user, the search information can also be matched with historical related information to get Candidate contacts who can help target users with their usage needs.
[0063] In an optional embodiment, the historical related information provided in this step includes one or more of the following information: historically published related materials and historically answered users. Here, the historical answering user refers to a user who has historically answered a related question, and the related question refers to a question related to search information.
[0064] That is, in this step, the search information can be matched with the relevant information released in the history, so as to determine the candidate contacts from the responsible users who have released the relevant materials in the history, and/or, the search information can be matched with the historical answer users, so as to Identify candidate contacts from historical answer users.
[0065] Step S104 , according to the search information and the pre-established knowledge graph and standard subject corpus, determine the associated data set corresponding to the search information.
[0066] Specifically, in this step, a knowledge graph can be generated based on the system log data and the stock solution data, and the knowledge graph associates various information in the system log data and the stock solution data. In this step, a standard subject corpus can also be generated based on the existing solution data. Optionally, a standard subject corpus can be established based on the structured data of [solution ID, solution title, solution content, solution label]. Here, the solutions in the standard question subject corpus are titled standard questions.
[0067] Optionally, the process of generating the knowledge graph based on the system log data and the stock solution data in this step may include the following S1 to S4:
[0068] S1. Acquire system log data and stock solution data, wherein the system log data is used to record the historical usage of the user when solving the problem.
[0069] S2. Several historical operation sequences are obtained based on system log data cleaning, and correlations and dependencies are extracted for several historical operation sequences to obtain the extracted correlations and dependencies.
[0070] In this step, the system log data is data in the form of [User User][execution operation Op][execution time T]. In this embodiment, the system log data can be used to cleanse data in the form of [operation A, operation B, operation C] Several historical operation sequences of , which are a set of standard solution steps generated based on system log data.
[0071] In this step, the relationship between the solutions can be extracted through the operation sequence, and the related (RELATED) relationship and the dependent (RELY) relationship can be obtained.
[0072] S3. Extract the label relationship and author relationship based on the stock solution data, and obtain the extracted label relationship and author relationship.
[0073] For example, optionally, the solution ID and solution label information in the stock solution data can be used to extract the label (LABEL) relationship, and the solution ID and solution author information in the stock solution data can be used to extract the author. (AUTHOR) relationship.
[0074] S4. Based on the relationship, dependency, label and author, the knowledge graph is established with the solution, the answerer and the topic label as the three nodes of the knowledge graph.
[0075] Specifically, when building a knowledge graph, the solution, the answerer, and the subject label are in a LABEL relationship, the solution and the solution are in a RELATED or RELY relationship, and the solution and the answerer are in a relationship. Is the author (AUTHOR) relationship.
[0076] After the knowledge map and the standard subject corpus are established, the associated data set corresponding to the search information can be determined according to the search information and the pre-established knowledge map and the standard subject corpus. Here, the related data set is regarded as "possible helpful information", which can assist the target user to solve the current usage needs.
[0077] That is, in this step, the implicit input information of the target user can be mined based on the knowledge graph, and then an association solution can be determined based on the implicit input information, so that the final search result is more comprehensive.
[0078] Step S105: Determine a search result corresponding to the search information according to the candidate data set, the candidate contact person, and the associated data set.
[0079] Specifically, in this step, the candidate data set, the candidate contact and the associated data set can be directly used as the search results corresponding to the search information, and the candidate data set, the candidate contact and the associated data set can be further processed, and the processed The candidate data set, the candidate contact and the associated data set are used as the search results corresponding to the search information.
[0080] The method for generating search results provided by the present application firstly obtains the search information input by the target user, then selects a data set related to the search information from a preset full data set as a candidate data set, and then compares the search information with historical related information. Matching, take the user corresponding to the matched historical related information as a candidate contact, and then determine the associated data set corresponding to the search information according to the search information and the pre-established knowledge graph and standard subject corpus, and finally according to the candidate data set, candidate contact A collection of people and related data to determine the search results corresponding to the search information. It can be seen that the present application can return the search results determined based on the candidate data set, the candidate contact and the associated data set only by inputting the search information once, and the search results are more comprehensive and more suitable for the user's use needs.
[0081] Considering that in some scenarios, it is sufficient to only return the corresponding answer (also referred to as a solution) for the search information input by the target user, while in other scenarios, it is necessary to return the corresponding answer for the search information input by the target user and Pages that contain action shortcut controls so that target users can quickly populate relevant information on the results page and perform actions quickly. For example, when the target user needs to perform the approval operation on the special system, the search information "I want to approve the order, the order number is xx, ..." can be entered in the search bar, and this embodiment needs to return the corresponding answer based on the search information and A page that contains an action shortcut control so that the target user can quickly go through the approval process after filling in the relevant information in the action shortcut control.
[0082] Based on this, optionally, the process of the aforementioned step "step S102, screening a data set related to the search information from the preset full data set as a candidate data set" may include:
[0083] Step S1021, judging whether the search information is the target type.
[0084] Here, the purpose type refers to the type of problem that can provide solutions and action shortcut controls (ie, pages).
[0085] Step S1022, if yes, filter the solution set related to the search information from the preset stock solution corpus, and filter the page set related to the search information from the preset full page set, and the filtered solution set and page collection as a candidate data collection.
[0086]When it is determined in the preceding steps that the search information is the purpose type, the solution set related to the search information can be screened from the stock solution corpus through this step as a candidate solution set, and the search can be filtered from the preset full-page set The information-related page set is regarded as the candidate page set, and the candidate solution set and the candidate page set are the candidate data sets in this step.
[0087] Step S1023 , if not, filter the solution set related to the search information from the existing solution corpus, and use the filtered solution set as the candidate data set.
[0088] If the search information is not the purpose type, it is sufficient to return solutions only for the search information. Based on this, this step only needs to filter the set of solutions related to the search information from the existing solution corpus.
[0089] According to the different types of search information input by the user, this embodiment filters out different candidate data sets, so that this embodiment can use more scenarios, and when the search information is the purpose type, pages containing operation shortcut controls can be filtered out , which is convenient for the target user to perform quick operation and improves the operation efficiency of the target user.
[0090] In an embodiment of the present application, the aforementioned process of "step S103, matching the search information with the historical related information, and using the user corresponding to the matched historical related information as a candidate contact" is introduced.
[0091] Optionally, the process of "step S103, matching the search information with historical related information, and using the user corresponding to the matched historical related information as a candidate contact" may include at least one of the following steps S1031 and S1032:
[0092] Step S1031 , match the search information with the historically released related materials through a data matching model, and use the user corresponding to the matched related data as a candidate contact.
[0093] If the historically related information includes historically released related materials, in this step, the search information can be matched with the historically released related materials through a pre-trained data matching model, and the users corresponding to the matched related materials are used as candidate contacts.
[0094] Here, the data matching model is obtained by training the first training search information as training data and using the user corresponding to the training related data matched with the marked first training search information as the sample label.
[0095] It is worth noting that in this step, the search information is matched with each relevant material published in the history through the material matching model, so the "matched relevant material" in this step refers to the first several relevant materials that are most matched.
[0096] Optionally, the above-mentioned data matching model may specifically be a q2a2u matching model (the first training search information query-matching criteria asks-matching contacts), and the q2a2u matching model is based on [user, sequence of operations] data to edit a specific contact. The joint probability space of the solutions is modeled.
[0097] In an optional embodiment, the User2Vec model may be used to obtain the user representation vector of the target user (that is, the user embedding, that is, the user's low-dimensional dense vector), and obtain the user representation vector of the user corresponding to each relevant data based on each relevant data released in the history, then this step “matches the search information with the relevant data released in the history through the data matching model and matches to The user corresponding to the relevant data is regarded as a candidate contact". Specifically, the user character vector of the target user and the user character vector of the user corresponding to each relevant data published in the history can be used to calculate the cosine similarity through the data matching model to obtain the similarity calculation result. The largest top K1 users, as candidate contacts.
[0098] Here, the training dataset of User2Vec is divided into user answer order S, that is, given user User i Time series records for answer or solution edits. The training goal is to maximize the log-maximum likelihood value ζ=Σlog p(Context(item)item) of the full user solution sequence. (Context refers to the context item' of a specific solution item in the solution sequence sequence S).
[0099] Step S1032: Match the search information with the historical answering user through the user matching model, and use the matched historical answering user as a candidate contact.
[0100] If the history-related information includes the historical answering user, in this step, the search information may be matched with the historical answering user through a pre-trained user matching model, and the matched historical answering user may be used as a candidate contact.
[0101] Here, the user matching model is obtained by using the second training search information as training data and the training answer user matched with the marked second training search information as the sample label.
[0102] It is worth noting that, in this step, the search information is matched with each historical answer user through the user matching model, so the "matched historical answer user" in this step refers to the first few historical answer users that are most matched.
[0103] Optionally, the above-mentioned user matching model may specifically be a q2u matching model (second training search information query-matching contacts), and the q2u matching model is based on [user, user corpus] data for a specific contact to answer a certain question. modeled in the joint probability space of .
[0104] In an optional embodiment, before the user matching model matches the search information with the historically answered users, the User2Vec model can be used to obtain the user representation vector of the target user based on the searched information, and based on the historical answers to the relevant questions answered by the users, each user can be obtained. The user representation vector of the historical answering users, then this step "match the search information with the historical answering users through the user matching model, and use the matched historical answering users as candidate contacts". Specifically, the target user's The cosine similarity calculation is performed between the user characterization vector and the user characterization vector of each historical answering user, and the top K2 users with the largest similarity calculation result are obtained as candidate contacts.
[0105] To sum up, this embodiment performs joint recall calculation based on the data matching model and/or the user matching model, so that the application can obtain more matching candidate contacts.
[0106] As introduced in the Background Art, "For some users who cannot accurately describe their own help needs, such as colloquial expressions or typos, empty results will be returned in the search bar, resulting in users being unable to find solutions." In order to enable users to obtain better search results even if they cannot accurately describe their own help needs, in step S104 in this embodiment, for the two situations in which the search information is valid input and invalid input, two methods are provided. The determination process of different associated data sets is described next. The two determination processes are introduced separately.
[0107] Based on this, optionally, the process of "step S104, according to the search information and the pre-established knowledge graph and standard subject corpus, determine the associated data set corresponding to the search information" may include the following steps S1041 or S1042:
[0108] Step S1041, if the search information is valid input, then match the search information with the standard question subject corpus to obtain the standard question corresponding to the search information, and match the standard question corresponding to the search information with the knowledge graph to obtain the association corresponding to the search information. Data collection.
[0109] Optionally, if the search information is the purpose type, in this step, the knowledge graph of the full log information (a part of the knowledge graph established in this embodiment) can be used as the target pool to be matched, and the user implicit input information (including the system usage log) can be used. etc.) to match the set of related pages related to the search information; at the same time, this step can use another part of the knowledge graph as the target pool to be matched, and use the implicit input information of the user to match the set of related solutions related to the search information.
[0110] Step S1042, if the search information is invalid input, match the historical operation sequence related to the search information with the standard question subject corpus, obtain the standard question corresponding to the historical operation sequence, and match the standard question corresponding to the historical operation sequence with the knowledge graph. , the matched set of associated data is used as the set of associated data corresponding to the search information, wherein the historical operation sequence is determined by the system log data.
[0111] Optionally, in this step, the historical operation sequence related to the search information can be vectorized to the user representation through the User2Vec model. The adjacent content in the knowledge graph of the standard corresponding to the operation sequence is used as the associated data set corresponding to the search information.
[0112] Optionally, when the related data set corresponding to the search information is displayed and output, the most similar standard questions may also be displayed and output together in the form of abbreviations.
[0113] This embodiment provides a method for determining a set of association schemes for the two situations in which the search information is valid input and invalid input, so that the final search result will not be empty regardless of whether the user can accurately describe his or her own need for help. , the user's search experience is better.
[0114] The aforementioned step S102 is a method of using a full-text search engine to search for a candidate data set, which is a coarse-grained screening scheme that does not consider the semantic similarity between the search information and the candidate data set, which results in some candidate data in the candidate data set. May not match the semantics of the search information. If the candidate data sets are directly displayed in the search results, the target user also needs to browse the candidate data sets one by one, and select helpful data from them, which is time-consuming.
[0115] In order to save the time of the target user, the coarse-grained screening result can be fine-grained screening in this embodiment, so as to screen out candidate data with inconsistent semantics. Based on this, optionally, the aforementioned process of "step S105, determining the search result corresponding to the search information according to the candidate data set, the candidate contact and the associated data set" may include:
[0116] Step S1051: Perform semantic similarity matching between the candidate data in the candidate data set and the search information respectively, and obtain several candidate data with high semantic similarity as the target data set.
[0117] That is to say, in this step, the search information can be respectively matched with each candidate data in the candidate data set for semantic similarity, and then the top K3 candidate data with the highest semantic similarity can be used as the target data set.
[0118] Optionally, this step can be implemented by using a deep learning pre-training language model, that is, using the deep learning pre-training language model to perform semantic similarity matching between the candidate data in the candidate data set and the search information, respectively, to obtain a number of high semantic similarity. candidate data.
[0119] Here, the deep learning pre-training language model is obtained by training the training search information and the corresponding training data as training data, and using the semantic similarity between the marked training search information and the training data as the sample label.
[0120]Optionally, the deep learning pre-trained language model may be composed of a pre-trained language model under a multi-head attention mechanism, and this embodiment may train different pre-trained language models for different subject domains. Here, the Multi-head Attention mechanism is the core processing mechanism in mainstream deep learning natural language processing models, such as Transformer and BERT models. The multi-head attention mechanism obtains the vector representation of the text containing contextual semantics after the matrix operation of multiple attention head devices and the linear transformation of the matrix. In this embodiment, the multi-head attention mechanism mainly exerts its significant advantages in processing the contextual semantics of long texts, improves the accuracy of tasks such as global log information mining, and supports the establishment of knowledge graphs.
[0121] Step S1052 , take the target data set, the candidate contact person and the associated data set as the search result corresponding to the search information.
[0122] This embodiment can perform fine-grained screening on the candidate data set, and filter out the candidate data with inconsistent semantics, so that the final search result is more accurate, and the time of the target user is saved.
[0123] Combining the above embodiments, optionally, if the search information is valid input and is of the purpose type, the search result in this embodiment includes three parts, namely, the target solution set and the target page set, and the candidate contact (contact for help). person), a set of associated solutions and a set of associated pages (possibly helpful information); if the search information is valid input and not of a purpose type, the search result in this embodiment includes three parts, namely the target solution set, Candidate contacts (contacts for help) and associated solution sets (possible helpful information); if the search result is invalid input, the search results in this embodiment include two parts, namely candidate contacts and associated data sets.
[0124] In this embodiment, for the specific process of obtaining the search result, reference may be made to the introduction in the foregoing embodiment, and details are not repeated here.
[0125] In an optional embodiment, considering that the prior art can only return search results containing keywords in the search information when the search information is valid input, but cannot return the reason for generating the search results, and users often need a set of Therefore, this embodiment can also generate a set of search return reasons corresponding to the search results when obtaining the search results, so as to explain the reasons for returning the search results.
[0126] To sum up, this embodiment integrates information from different dimensions such as page content, page title, page function description, historical question and answer information, and the person in charge of related functions, and uses the information of each dimension to generate a solution, which is convenient for users to understand system functions and all In addition, in this embodiment, the knowledge graph information is used to mine the global behavior elements of the system, and each dimension information corresponding to the high-relevance information nodes is matched in the graph as a candidate answer, so that the user has no valid input. In the case of a user, it can also obtain valuable information based on implicit data such as its own behavior. In addition, the results of knowledge graph matching are interpretable, which better meets users' understanding needs for solution attribution.
[0127] Provides an end-to-end solution from search information to corresponding function pages, from purpose description to transaction instructions or corresponding function pages, and at the same time based on full search content and other explicit input information, mining knowledge graphs and user continuous operation records and other implicit input information, solve user operation problems, and provide more complete solutions to improve user operation efficiency.
[0128] The embodiments of the present application further provide a search result generating apparatus. The following describes the search result generating apparatus provided by the embodiments of the present application. The search result generating apparatus described below and the search result generating method described above can be referred to each other correspondingly.
[0129] see figure 2 , showing a schematic structural diagram of the apparatus for generating search results provided by the embodiments of the present application, such as figure 2 As shown, the apparatus for generating search results may include: a search information acquisition module 201 , a candidate data set determination module 202 , a candidate contact determination module 203 , a related data set determination module 204 and a search result determination module 205 .
[0130] The search information acquisition module 201 is configured to acquire search information input by a target user.
[0131] The candidate data set determination module 202 is configured to filter a data set related to the search information from a preset full data set as a candidate data set.
[0132] A candidate contact determination module 203, configured to match the search information with historical related information, and use the user corresponding to the matched historical related information as a candidate contact, wherein the historical related information includes historically released related materials and/or historical answers Users, historical answer users refer to users who have historically answered related questions.
[0133] The associated data set determination module 204 is configured to determine the associated data set corresponding to the search information according to the search information and the pre-established knowledge graph and standard question subject corpus, wherein the knowledge graph is generated based on the system log data and the stock solution data, and the standard question The topic corpus is generated based on the stock solution data.
[0134] The search result determination module 205 is configured to determine the search result corresponding to the search information according to the candidate data set, the candidate contact and the associated data set.
[0135] The device for generating search results provided by the present application first obtains the search information input by the target user, and then selects the data sets related to the search information from the preset full data set as candidate data sets, and then performs the search information and historical related information. Matching, take the user corresponding to the matched historical related information as a candidate contact, and then determine the associated data set corresponding to the search information according to the search information and the pre-established knowledge graph and standard subject corpus, and finally according to the candidate data set, candidate contact A collection of people and related data to determine the search results corresponding to the search information. It can be seen that the present application can return the search results determined based on the candidate data set, the candidate contact and the associated data set only by inputting the search information once, and the search results are more comprehensive and more suitable for the user's use needs.
[0136] In a possible implementation manner, the above-mentioned candidate data set determination module 202 may include: a purpose type determination submodule, a first candidate data set determination submodule, and a second candidate data set determination submodule.
[0137] Among them, the purpose type judgment sub-module is used to judge whether the search information is the purpose type.
[0138] The first candidate data set determination sub-module is configured to filter the solution set related to the search information from the preset stock solution corpus if the destination type determination sub-module determines that the search information is the destination type, and select the solution set related to the search information from the preset total solution corpus. The page collection related to the search information is filtered in the page collection, and the filtered solution collection and page collection are used as the candidate data collection.
[0139] The first candidate data set determination submodule is used to filter the solution set related to the search information from the stock solution corpus if the target type judgment submodule judges that the search information is not the target type, and the filtered solution set is used as a candidate Data collection.
[0140] In a possible implementation manner, the above-mentioned candidate contact determination module 203 may include: a first candidate contact matching submodule and/or a second candidate contact matching submodule.
[0141] Wherein, the first candidate contact matching sub-module is used to match the search information with the historically released related materials through a data matching model, and use the user corresponding to the matched related data as a candidate contact, wherein the data matching model is: The first training search information is used as training data, and the user corresponding to the training related data matched with the marked first training search information is used as the sample label for training.
[0142] The second candidate contact matching sub-module is used to match the search information with the historical answer users through the user matching model, and use the matched historical answer users as candidate contacts, wherein the user matching model is based on the second training search information as The training data is obtained by training the user who matches the marked second training search information as the sample label.
[0143] In a possible implementation manner, the above-mentioned related data set determining module 204 may include: a first related data set determining submodule or a second related data set determining submodule.
[0144] Wherein, the first associated data set determination sub-module is used to match the search information with the standard question subject corpus if the search information is valid input, obtain the standard question corresponding to the search information, and combine the standard question corresponding to the search information with the knowledge The graph is matched to obtain the associated data set corresponding to the search information.
[0145] The first associated data set determination submodule is used to match the historical operation sequence related to the search information with the standard subject corpus if the search information is invalid input, obtain the standard question corresponding to the historical operation sequence, and match the historical operation sequence to the corresponding standard question. The standard question of the search is matched with the knowledge graph, and the matched set of related data is used as the set of related data corresponding to the search information, wherein the historical operation sequence is determined by the system log data.
[0146] In a possible implementation manner, the foregoing search result determination module 205 may include: a target data set determination submodule and a search result determination submodule.
[0147] The target data set determination sub-module is used to perform semantic similarity matching between the candidate data in the candidate data set and the search information respectively, and obtain several candidate data with high semantic similarity as the target data set.
[0148] The search result determination sub-module is used for taking the target data set, the candidate contact person and the associated data set as the search result corresponding to the search information.
[0149] In a possible implementation manner, the above target data set determination sub-module can be specifically used to perform semantic similarity matching between the candidate data in the candidate data set and the search information through a deep learning pre-training language model, so as to obtain a high semantic similarity. Among them, the deep learning pre-training language model is obtained by training the training search information and the corresponding training data as the training data, and using the semantic similarity between the marked training search information and the training data as the sample label.
[0150] In a possible implementation manner, the apparatus for generating search results provided by the embodiments of the present application may further include: a search return reason generating module.
[0151] The search return reason generating module is used to generate a search return reason corresponding to the search result, wherein the search return reason is used to explain the reason for returning the search result.
[0152] In a possible implementation manner, the process of establishing the knowledge graph by the related data set determining module 204 may include: a data acquisition submodule, a first relation extraction submodule, a second relation extraction submodule, and a knowledge graph establishment submodule.
[0153] The data acquisition sub-module is used to acquire system log data and stock solution data, wherein the system log data is used to record the historical usage of the user when solving the problem.
[0154] The first relationship extraction sub-module is used for obtaining several historical operation sequences based on system log data cleaning, and extracting the correlation and dependency relationship for the several historical operation sequences to obtain the extracted correlation and dependency relationship.
[0155] The second relationship extraction sub-module is used to extract the label relationship and author relationship based on the stock solution data, and obtain the extracted label relationship and author relationship.
[0156]The knowledge graph building sub-module is used to build a knowledge graph based on correlation, dependency, labeling and author relations, with solutions, answerers and topic labels as the three nodes of the knowledge graph.
[0157] The embodiments of the present application also provide a device for generating search results. optional, image 3 A block diagram of the hardware structure of the search result generating device is shown, refer to image 3 , the hardware structure of the search result generating device may include: at least one processor 301, at least one communication interface 302, at least one memory 303 and at least one communication bus 304;
[0158] In this embodiment of the present application, the number of the processor 301, the communication interface 302, the memory 303, and the communication bus 304 is at least one, and the processor 301, the communication interface 302, and the memory 303 communicate with each other through the communication bus 304;
[0159] The processor 301 may be a central processing unit (CPU), or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention, etc.;
[0160] The memory 303 may include high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), etc., such as at least one disk memory;
[0161] Wherein, the memory 303 stores a program, and the processor 301 can call the program stored in the memory 303, and the program is used for:
[0162] Obtain the search information entered by the target user;
[0163] Filter the data sets related to the search information from the preset full data sets as candidate data sets;
[0164] Match the search information with historical related information, and use the user corresponding to the matched historical related information as a candidate contact, where the historical related information includes historically published related materials and/or historical answer users, and historical answer users refer to historical answer users users who have experienced related issues;
[0165] Determine the associated data set corresponding to the search information according to the search information and the pre-established knowledge graph and standard topic corpus, wherein the knowledge graph is generated based on the system log data and the stock solution data, and the standard topic corpus is generated based on the stock solution data;
[0166] The search result corresponding to the search information is determined according to the candidate data set, the candidate contact and the associated data set.
[0167] Optionally, the refinement function and extension function of the program may refer to the above description.
[0168] Embodiments of the present application further provide a readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the above-mentioned method for generating a search result.
[0169] Optionally, the refinement function and extension function of the program may refer to the above description.
[0170] Finally, it should also be noted that in this document, relational terms such as and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply these entities or operations. There is no such actual relationship or order between them. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion such that a process, method, article or device comprising a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element.
[0171] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the various embodiments can be referred to each other.
[0172] The above description of the disclosed embodiments enables any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, this application is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
PUM


Description & Claims & Application Information
We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.