Knowledge graph-based search request processing method and device, equipment and medium

Abstract

The invention discloses a search request processing method and device based on a knowledge graph, equipment and a medium, and the method comprises the steps: obtaining a first search statement input by a target user, determining a plurality of first keywords according to the first search statement, and carrying out the matching and screening of the first keywords according to a preset keyword standard library, and obtaining a plurality of second keywords; performing matching query on the second keyword through a pre-constructed search engine knowledge graph to obtain a target keyword node; determining a plurality of target search paths and corresponding associated keywords and target search results in a search engine knowledge graph according to the target keyword nodes; and generating a keyword prompt list according to the associated keyword, generating a search result list according to the target search result, and further displaying the keyword prompt list and the search result list. The search efficiency of the user and the search result recommendation accuracy are improved, and the method can be applied to the technical field of artificial intelligence.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a search request processing method, device, equipment and medium based on a knowledge graph. Background Art

[0002] Search engine technology has developed along with the development of Internet technology. Its main function is to deal with the retrieval of massive information and meet people's needs for information in the context of information explosion. A core aspect of search engine technology is to use various advanced algorithms to calculate the correlation between user input keywords and questions and the information in the background massive data index. The algorithms adopted evaluate the relationship between a large number of information and keywords from various angles. The evaluation result is a list of information according to the relevant size. When users obtain this list, they need to actively filter out suitable search results. In the process of filtering the list, users gradually understand how to use the search engine, understand the skills of keyword input, and adjust the keyword input to re-express information needs. From this perspective, people participate in the search process, and the characteristics and service capabilities of information service systems such as search engines are invisible to users. Users rely on input adjustments and judgments on results to learn how to use the system, which wastes the time of users to obtain information.

[0003] The understanding of user questions by search engines is always the goal pursued by search engines. There are three core links in the entire search engine service process. The first core link is the understanding of the problem. The engine understands the core points of the user's query through various algorithms and methods, and knows what the user is looking for; the other is how to organize and index the background information resources so that the background information can be stored in a way that can express the essence of the information and adapt to information queries; the third aspect is the direct matching of search needs and search results, that is, how to have a highly adaptable way to match user needs with stored information and perform reasonable calculations, so that the calculation results not only meet the user's information needs but also reflect the essence of the information.

[0004] In real-world applications, the formation of a standardized knowledge graph and its application in specific fields often seem to be inadequate. One of the reasons is that people have different understandings and descriptions of problems. Individual differences lead to insufficient adaptability of standardized products. How to introduce knowledge graph technology into information service fields such as search engines, and how to allow knowledge graphs to evolve autonomously based on information service applications for individuals, groups, and fields, and ultimately form an understanding service engine for specific people, groups, and fields, is an urgent problem to be solved.

[0005] Terminology explanation:

[0006] Knowledge Graph: Knowledge graph is a semantic network that reveals the relationship between entities. It is mainly used to describe various entities and concepts in the real world and the relationship between them. The basic components of the knowledge graph are "entity-relationship-entity" triples and entity and its related attribute-value pairs. Entities are connected to each other through relationships to form a network of knowledge structures. Each node represents an entity that also has some attributes. Knowledge graphs are widely used in intelligent search, intelligent question and answer, personalized recommendations and other fields. They can improve search quality, enhance users' ability to query complex related information, and understand user intentions from a semantic level.

[0007] Search engine: It is a search technology that uses specific strategies to retrieve specified information from the Internet and feed it back to users based on user needs and certain algorithms. Search engines rely on a variety of technologies, such as web crawler technology, search and ranking technology, web page processing technology, big data processing technology, natural language processing technology, etc., to provide fast and highly relevant information services for information retrieval users. The core modules of search engine technology generally include crawlers, indexing, retrieval and ranking, etc., and a series of other auxiliary modules can be added to create a better network usage environment for users. Summary of the invention

[0008] The purpose of the present invention is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0009] To this end, an object of an embodiment of the present invention is to provide a search request processing method based on a knowledge graph, which improves the user's search efficiency and the accuracy of search result recommendations.

[0010] Another object of an embodiment of the present invention is to provide a search request processing device based on a knowledge graph.

[0011] In order to achieve the above technical objectives, the technical solutions adopted by the embodiments of the present invention include:

[0012] On the one hand, an embodiment of the present invention provides a method for processing a search request based on a knowledge graph, comprising the following steps:

[0013] Acquire a first search statement input by a target user, determine a plurality of first keywords according to the first search statement, and then match and screen the first keywords according to a preset keyword standard library to obtain a plurality of second keywords;

[0014] Perform a matching query on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node;

[0015] Determine multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node;

[0016] A keyword prompt list is generated according to the associated keywords, and a search result list is generated according to the target search result, and then the keyword prompt list and the search result list are displayed.

[0017] Further, in one embodiment of the present invention, the determining of a plurality of first keywords according to the first search statement, and then matching and screening the first keywords according to a preset keyword standard library to obtain a plurality of second keywords specifically includes:

[0018] Performing word segmentation processing on the first search sentence to obtain the first keyword;

[0019] Traversing the keyword standard library to obtain a plurality of standard keywords, and determining a first semantic similarity between each of the standard keywords and the first keyword;

[0020] When the first semantic similarity is greater than or equal to a preset first threshold, it is determined that the corresponding first keyword is the second keyword.

[0021] Furthermore, in one embodiment of the present invention, the search engine knowledge graph is constructed by the following steps:

[0022] Acquire multiple historical search records, and determine a first keyword change sequence and corresponding selection results according to the historical search records;

[0023] Determine a plurality of keyword nodes and connection relationships between the keyword nodes according to the first keyword change sequence, and determine a first search path according to the keyword nodes, the connection relationships and the selection result;

[0024] Constructing an initial knowledge graph according to the first search path;

[0025] The initial knowledge graph is completed to obtain the search engine knowledge graph.

[0026] Further, in one embodiment of the present invention, the completion of the initial knowledge graph to obtain the search engine knowledge graph specifically includes:

[0027] Obtaining user information, industry field information, and search scenario information corresponding to each of the historical search records;

[0028] Generate labeled samples according to the user information, the industry field information, the search scenario information, and the first keyword variation sequence, and determine the true labels of the labeled samples according to the selection results, and then train a first selection result prediction model according to the labeled samples and the true labels;

[0029] Acquire multiple preset second keyword change sequences, generate unlabeled samples according to the second keyword change sequences, and then input the unlabeled samples into the first selection result prediction model to obtain the predicted label category and corresponding confidence of each of the unlabeled samples;

[0030] Determining a global self-feedback threshold and a local self-feedback threshold of each of the predicted label categories, and adjusting the global self-feedback threshold according to the local self-feedback threshold to obtain a self-feedback confidence threshold;

[0031] The unlabeled samples are screened according to the self-feedback confidence threshold and the confidence to obtain confident samples, and the corresponding predicted label categories are used as pseudo labels of the confident samples;

[0032] Jointly training the first selection result prediction model according to the labeled samples and the confidence samples to obtain a second selection result prediction model;

[0033] Acquire a plurality of preset third keyword change sequences, predict the predicted selection results corresponding to the third keyword change sequences according to the second selection result prediction model, and then determine the second search path according to the third keyword change sequences and the predicted selection results;

[0034] The second search path is added to the initial knowledge graph to obtain the search engine knowledge graph.

[0035] Further, in one embodiment of the present invention, performing a matching query on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node specifically includes:

[0036] Determine a plurality of keyword nodes according to the search engine knowledge graph, and determine a second semantic similarity between each of the keyword nodes and the second keyword;

[0037] When the second semantic similarity is greater than or equal to a preset second threshold, the corresponding keyword node is determined to be the target keyword node.

[0038] Further, in one embodiment of the present invention, the determining of multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node specifically includes:

[0039] Taking the target keyword node as a starting point and each selection result as an end point, determining the shortest search path corresponding to each selection result in the search engine knowledge graph as an alternative search path;

[0040] Determine the corresponding path length according to the number of keyword nodes on each candidate search path, and determine the number of historical search successes for each candidate search path according to the historical search records;

[0041] Filtering the target search path from the candidate search paths according to the path length and the number of historical search successes, and determining a recommended ranking value corresponding to each of the target search paths;

[0042] The associated keywords are determined according to other keyword nodes on the target search path that are not the target keyword node, and the target search result is determined according to the selection result corresponding to the target search path.

[0043] Further, generating a keyword prompt list according to the associated keywords, and generating a search result list according to the target search results, specifically includes:

[0044] Determine the node distance between each of the associated keywords and the target keyword node according to the target search path;

[0045] Determine a first arrangement order of each of the associated keywords according to the node distance and the recommended ranking value, and sort the associated keywords according to the first arrangement order to obtain the keyword prompt list;

[0046] The target search results are sorted according to the recommended sorting values ​​to obtain the search result list.

[0047] On the other hand, an embodiment of the present invention provides a search request processing device based on a knowledge graph, comprising:

[0048] A keyword screening module, used to obtain a first search statement input by a target user, determine a number of first keywords according to the first search statement, and then match and screen the first keywords according to a preset keyword standard library to obtain a number of second keywords;

[0049] A keyword node query module, used to perform a matching query on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node;

[0050] A target search path determination module, used to determine multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node;

[0051] The search recommendation module is used to generate a keyword prompt list according to the associated keywords, and to generate a search result list according to the target search result, and then to display the keyword prompt list and the search result list.

[0052] On the other hand, an embodiment of the present invention provides an electronic device, comprising a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, wherein the program, when executed by the processor, realizes the search request processing method based on the knowledge graph as described above.

[0053] On the other hand, an embodiment of the present invention also provides a storage medium, which is a computer-readable storage medium for computer-readable storage, and the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the search request processing method based on the knowledge graph as described above.

[0054] The advantages and beneficial effects of the present invention will be partly given in the following description, partly become apparent from the following description, or be understood through the practice of the present invention:

[0055] The embodiment of the present invention obtains a first search statement input by a target user, determines a number of first keywords according to the first search statement, and then matches and screens the first keywords according to a preset keyword standard library to obtain a number of second keywords, matches and queries the second keywords through a pre-constructed search engine knowledge graph to obtain a target keyword node, determines multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node, generates a keyword prompt list according to the associated keywords, and generates a search result list according to the target search results, and then displays the keyword prompt list and the search result list. The embodiment of the present invention can provide a user with a keyword prompt list and a search result list for the user to choose when the user performs a search input, and can obtain accurate search results without the user entering a complete search statement, thereby improving the user's search efficiency and the accuracy of search result recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solution in the embodiments of the present invention, the following introduction is made to the drawings required for use in the embodiments of the present invention. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solution of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1A flowchart of a method for processing a search request based on a knowledge graph provided in an embodiment of the present invention;

[0058] Figure 2 A step flow chart of step S101 provided in an embodiment of the present invention;

[0059] Figure 3 A flowchart of the steps for constructing a search engine knowledge graph provided by an embodiment of the present invention;

[0060] Figure 4 A step flow chart of step S204 provided in an embodiment of the present invention;

[0061] Figure 5 A step flow chart of step S102 provided in an embodiment of the present invention;

[0062] Figure 6 A step flow chart of step S103 provided in an embodiment of the present invention;

[0063] Figure 7 A step flow chart of step S104 provided in an embodiment of the present invention;

[0064] Figure 8 A schematic diagram of the structure of a search request processing device based on a knowledge graph provided in an embodiment of the present invention;

[0065] Fig. 9 A schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention;

[0066] Fig.10 A schematic diagram of the structure of a storage medium provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0067] Embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limitations on the present application. It should be noted that, although the functional module division is performed in the system schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in a different order from the module division in the system schematic diagram or the flow chart. For the step numbers in the following embodiments, they are only set for the convenience of explanation, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0068] In the description of the present invention, the meaning of "a plurality" is two or more. If there is a description of the first or the second, it is only for the purpose of distinguishing the technical features, and it cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features. In addition, unless otherwise defined, all technical and scientific terms used in this document have the same meaning as those commonly understood by technicians in the technical field of this application. The terms used in this document are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0069] There are currently two types of input prompts: one is based on statistics, which counts and prompts high-frequency words similar to the current input word; the other is the automatic association of the pinyin input method; these two methods are independent of the background serviceable data, and also do not consider the user's historical input before obtaining the target result, which may be another way of expressing the target result. The present invention combines the semantic understanding of the background serviceable information and the input concept understanding of the historical information obtained by different users to form a search input request prompt method based on concept understanding.

[0070] The present invention solves the problem of understanding user search requests based on the conceptual understanding of service domain knowledge and record mining during the search engine interaction process, realizes the recognition of different expressions of the same question, applies historical records, learns the historical input expressions corresponding to the successful search service (such as the historical input of the target result obtained by adjusting the keywords), and performs conceptual understanding and feature extraction together with the background serviceable information content to form input method prompts, or possible user input keyword prompts and corresponding target result prompts.

[0071] Different users have different expressions for the same question, and service domain knowledge is difficult for users to understand. Users need to learn knowledge in unfamiliar fields and constantly adjust their input expressions to obtain the information they need in an unfamiliar field.

[0072] The present invention learns the conceptual matching relationship between a number of questions input by a user in the process of obtaining target information and the final results, and obtains a search input request prompt method based on concept understanding, which reduces the time for users to learn domain knowledge and also realizes the intrinsic concept learning and application between different expressions of the same question.

[0073] The present invention also proposes a method for forming and applying knowledge graphs that can be used for individuals, groups, and fields. The method forms a knowledge graph that can perceive changes in fields and changes in the expression methods of a certain type of problem over a period of time. Geographical location, terminal characteristics, usage scenarios, time and other factors are considered in search and completion. Based on this knowledge graph, the differences in different descriptions of the same thing by different users can be overcome, so that different users can obtain the same results with different descriptions of the same thing.

[0074] The search request processing method based on the knowledge graph provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a set-top box, etc.; the server side can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the search request processing method based on the knowledge graph, etc., but is not limited to the above forms.

[0075] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0076] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with the relevant laws, regulations, and standards of the relevant countries and regions. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0077] like Figure 1 FIG. 1 is a flowchart of a method for processing a search request based on a knowledge graph according to an embodiment of the present invention. Figure 1 The embodiment of the present invention provides a search request processing method based on a knowledge graph, which specifically includes the following steps:

[0078] S101, obtaining a first search statement input by a target user, determining a plurality of first keywords according to the first search statement, and then matching and screening the first keywords according to a preset keyword standard library to obtain a plurality of second keywords;

[0079] S102, performing a matching query on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node;

[0080] S103, determining multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node;

[0081] S104: Generate a keyword prompt list according to the associated keywords, and generate a search result list according to the target search result, and then display the keyword prompt list and the search result list.

[0082] The embodiment of the present invention can provide the user with a keyword prompt list and a search result list for the user to choose when the user performs a search input, and accurate search results can be obtained without the user entering a complete search statement, thereby improving the user's search efficiency and the accuracy of search result recommendations.

[0083] like Figure 2 FIG. 1 is a flowchart of step S101 provided in an embodiment of the present invention, referring to FIG. Figure 2 As an optional implementation, a plurality of first keywords are determined according to the first search statement, and then the first keywords are matched and screened according to a preset keyword standard library to obtain a plurality of second keywords, which specifically include:

[0084] S1011, performing word segmentation processing on the first search sentence to obtain a first keyword;

[0085] S1012, traversing the keyword standard library to obtain multiple standard keywords, and determining a first semantic similarity between each standard keyword and the first keyword;

[0086] S1013: When the first semantic similarity is greater than or equal to a preset first threshold, determine that the corresponding first keyword is the second keyword.

[0087] Specifically, in the keyword standard library accumulated by the search service, it is determined whether the keyword input by the user is a complete word or a complete expression. There are many judgment algorithms, such as using basic binary or ternary word segmentation methods to segment the search sentence input by the user, and calculating the first semantic similarity between the keyword obtained and each standard keyword in the keyword standard library. When the first semantic similarity is less than the preset first threshold, it is considered that the keyword is not a standard information expression and the keyword is discarded; when the first semantic similarity is greater than or equal to the preset first threshold, the corresponding first keyword is determined to be the second keyword.

[0088] like Figure 3 The figure is a flowchart of the steps of constructing a search engine knowledge graph according to an embodiment of the present invention, referring to Figure 3 As an optional implementation, the search engine knowledge graph is constructed by the following steps:

[0089] S201, obtaining multiple historical search records, and determining a first keyword change sequence and corresponding selection results according to the historical search records;

[0090] S202, determining a plurality of keyword nodes and connection relationships between the keyword nodes according to the first keyword change sequence, and determining a first search path according to the keyword nodes, the connection relationships, and the selection results;

[0091] S203, constructing an initial knowledge graph according to the first search path;

[0092] S204: Complete the initial knowledge graph to obtain a search engine knowledge graph.

[0093] Specifically, by adjusting the search service, the changes in the keywords entered by users during each search and the final selection results are accumulated to form a knowledge graph corresponding to the keywords and target information. To a certain extent, this knowledge graph can be considered as an information description ontology adapted to the service field, that is, how users describe their needs in this field.

[0094] The search engine knowledge graph consists of multiple search paths, each of which includes several keyword nodes and a corresponding selection result. The last keyword node of the search path corresponds to the keyword finally entered by the user, and the selection result of the search path corresponds to the final target information selected by the user. After long-term learning and optimization, when this search engine knowledge graph reaches a certain information capacity, it can be applied in real systems, that is, when the user enters keywords that may not have search results, the associated keywords of the keyword and the search results corresponding to the keyword are obtained according to the knowledge graph.

[0095] In the field of information services, the information service system can record the user's input keywords and final selection results for individuals, groups, vertical fields, or a comprehensive service field. This service field can be a traditional Internet field or an information service platform. Furthermore, the platform can record the changes in the user's keywords during the use of the information service system and obtain a keyword change sequence.

[0096] A complete information service process can be expressed as follows: the user first inputs a keyword, then observes the search result list, and if the user clicks on a result in the list, each result clicked by the user and the number of clicked results N are recorded. In a normal service process, when N is not zero, it is considered to be a complete information service process; in certain specific environments or specific fields, N can take an empirical value of the system, and a value greater than this empirical value is considered to be a complete information service process, and a value less than this empirical value is considered to be an incomplete information service process. The present invention will count the historical search records corresponding to each complete information service process.

[0097] In a complete information service process, there are many reasons why users change keywords but do not generate search results. Taking the search process of the voice information service platform seat service personnel as an example, this change of keywords may be caused by the following reasons:

[0098] 1. The difference in the information expression of forward users. The expression of forward users is based on their own understanding of the description of the required problem. However, at the beginning of the service, the user himself may not know whether this understanding is correct or whether the corresponding information is stored in the background. As a result, the first keyword search does not find the required results.

[0099] 2. Users adjust input keywords through understanding and interaction with the system. Based on the accumulation of service experience, users abstract the core point of (forward user) information needs, but this center point changes to a certain extent during the service process.

[0100] like Figure 4 FIG. 1 is a flow chart of step S204 provided in an embodiment of the present invention, referring to FIG. Figure 4 As an optional implementation, the initial knowledge graph is completed to obtain a search engine knowledge graph, which specifically includes:

[0101] S2041. Obtain user information, industry field information, and search scenario information corresponding to each historical search record;

[0102] S2042, generating labeled samples according to the user information, industry field information, search scenario information, and the first keyword variation sequence, and determining true labels of the labeled samples according to the selection results, and then training a first selection result prediction model according to the labeled samples and the true labels;

[0103] S2043, obtaining a plurality of preset second keyword change sequences, generating unlabeled samples according to the second keyword change sequences, and then inputting the unlabeled samples into the first selection result prediction model to obtain a predicted label category and a corresponding confidence level of each unlabeled sample;

[0104] S2044, determining a global self-feedback threshold and a local self-feedback threshold of each predicted label category, and adjusting the global self-feedback threshold according to the local self-feedback threshold to obtain a self-feedback confidence threshold;

[0105] S2045, screening the unlabeled samples according to the self-feedback confidence threshold and the confidence to obtain the confident samples, and using the corresponding predicted label category as the pseudo label of the confident samples;

[0106] S2046, jointly training the first selection result prediction model according to the labeled samples and the confidence samples to obtain a second selection result prediction model;

[0107] S2047, obtaining a plurality of preset third keyword change sequences, predicting the predicted selection results corresponding to the third keyword change sequences according to the second selection result prediction model, and then determining the second search path according to the third keyword change sequences and the predicted selection results;

[0108] S2048. Add the second search path to the initial knowledge graph to obtain a search engine knowledge graph.

[0109] Specifically, the platform records the keywords input by the user and the user's own knowledge background, industry field information and search scenario information. In this way, after a period of accumulation, a self-feedback semi-supervised learning model is used to combine the user's own knowledge background, industry field information and search scenario information to form a search engine knowledge graph of keywords and search results. The process of this semi-supervised learning model is mainly divided into two stages. In the first stage, the model uses a small amount of labeled data in the early stage to train the classifier for several rounds so that it can obtain preliminary classification capabilities. In the second stage, the model uses various types of unlabeled data gradually formed during the operation of the system, and screens out confidence samples through the self-feedback threshold calculation module, and generates corresponding pseudo-labels. These confidence samples with pseudo-labels are input into the classifier together with the original labeled data for joint training, thereby further improving the performance of the classifier. At the stage transition, the confidence samples need to be input into the classifier together with the original labeled samples, and the unsupervised learning loss is added to the loss function. Through this semi-supervised learning model, the correspondence between the keyword change sequence and the selection result can be continuously learned, so as to generate a second search path to complete the initial knowledge graph and obtain a search engine knowledge graph for specific people, groups, and fields.

[0110] The self-feedback semi-supervised learning mode of an embodiment of the present invention is described below.

[0111] The training data includes two types of data: labeled and unlabeled. The labeled samples are defined as Where N l Represents the number of labeled samples, and the label set of the corresponding samples is defined as Y l , where each label All are taken from [1, N c ], N c Represents the number of categories.

[0112] The characteristics of sample data should include keywords, user's own knowledge background, industry field and other information, as well as geographic location, terminal characteristics, usage scenarios, usage time and other search scenario information. Cross entropy loss is introduced to train the neural network. Through the initial training of the supervised loss function, the neural network has a certain ability to extract data features and can be used to predict the probability distribution of unlabeled samples. Based on this, the pseudo-label generation process for unlabeled samples is realized. Given an unlabeled sample, the prediction score for each possible category can be obtained. This score represents the model's confidence that the sample belongs to each category, and the category corresponding to the maximum probability of the predicted value is the predicted category.

[0113] In the process of generating pseudo labels for unlabeled data, some noisy samples with wrong labels will inevitably be generated. Therefore, it is crucial to remove these noisy pseudo-labeled samples. The method in the present invention mainly completes sample selection through two parts: global self-feedback threshold learning and local self-feedback threshold learning. In order to overcome the limitations of traditional fixed screening thresholds, a novel unlabeled sample model selection strategy is proposed. This strategy is based on adaptive feedback threshold learning and covers two mechanisms: global adaptive feedback and category-specific adaptive feedback.

[0114] Learning effectiveness can be effectively judged by evaluating the prediction confidence of a well-calibrated model. This method uses the prediction results of the model during training to automatically define and dynamically adjust the confidence threshold for each category. Initially, a global threshold is estimated as the exponential moving average (EMA) of the model confidence, and then the EMA of the probability of each category is evaluated using the model to adjust the global threshold.

[0115] First, the global self-feedback threshold module is gradually optimized according to the average confidence of unlabeled data during the model iteration process. Specifically, in each iteration, the module determines a global threshold by calculating the average confidence of all unlabeled samples to control the sample screening process as a whole. Secondly, the local self-feedback threshold module considers the diagnosis probability of unlabeled data for each category in more detail and comprehensively evaluates the thresholds of each category. The local threshold is calculated based on the classification probability of unlabeled samples in each category to ensure that samples of different categories are treated differently, so as to more comprehensively evaluate and screen confident samples. Finally, the global self-feedback threshold and the local self-feedback threshold are comprehensively calculated to obtain the final self-feedback threshold. This dynamic and adaptive threshold adjustment mechanism enables the model to significantly improve the classification performance and prediction accuracy when training with unlabeled data.

[0116] The setting of the global threshold needs to be closely related to the model's confidence in the unlabeled data in order to accurately reflect the overall learning situation. In addition, in order to effectively exclude those erroneous pseudo-labels, this global threshold should be gradually increased during the training process. In this patent, the global threshold is defined as the average confidence of the model in the unlabeled data at the tth iteration:

[0117]

[0118] Among them, λ is the parameter of EMA, Represents the prediction score of the jth sample in each category in the unlabeled data.

[0119] The purpose of the local self-feedback threshold is to adjust the global self-feedback threshold in a category-specific way to take into account intra-class diversity and potential category proximity. The specific learning state of each category is evaluated by calculating the model's predicted expected value for that category:

[0120]

[0121] From the above, we get the self-feedback threshold T for each category r (t) = [T r (t,1),…,T r (t,c),…T r (t,N c )], Represents the prediction score of the jth sample for the cth category in the unlabeled data.

[0122] Through global self-feedback and local self-feedback threshold learning, the final self-feedback threshold T can be obtained:

[0123]

[0124] Among them, the calculation form of MaxNorm is

[0125] According to the above self-feedback threshold calculation formula, a more reliable unlabeled sample set after screening can be obtained as

[0126] Finally, the threshold learning loss for unsupervised learning during the tth iteration is defined as follows:

[0127]

[0128] Among them, l represents the indicator function of the sample exceeding the fixed probability threshold, Represents the pseudo label of the jth sample in the unlabeled data.

[0129] The final loss function is as follows:

[0130] L=λ1L ce +λ2L T

[0131] Among them, λ1 and λ2 are the balance coefficients in the loss function, L ce represents the cross entropy loss.

[0132] like Figure 5 FIG. 1 is a flow chart of step S102 provided in an embodiment of the present invention, referring to FIG. Figure 5 As an optional implementation, a matching query is performed on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node, which specifically includes:

[0133] S1021, determining a plurality of keyword nodes according to the search engine knowledge graph, and determining a second semantic similarity between each keyword node and a second keyword;

[0134] S1022: When the second semantic similarity is greater than or equal to a preset second threshold, determine the corresponding keyword node as a target keyword node.

[0135] Specifically, determine whether the second keyword matches the recorded user input keyword and there is a certain matching relationship. This recorded keyword is the keyword of the search engine knowledge graph node. Two rules, full match and similarity match, can be used. The selection of specific rules needs to be determined according to the specific service field. If it is a knowledge graph serving individual users, the full match rule can be used (the second threshold is set to 1); if it is a knowledge graph serving a group or a vertical field, the similarity-based rule can be used; if it is a knowledge graph with a larger service scope, it is best to use the similarity rule.

[0136] When the second semantic similarity is greater than or equal to the preset second threshold, the corresponding keyword node is determined to be the target keyword node; when the second semantic similarity is less than the preset second threshold, it is considered that the user input is a qualified keyword that can be used by the system, then this node can be added to the corresponding position of the knowledge graph, and at the same time, it is recorded whether there are subsequent keywords to be entered, and the subsequent user selection information is recorded, so as to further complete the knowledge graph.

[0137] like Figure 6 FIG. 1 is a flow chart of step S103 provided in an embodiment of the present invention, referring to FIG. Figure 6 As an optional implementation, multiple target search paths and corresponding associated keywords and target search results are determined in the search engine knowledge graph according to the target keyword node, which specifically includes:

[0138] S1031, taking the target keyword node as the starting point and each selection result as the end point, determining the shortest search path corresponding to each selection result in the search engine knowledge graph as an alternative search path;

[0139] S1032, determining the corresponding path length according to the number of keyword nodes on each candidate search path, and determining the number of historical search successes of each candidate search path according to historical search records;

[0140] S1033, screening out target search paths from candidate search paths according to path lengths and the number of historical search successes, and determining a recommended ranking value corresponding to each target search path;

[0141] S1034: Determine associated keywords based on other keyword nodes other than the target keyword node on the target search path, and determine the target search result based on the selection result corresponding to the target search path.

[0142] Specifically, when a user enters a keyword for any node in the knowledge graph, if this keyword is in the last layer of the knowledge graph network, the selection results triggered by this keyword can be ranked according to the weight (frequency), and this relationship can be superimposed on the relevance judgment calculation formula of the search service system used by the system to improve the relevant calculation process and results of the formula. When this keyword is not at the last node of the network, the path from this node to the target result can be calculated, and several search paths with final results that reach above the threshold can be obtained, and the associated keywords on the path can be obtained. At this time, the system can prompt the user with search inputs that may be related to this keyword or the target result of this keyword.

[0143] After entering keywords into a search engine, the user obtains a list of services that may provide services to the user or meet the user's information needs. The agent selects the final result by communicating with the user (or the user selects from the list himself).

[0144] In this service process, through the accumulation of service logs, a many-to-many list of keyword-corresponding selection results can be obtained. Furthermore, since the keywords adjusted by the user in the information search process are obtained, a knowledge graph network with a limited number of layers can be formed. In the information service process, when the agent explicitly enters a keyword, it can determine the companies that have been selected and provided in the background. Therefore, every time a user searches, the search results triggered by this keyword and that have provided satisfactory services to the user can be arranged in front, which improves the efficiency of the user's information search.

[0145] The comprehensive factor of multiple alternative search paths from the target keyword node to the final user selection result is obtained, denoted as β. The comprehensive factor β is calculated by two factors: factor β1, obtained by calculating the shortest path from the target keyword node to the final user selection result. Factor β2, i.e., the weight coefficient from the target keyword node to the final user selection result, is obtained by counting the number of historical service successes from the target keyword node to the final user selection result. The more successful times, the greater the weight.

[0146] A certain optimization algorithm can be used to comprehensively calculate the comprehensive factor β of factor β1 and factor β2. For example, several expressions with the shortest path can be selected first, and then the expression with the largest path weight can be selected from the expressions with the shortest path; several expressions with the largest weight can be selected first, and then the expression with the shortest path can be selected from the expressions with the largest weight; or the composite weighting of the path and weight can be comprehensively considered, etc. The specific algorithm is not described in detail here. Two types of data are arranged according to the comprehensive factor β for users to choose. The first type is the search prompt keywords recommended for users, and the second type is the possible search results recommended for users.

[0147] like Figure 7 FIG. 1 is a flowchart of step S104 provided in an embodiment of the present invention, referring to FIG. Figure 7 As an optional implementation, a keyword prompt list is generated according to the associated keywords, and a search result list is generated according to the target search result, which specifically includes:

[0148] S1041, determining the node distance between each associated keyword and the target keyword node according to the target search path;

[0149] S1042, determining a first ranking order of each associated keyword according to the node distance and the recommended ranking value, and sorting the associated keywords according to the first ranking order to obtain a keyword prompt list;

[0150] S1043. Sort the target search results according to the recommended sorting values ​​to obtain a search result list.

[0151] Specifically, for the search result list, it can be sorted directly according to the recommended ranking value obtained in the previous steps; and for the associated keywords, since there may be multiple associated keywords corresponding to the same target search path, it is necessary to group the associated keywords corresponding to the same target search path, first sort the associated keyword group according to the recommended ranking value obtained in the previous steps, and then sort the node distance between each associated keyword and the target keyword node within the group. The closer the node distance, the closer it is to the front, and then you can get a keyword prompt list.

[0152] It should be noted that the recommended results of the knowledge graph give users certain search prompts or display search results, after which the user may take the following steps: select recommended keywords or select recommended target results.

[0153] When the user only selects the recommended keywords and does not select the recommended search results, the system needs to search again in the database according to the user's search click instructions. At this time, the search results are arranged according to the relevance calculation algorithm originally set by the system. The relevance accumulated by the knowledge graph can be used as an input parameter of the system relevance algorithm, but it does not play a decisive role.

[0154] In addition, the role of the present invention in the user interface is not to completely replace the search results. It has two main roles: one is to recommend information related to the input keywords to the user based on the user's input and the formation of the knowledge graph. This related information is comprehensively calculated through the successful search result records of historical services; the other role is to recommend input keywords related to the user's current input, correct user errors or input without search results, or provide users with possible correct inputs. When the user does not select these recommended keywords and search results, the system can still input according to the user's input. For the sorting of search results according to user input, you can choose to consider the accumulation of the knowledge graph to adjust the sorting results, or you can choose not to consider the accumulation of the knowledge graph and sort according to the system's own relevance calculation method.

[0155] The method steps of the embodiment of the present invention are described above. It can be appreciated that the embodiment of the present invention can provide a keyword prompt list and a search result list for the user to select when the user performs a search input, and accurate search results can be obtained without the user having to enter a complete search statement, thereby improving the user's search efficiency and the accuracy of the search result recommendation.

[0156] like Figure 8 FIG. 1 is a schematic diagram of a structure of a search request processing device based on a knowledge graph according to an embodiment of the present invention. Figure 8 , an embodiment of the present invention provides a search request processing device based on a knowledge graph, comprising:

[0157] A keyword screening module is used to obtain a first search statement input by a target user, determine a number of first keywords according to the first search statement, and then match and screen the first keywords according to a preset keyword standard library to obtain a number of second keywords;

[0158] A keyword node query module is used to perform a matching query on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node;

[0159] A target search path determination module is used to determine multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node;

[0160] The search recommendation module is used to generate a keyword suggestion list based on the associated keywords, and generate a search result list based on the target search results, and then display the keyword suggestion list and the search result list.

[0161] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0162] The embodiment of the present invention further provides an electronic device, the electronic device comprising: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, wherein when the program is executed by the processor, the above-mentioned search request processing method based on the knowledge graph is realized. The electronic device may be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0163] like Fig. 9 FIG. 1 is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention, referring to FIG. Fig. 9 , an embodiment of the present invention provides an electronic device, including:

[0164] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention;

[0165] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the search request processing method based on the knowledge graph in the embodiment of the present invention;

[0166] Input / output interface 903, used to implement information input and output;

[0167] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0168] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0169] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0170] like Fig.10 FIG. 1 is a schematic diagram of a storage medium according to an embodiment of the present invention. Fig.10 An embodiment of the present invention further provides a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs 1001, and the one or more programs 1001 can be executed by one or more processors to implement the above-mentioned search request processing method based on the knowledge graph.

[0171] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0172] The embodiment of the present invention also discloses a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 1 The method shown.

[0173] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the above-mentioned boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0174] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified to the contrary, one or more of the above-mentioned functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0175] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the above methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0176] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0177] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the above-mentioned program is printed, since the above-mentioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or processing in other suitable ways as necessary, and then stored in a computer memory.

[0178] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0179] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0180] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0181] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A search request processing method based on knowledge graph, characterized in that: The following steps are involved: Acquire a first search statement input by a target user, determine a plurality of first keywords according to the first search statement, and then match and screen the first keywords according to a preset keyword standard library to obtain a plurality of second keywords; Perform a matching query on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node; Determine multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node; A keyword prompt list is generated according to the associated keywords, and a search result list is generated according to the target search result, and then the keyword prompt list and the search result list are displayed.

2. According to claim 1, a search request processing method based on knowledge graph is characterized in that: The determining of a plurality of first keywords according to the first search statement, and then matching and screening the first keywords according to a preset keyword standard library to obtain a plurality of second keywords specifically includes: Performing word segmentation processing on the first search sentence to obtain the first keyword; Traversing the keyword standard library to obtain a plurality of standard keywords, and determining a first semantic similarity between each of the standard keywords and the first keyword; When the first semantic similarity is greater than or equal to a preset first threshold, it is determined that the corresponding first keyword is the second keyword.

3. According to claim 1, a search request processing method based on knowledge graph is characterized in that: The search engine knowledge graph is constructed by the following steps: Acquire multiple historical search records, and determine a first keyword change sequence and corresponding selection results according to the historical search records; Determine a plurality of keyword nodes and connection relationships between the keyword nodes according to the first keyword change sequence, and determine a first search path according to the keyword nodes, the connection relationships and the selection result; Constructing an initial knowledge graph according to the first search path; The initial knowledge graph is completed to obtain the search engine knowledge graph.

4. A method for processing search requests based on knowledge graph according to claim 3, characterized in that: The step of completing the initial knowledge graph to obtain the search engine knowledge graph specifically includes: Obtaining user information, industry field information, and search scenario information corresponding to each of the historical search records; Generate labeled samples according to the user information, the industry field information, the search scenario information, and the first keyword variation sequence, and determine the true labels of the labeled samples according to the selection results, and then train a first selection result prediction model according to the labeled samples and the true labels; Acquire multiple preset second keyword change sequences, generate unlabeled samples according to the second keyword change sequences, and then input the unlabeled samples into the first selection result prediction model to obtain the predicted label category and corresponding confidence of each of the unlabeled samples; Determining a global self-feedback threshold and a local self-feedback threshold of each of the predicted label categories, and adjusting the global self-feedback threshold according to the local self-feedback threshold to obtain a self-feedback confidence threshold; The unlabeled samples are screened according to the self-feedback confidence threshold and the confidence to obtain confident samples, and the corresponding predicted label categories are used as pseudo labels of the confident samples; Jointly training the first selection result prediction model according to the labeled samples and the confidence samples to obtain a second selection result prediction model; Acquire a plurality of preset third keyword change sequences, predict the predicted selection results corresponding to the third keyword change sequences according to the second selection result prediction model, and then determine the second search path according to the third keyword change sequences and the predicted selection results; The second search path is added to the initial knowledge graph to obtain the search engine knowledge graph.

5. According to claim 3, a search request processing method based on knowledge graph is characterized in that: The matching query of the second keyword by using the pre-built search engine knowledge graph to obtain the target keyword node specifically includes: Determine a plurality of keyword nodes according to the search engine knowledge graph, and determine a second semantic similarity between each of the keyword nodes and the second keyword; When the second semantic similarity is greater than or equal to a preset second threshold, the corresponding keyword node is determined to be the target keyword node.

6. A method for processing search requests based on knowledge graph according to claim 5, characterized in that: The step of determining a plurality of target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node specifically includes: Taking the target keyword node as a starting point and each selection result as an end point, determining the shortest search path corresponding to each selection result in the search engine knowledge graph as an alternative search path; Determine the corresponding path length according to the number of keyword nodes on each candidate search path, and determine the number of historical search successes for each candidate search path according to the historical search records; Filtering the target search path from the candidate search paths according to the path length and the number of historical search successes, and determining a recommended ranking value corresponding to each of the target search paths; The associated keywords are determined according to other keyword nodes on the target search path that are not the target keyword node, and the target search result is determined according to the selection result corresponding to the target search path.

7. A method for processing search requests based on knowledge graph according to claim 6, characterized in that: The step of generating a keyword prompt list according to the associated keywords and generating a search result list according to the target search results specifically includes: Determine the node distance between each of the associated keywords and the target keyword node according to the target search path; Determine a first arrangement order of each of the associated keywords according to the node distance and the recommended ranking value, and sort the associated keywords according to the first arrangement order to obtain the keyword prompt list; The target search results are sorted according to the recommended sorting values ​​to obtain the search result list.

8. A search request processing device based on knowledge graph, characterized in that: include: A keyword screening module, used to obtain a first search statement input by a target user, determine a number of first keywords according to the first search statement, and then match and screen the first keywords according to a preset keyword standard library to obtain a number of second keywords; A keyword node query module, used to perform a matching query on the second keyword through a pre-built search engine knowledge graph to obtain a target keyword node; A target search path determination module, used to determine multiple target search paths and corresponding associated keywords and target search results in the search engine knowledge graph according to the target keyword node; The search recommendation module is used to generate a keyword prompt list according to the associated keywords, and to generate a search result list according to the target search result, and then to display the keyword prompt list and the search result list.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the search request processing method based on the knowledge graph are realized as described in any one of claims 1 to 7.

10. A storage medium, the storage medium being a computer-readable storage medium, used for computer-readable storage, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the search request processing method based on the knowledge graph as described in any one of claims 1 to 7.

Images