Intelligent screening method and system based on target search data

Abstract

The invention relates to the technical field of intelligent data screening, in particular to an intelligent screening method and system based on target search data. The method comprises the following steps: acquiring an original search data set; performing data preprocessing on the original search data set to generate a standard search data set; performing database construction based on the standard search data set to obtain a search database; acquiring user input data; performing keyword analysis on the user input data to generate user input keyword data; performing search association on the search database according to keyword data input by a user to generate a first search association data set; according to the method, through data preprocessing, keyword analysis, attention weight distribution and intelligent recommendation, the accuracy and intelligence of search results are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent data screening, and in particular to an intelligent screening method and system based on target search data. Background Art

[0002] With the explosive growth of Internet content, researchers have begun to explore smarter screening methods. The development of recommendation systems has become an important direction, which uses users' historical behaviors and interests to recommend relevant content. This method provides personalized recommendations by analyzing users' preferences and behavior patterns, significantly improving the efficiency and accuracy of information retrieval. However, traditional recommendation systems are still subject to some limitations, such as data sparsity and cold start problems. To address these problems, intelligent screening methods based on target search data have begun to emerge. This method uses users' clear goals or needs, combined with big data and machine learning technology, to achieve more accurate data filtering and recommendation. With the development of deep learning and natural language processing technology, intelligent screening methods based on target search data have made great progress. The model can understand the intention of the user from the target description provided by the user, and screen the data according to the context and context. This method not only improves the accuracy of the search, but also can handle more complex queries and needs, providing users with more personalized and accurate services. However, current traditional search methods cannot effectively deal with the problem of information overload because they mainly rely on keyword matching without considering the user's actual needs and intentions. At the same time, they are limited by data sparsity and cold start problems and cannot provide users with accurate personalized recommendations, resulting in low screening accuracy and intelligence. Summary of the invention

[0003] Based on this, it is necessary to provide an intelligent screening method and system based on target search data to solve at least one of the above technical problems.

[0004] To achieve the above object, an intelligent screening method based on target search data is provided, the method comprising the following steps:

[0005] Step S1: obtaining an original search data set; performing data preprocessing on the original search data set to generate a standard search data set; constructing a database based on the standard search data set to obtain a search database;

[0006] Step S2: obtaining user input data; performing keyword analysis on the user input data to generate user input keyword data; performing search association on the search database according to the user input keyword data to generate a first search association data set; comparing the first search association data weight with a preset search association weight threshold, and when the first search association data weight is greater than or equal to the preset search association weight threshold, sorting and displaying the first search association data set corresponding to the first search association data weight to generate first search association sorting data;

[0007] Step S3: allocating attention weights to the first search-related ranking data to generate first search-related data attention weights; comparing the first search-related data attention weights with a preset standard data attention weight threshold to obtain high search-level screening data; performing similarity calculations on the high search-level screening data to obtain high search-level similarities; performing secondary screening on the high search-level screening data according to the high search-level similarities to obtain second search-related data;

[0008] Step S4: Utilize blockchain to perform user search behavior preference analysis on the second search-related data to generate user search behavior preference data; construct an intelligent user behavior profile for the user search behavior preference data to generate a user search behavior profile; utilize a deep neural network algorithm to perform intelligent search recommendations on the user search behavior profile to generate a recommendation result set to perform intelligent screening of target search data.

[0009] The present invention can improve the quality and consistency of data by preprocessing and standardizing the original search data, thereby constructing a search database more effectively. This helps to improve the performance and accuracy of the search engine. Through keyword analysis and search association, relevant search results can be provided according to the user's input. Sorting display enables users to find content of interest more quickly, thereby improving the search experience. Through attention weight allocation and similarity calculation, search results can be further screened and refined, so that users can get more relevant and accurate search recommendations. This can improve user satisfaction and reduce the redundancy of search results. By performing preference analysis and behavior portrait construction on user search behavior, the search engine can better understand the user's interests and needs. Combining deep neural network algorithms for intelligent search recommendations can provide personalized search results, increase user satisfaction and search efficiency. At the same time, blockchain technology is used to ensure the security and credibility of data, further improving the reliability of the search engine. Therefore, the present invention improves the accuracy and intelligence of search results through data preprocessing, keyword analysis, attention weight allocation and intelligent recommendation.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: obtaining an original search data set;

[0012] Step S12: performing data cleaning on the original search data set to generate an original search cleaning data set; performing data filtering on the original search cleaning data set to generate an original search filtering data set; performing data format recognition on the original search filtering data set, and performing data format standardization on the recognized original search filtering data set to generate a search format data set;

[0013] Step S13: classifying the search format data set for data duplication to obtain search non-duplicate data and search duplicate data and remove the search duplicate data;

[0014] Step S14: performing data missing value detection on the search non-duplicate data to generate a search non-duplicate data missing detection result; performing data repair on the search non-duplicate data using the search non-duplicate data missing detection result to generate a standard search data set;

[0015] Step S15: construct a database based on the standard search data set to obtain a search database.

[0016] The present invention can remove invalid or redundant data and improve the quality and accuracy of data by cleaning and filtering the original search data set. Data format identification and standardization of the original search filter data set can ensure the consistency and uniformity of the data, and facilitate subsequent data processing and analysis. The search format data set is repeated and classified, and repeated data is identified and eliminated, which avoids the influence of redundant information and provides a cleaner and more efficient data set. Missing value detection is performed on the search non-duplicate data to find the missing value problem in the data, and then data repair is performed to fill the missing values, making the data more complete and reliable. Constructing a search database based on the standard search data set will provide a structured and well-organized database that can support fast data query and retrieval, and facilitate users to perform search and analysis.

[0017] Preferably, step S14 comprises the following steps:

[0018] Step S141: using the conditional random field to perform data missing value detection on the search non-repeated data to obtain the search non-repeated data missing detection result;

[0019] Step S142: filling missing values ​​in the preliminary detected abnormal data set according to the search non-repetitive data missing detection result to generate a search filled data set; performing data consistency detection on the search filled data to obtain a data consistency detection result; performing data anomaly identification on the search filled data set according to the data consistency detection result to generate a data anomaly type identification result;

[0020] Step S143: When it is confirmed that the data anomaly type identification result is a semantic anomaly, a knowledge graph method is used to perform semantic repair on the search filling data set to generate a semantic repair data set;

[0021] Step S144: when it is confirmed that the data anomaly type identification result is a statistical anomaly, Bayesian inference is used to statistically repair the search filling data set to generate a statistical repair data set;

[0022] Step S145: Perform repair verification on the semantic repair data set and the statistical repair data set to generate a standard search data set.

[0023] The present invention can identify the missing value problem in the data by performing missing value detection on the search non-repetitive data, which is helpful to find the incompleteness of the data set. According to the missing detection result of the search non-repetitive data, the missing value filling is performed on the abnormal data set initially detected to generate a search filling data set. In this way, the missing data can be supplemented, making the data set more complete and improving the accuracy of subsequent analysis. The search filling data is subjected to data consistency detection to obtain a data consistency detection result. This helps to find inconsistency problems in the data, such as differences in data format, data type, etc., and further improve the quality and consistency of the data. According to the data consistency detection result, data anomaly identification is performed in step S142, and corresponding repair is performed according to the identification result. If the anomaly type is confirmed to be a semantic anomaly (step S143), the knowledge graph method can be used to perform semantic repair to improve the accuracy of the data semantics. If the anomaly type is confirmed to be a statistical anomaly (step S144), the Bayesian inference method can be used to perform statistical repair to improve the statistical properties of the data. The repaired semantic repair data set and the statistical repair data set are repaired and checked to generate a standard search data set, which can verify the effectiveness and correctness of the repair and ensure that the generated data set meets the expected standards and requirements.

[0024] Preferably, step S2 comprises the following steps:

[0025] Step S21: Obtain user input data;

[0026] Step S22: performing input type identification on the user input data to obtain user input type data; performing keyword analysis on the user input data based on the user input type data to generate user input keyword data;

[0027] Step S23: performing search association on the search database according to the keyword data input by the user to generate a first search association data set; performing association weight calculation on the first search association data set to obtain a first search association data weight;

[0028] Step S24: Compare the first search association data weight with the preset search association weight threshold. When the first search association data weight is greater than or equal to the preset search association weight threshold, sort and display the first search association data set corresponding to the first search association data weight to generate first search association sorting data.

[0029] Through keyword analysis and association weight calculation, the system of the present invention can more accurately identify user intentions, thereby providing more relevant search results. By identifying the input type and parsing specific keywords, the system can effectively process different types of input data, improving the efficiency and accuracy of data processing. By setting the search association weight threshold, the system can intelligently filter and sort search results to provide users with a more intelligent and personalized search experience. Accurate and efficient search results can significantly improve user satisfaction, and users can find the information they need more quickly. This process can be applied to different types of data and input methods, making the system highly scalable and adaptable to various application scenarios.

[0030] Preferably, step S22 includes the following steps:

[0031] Step S221: performing input type recognition on the user input data to obtain user input type data, wherein the input type recognition includes text recognition, voice recognition, image recognition and video recognition;

[0032] Step S222: when it is confirmed that the user input type data is text, a text semantic analysis is performed on the user input data to generate text semantic analysis data; keywords are extracted from the user input data according to the text semantic analysis data to generate text keyword data;

[0033] Step S223: When it is confirmed that the user input type data is audio, Fourier transform is performed on the user input data to obtain a user input spectrum graph; Mel frequency cepstral coefficients are extracted from the user input spectrum graph to obtain user input audio features; text transcription is performed on the user input audio features to generate audio conversion keyword data;

[0034] Step S224: when it is confirmed that the user input type data is an image, image content recognition is performed on the user input data to generate image content recognition data, wherein the image content recognition includes image object recognition and image scene recognition; image content description is performed on the image content recognition data to generate image conversion keyword data;

[0035] Step S225: when it is confirmed that the user input type data is a video, extracting video frames from the user input data based on a preset time interval to obtain a user input video frame set; performing frame set change screening on the user input video frame set to obtain a user input change frame set; performing frame set image content recognition on the user input change frame set to generate frame set image content data; performing frame set image content description on the frame set image content data to generate video conversion keyword data;

[0036] Step S226: Comprehensively extract the text keyword data, the audio conversion keyword data, the image conversion keyword data, and the video conversion keyword data to obtain the user input keyword data.

[0037] The present invention enables the system to adapt to diverse user needs and input methods by processing different types of user input (text, audio, image and video). By performing specific processing and keyword extraction on each input type, the system can generate more accurate and relevant keyword data and improve the accuracy of the search. The processing is carried out in steps, and each step is optimized for a specific input type to ensure the efficiency and accuracy of the processing. Through semantic analysis, spectrum analysis, content recognition and other technologies, the system can deeply understand the content input by the user, thereby generating more valuable keywords. The different types of keyword data are processed comprehensively to ensure that the keyword data finally generated can fully reflect the user's search intention, improve the relevance of the search results and user satisfaction. By processing multiple input methods of the user and providing high-quality search results, the user experience is significantly improved.

[0038] Preferably, step S3 comprises the following steps:

[0039] Step S31: collecting user feedback data on the first search related ranking data to obtain first search user feedback data, wherein the feedback data collection includes user click frequency collection and user stay time collection;

[0040] Step S32: allocating attention weights to the first search related ranking data according to the first search user feedback data to generate first search related data attention weights;

[0041] Step S33: comparing the attention weight of the first search-related data with a preset standard data attention weight threshold, and when the attention weight of the first search-related data is greater than or equal to the preset standard data attention weight threshold, performing high search level screening on the first search user feedback data to obtain high search level screening data;

[0042] Step S34: perform similarity calculation on the high search level screening data using a search similarity calculation formula to obtain high search level similarity; perform secondary screening on the high search level screening data according to the high search level similarity to obtain second search related data.

[0043] Through user feedback data, the present invention enables the system to better understand user preferences and behaviors, thereby optimizing the ranking of search results and making the results more in line with user needs. Through attention weight allocation and high search volume screening, the system can dynamically adjust the search algorithm to ensure that the most relevant search results are ranked first. Collecting data such as click frequency and dwell time enables the system to more accurately analyze user behavior and improve the accuracy of search results. Through staged screening and similarity calculation, the system can efficiently process large amounts of search data and ensure the quality and performance of search results. The optimized search results are more in line with user expectations and can significantly improve the user's search experience and satisfaction. By continuously collecting and analyzing user feedback data, the system can continuously improve the search algorithm and results to maintain the efficiency and relevance of the search engine.

[0044] Preferably, the search similarity calculation formula in step S34 is as follows:

[0045]

[0046] In the formula, S(Q,D) represents the similarity between query Q and document D, a represents the lower limit of search time, b represents the upper limit of search time, n represents the number of queries, and w represents the number of queries. i Expressed as the weight of the i-th screening feature, λ i Expressed as the attenuation coefficient of the i-th screening feature, q i Represented as the value of the i-th feature of the query Q, d i It is represented as the value of the i-th feature of document D, exp(-λ i (q i -d i ) 2 ) represents the similarity between the i-th features, g(t) represents the contribution adjustment coefficient that changes with time point t, and σ represents the standard deviation of the normal distribution.

[0047] The present invention analyzes and integrates a search similarity calculation formula. The integral term in the formula is used to accumulate similarity values ​​in the interval [a, b], which can capture the similarity changes between queries and documents in different feature spaces and time dimensions. Through integration, dynamic or time-varying similarity calculations can be processed to adapt to different application scenarios. Sum all n features and comprehensively consider the contribution of each feature to the similarity. Ensure that the similarity calculation covers all relevant features and provide a comprehensive similarity assessment. Weight w iRepresents the weight of each feature i, adjusting the importance of the feature in the similarity calculation. By adjusting the weight, important features can be given priority, the influence of unimportant features can be ignored or reduced, and the accuracy and relevance of the calculation can be improved. Gaussian function exp(-λ i (q i -d i ) 2 )) represents the similarity between the i-th features, as q i and d i The parameter λ decreases exponentially with the increase of the difference. i Control the influence of feature differences and provide adjustments for the sensitivity of different features. The form of the Gaussian function ensures that the similarity is high when the feature difference is small, and the similarity drops rapidly when the difference is large, which is in line with the actual situation. The weight function g(t) can capture the change of feature similarity over time and adapt to the processing of dynamic data and time series data. Normalization constant Ensure the normalization of the similarity calculation results so that they conform to the normal distribution and are easy to explain and compare. Standardization reduces possible deviations in the calculation process and provides stable similarity calculation results. When using the conventional search similarity calculation formula in the field, the similarity between the query Q and the document D can be obtained. By applying the search similarity calculation formula provided by the present invention, the similarity between the query Q and the document D can be calculated more accurately. The formula takes into account multiple features and their weights, and integrates them through Gaussian functions and integrals to provide a comprehensive similarity evaluation. The use of normalization constants and Gaussian functions ensures the stability and robustness of the similarity calculation and reduces the influence of noise and outliers. Through the synergistic effect of multiple parameters, accurate, comprehensive and stable similarity evaluation can be provided in high-search-level data processing to adapt to different application scenarios and needs.

[0048] Preferably, step S4 comprises the following steps:

[0049] Step S41: using blockchain to perform distributed and trusted storage of the second search-related data to generate second search-related storage data;

[0050] Step S42: performing a user search behavior preference analysis on the user input data according to the second search-related storage data to generate user search behavior preference data;

[0051] Step S43: construct an intelligent user behavior profile for the user search behavior preference data to generate a user search behavior profile; use a deep neural network algorithm to construct a personalized prediction model for the user search behavior profile to generate a personalized intelligent search recommendation model;

[0052] Step S44: Perform intelligent search recommendation on the user input data based on the personalized intelligent search recommendation model, generate a recommendation result set, and perform intelligent screening of the target search data.

[0053] The present invention uses blockchain technology for distributed storage to ensure the security, integrity and immutability of data, thereby enhancing users' trust in the system. By conducting in-depth analysis of user search behavior, the system can accurately understand the user's search preferences and provide a reliable basis for personalized recommendations. By building intelligent user behavior portraits and personalized prediction models, the system can provide personalized search recommendations based on the user's specific needs and interests, thereby improving user satisfaction. Using deep neural network algorithms, the system can quickly and efficiently perform intelligent search recommendations on user input data, thereby improving the relevance and accuracy of search results. By continuously analyzing user search behavior and preferences, the system can continuously optimize the recommendation model and maintain the efficiency and accuracy of the search engine.

[0054] Preferably, step S43 includes the following steps:

[0055] Step S431: performing context information aggregation on the user search behavior preference data to generate a user context information data set;

[0056] Step S432: performing time series analysis on the user context information data set to generate user time series feature data;

[0057] Step S433: Performing user behavior analysis on the user context information data set according to the user time series feature data to generate user behavior data; constructing a user profile based on the user behavior data to generate a user search behavior profile;

[0058] Step S434: performing multimodal feature fusion on the user context information dataset, the user time series feature data and the user search behavior profile to generate a fused feature set; performing dataset division on the fused feature set to generate a model training set and a model test set;

[0059] Step S435: Use the DNN neural network algorithm to perform model training on the model training set to generate a personalized intelligent search recommendation training model; perform model testing iteration on the personalized intelligent search recommendation training model through the model test set to generate a personalized intelligent search recommendation model.

[0060] Through context information aggregation and time series analysis, the system of the present invention can build a more accurate and comprehensive user search behavior portrait and improve the accuracy of recommendations. Through multimodal feature fusion, the features of multiple data sources are combined together, which improves the model's ability to understand user behavior and enhances the performance of the recommendation system. The fused feature set is trained using a deep neural network algorithm so that the model can learn from complex user behavior patterns and improve the recommendation effect. Through test iterations, the model can be continuously optimized and adjusted to improve the accuracy of recommendations. Based on user portraits and behavior analysis, the system can provide highly personalized search recommendations and improve the relevance of user experience and search results. Accurate and personalized recommendations can not only improve the relevance of search results, but also enhance user dependence and satisfaction with the system.

[0061] In this specification, an intelligent screening system based on target search data is provided, which is used to execute the above-mentioned intelligent screening method based on target search data. The intelligent screening system based on target search data includes:

[0062] The search library integration module is used to obtain the original search data set; perform data preprocessing on the original search data set to generate a standard search data set; and construct a database based on the standard search data set to obtain a search database;

[0063] A shallow search association module is used to obtain user input data; perform keyword analysis on the user input data to generate user input keyword data; perform search association on the search database according to the user input keyword data to generate a first search association data set; compare the first search association data weight with a preset search association weight threshold, and when the first search association data weight is greater than or equal to the preset search association weight threshold, sort and display the first search association data set corresponding to the first search association data weight to generate first search association sorting data;

[0064] A deep search association module is used to assign attention weights to the first search association ranking data to generate first search association data attention weights; compare the first search association data attention weights with a preset standard data attention weight threshold to obtain high search level screening data; perform similarity calculation on the high search level screening data to obtain high search level similarity; perform secondary screening on the high search level screening data according to the high search level similarity to obtain second search association data;

[0065] The intelligent recommendation module is used to use the blockchain to perform user search behavior preference analysis on the second search related data to generate user search behavior preference data; to construct an intelligent user behavior portrait for the user search behavior preference data to generate a user search behavior portrait; to use a deep neural network algorithm to perform intelligent search recommendations on the user search behavior portrait to generate a recommendation result set to perform intelligent screening of target search data.

[0066] The beneficial effect of the present invention is to generate a standard search data set by obtaining and preprocessing the original search data set. This step can help extract and clean the original data to meet the needs of subsequent processing and ensure the accuracy and consistency of the data. After the user input data is parsed by keywords, it will be used for associated search in the search database to generate a first search associated data set. According to the preset search associated weight threshold, the first search associated data is sorted and screened to generate the first search associated sorted data. This step helps to match and sort the user input with the content in the search database and provide search results related to the user query. By assigning attention weights to the first search associated sorted data, high search level screening data can be obtained. Then, similarity calculation is performed on the high search level screening data to further screen out the second search associated data with high similarity. This step helps to provide more relevant and accurate search results. By performing user search behavior preference analysis on the second search associated data, user search behavior preference data can be generated. Then, a deep neural network algorithm can be used to construct a user search behavior portrait, and intelligent search recommendation is performed based on the portrait to generate a recommended result set. This step helps to provide users with personalized search recommendation results that meet their preferences and needs, which helps to improve the search effect and user experience of the search engine. By processing and analyzing search data, and understanding and modeling user behavior, more accurate and personalized search results can be provided. Therefore, the present invention improves the accuracy and intelligence of search results through data preprocessing, keyword analysis, attention weight allocation and intelligent recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A schematic diagram of the steps of an intelligent screening method based on target search data;

[0068] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0069] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0070] Figure 4 for Figure 1Detailed implementation steps of step S4 in FIG.

[0071] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0072] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0073] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different network and / or processor methods and / or microcontroller methods.

[0074] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0075] To achieve this, please refer to Figures 1 to 4 , an intelligent screening method based on target search data, the method comprising the following steps:

[0076] Step S1: obtaining an original search data set; performing data preprocessing on the original search data set to generate a standard search data set; constructing a database based on the standard search data set to obtain a search database;

[0077] Step S2: obtaining user input data; performing keyword analysis on the user input data to generate user input keyword data; performing search association on the search database according to the user input keyword data to generate a first search association data set; comparing the first search association data weight with a preset search association weight threshold, and when the first search association data weight is greater than or equal to the preset search association weight threshold, sorting and displaying the first search association data set corresponding to the first search association data weight to generate first search association sorting data;

[0078] Step S3: allocating attention weights to the first search-related ranking data to generate first search-related data attention weights; comparing the first search-related data attention weights with a preset standard data attention weight threshold to obtain high search-level screening data; performing similarity calculations on the high search-level screening data to obtain high search-level similarities; performing secondary screening on the high search-level screening data according to the high search-level similarities to obtain second search-related data;

[0079] Step S4: Utilize blockchain to perform user search behavior preference analysis on the second search-related data to generate user search behavior preference data; construct an intelligent user behavior profile for the user search behavior preference data to generate a user search behavior profile; utilize a deep neural network algorithm to perform intelligent search recommendations on the user search behavior profile to generate a recommendation result set to perform intelligent screening of target search data.

[0080] The present invention can improve the quality and consistency of data by preprocessing and standardizing the original search data, thereby constructing a search database more effectively. This helps to improve the performance and accuracy of the search engine. Through keyword analysis and search association, relevant search results can be provided according to the user's input. Sorting display enables users to find content of interest more quickly, thereby improving the search experience. Through attention weight allocation and similarity calculation, search results can be further screened and refined, so that users can get more relevant and accurate search recommendations. This can improve user satisfaction and reduce the redundancy of search results. By performing preference analysis and behavior portrait construction on user search behavior, the search engine can better understand the user's interests and needs. Combining deep neural network algorithms for intelligent search recommendations can provide personalized search results, increase user satisfaction and search efficiency. At the same time, blockchain technology is used to ensure the security and credibility of data, further improving the reliability of the search engine. Therefore, the present invention improves the accuracy and intelligence of search results through data preprocessing, keyword analysis, attention weight allocation and intelligent recommendation.

[0081] In the embodiment of the present invention, reference Figure 1The above is a schematic diagram of the steps of an intelligent screening method based on target search data of the present invention. In this example, the intelligent screening method based on target search data includes the following steps:

[0082] Step S1: obtaining an original search data set; performing data preprocessing on the original search data set to generate a standard search data set; constructing a database based on the standard search data set to obtain a search database;

[0083] In an embodiment of the present invention, search data is collected from multiple sources, such as web crawlers, API interfaces, etc. The search data set may include information such as the user's search keywords, search time, clicked links, etc. The data preprocessing stage includes steps such as data cleaning, deduplication, and normalization. For example, duplicate search records are removed, the date and time format is standardized into a unified format, and missing values ​​are processed. This ensures the quality and consistency of the data and prepares for subsequent analysis and modeling. Use a suitable database management system (such as MySQL, MongoDB, etc.) to design a database model based on the structure of the standard search data set. The preprocessed data is then imported into the database. When designing a database, it is necessary to take into account the query requirements of the search data so that subsequent search operations can be performed efficiently.

[0084] Step S2: obtaining user input data; performing keyword analysis on the user input data to generate user input keyword data; performing search association on the search database according to the user input keyword data to generate a first search association data set; comparing the first search association data weight with a preset search association weight threshold, and when the first search association data weight is greater than or equal to the preset search association weight threshold, sorting and displaying the first search association data set corresponding to the first search association data weight to generate first search association sorting data;

[0085] In an embodiment of the present invention, the data is obtained through a search box, voice input, API interface, etc. of a search engine. These data are search keywords, query strings, voice commands, etc. input by users. Natural language processing (NLP) technology or keyword extraction algorithm is used to extract keywords from the data input by users. These keywords are key descriptions of the user's search intent or needs. Using the search engine's retrieval algorithm, relevant search results are retrieved from the search database based on the keywords input by the user. These search results include web pages, articles, pictures, and other content related to the keywords input by the user. The search results are weighted, and the weight calculation can be performed based on indicators such as the relevance, click-through rate, and page views of the search results. Then, the search results are compared with the preset search association weight threshold, and the search results that meet the conditions are screened out, and they are sorted and displayed according to the weights to generate the first search association sorting data.

[0086] Step S3: allocating attention weights to the first search-related ranking data to generate first search-related data attention weights; comparing the first search-related data attention weights with a preset standard data attention weight threshold to obtain high search-level screening data; performing similarity calculations on the high search-level screening data to obtain high search-level similarities; performing secondary screening on the high search-level screening data according to the high search-level similarities to obtain second search-related data;

[0087] In an embodiment of the present invention, an attention weight is assigned to each search result based on indicators such as the relevance, authority, and click-through rate of the search results. These weights can be adjusted according to business needs and user behavior data to ensure that more relevant search results obtain higher attention weights. A preset standard data attention weight threshold is set to compare the attention weights of the first search-related data. If the attention weight of a search result exceeds the preset threshold, it is identified as high-search-level data. For the screened high-search-level data, their relevance can be further evaluated by calculating the similarity between them. The similarity calculation can be performed using a text similarity algorithm, an image similarity algorithm, etc., and is calculated based on the content characteristics of the search results. Based on the calculated similarity value, the high-search-level screened data is screened for a second time to further refine and optimize the search results. Sorting can be performed according to the similarity value, or a similarity threshold can be set to screen out the most relevant search results to generate the second search-related data.

[0088] Step S4: Utilize blockchain to perform user search behavior preference analysis on the second search-related data to generate user search behavior preference data; construct an intelligent user behavior profile for the user search behavior preference data to generate a user search behavior profile; utilize a deep neural network algorithm to perform intelligent search recommendations on the user search behavior profile to generate a recommendation result set to perform intelligent screening of target search data.

[0089] In the embodiment of the present invention, the security, immutability and transparency of search behavior data are ensured by using blockchain technology. Search behavior data may include user clicks, favorites, sharing and other behaviors, and blockchain is used to record and store these data to ensure the credibility of the data. Then, these search behavior data are analyzed to extract the user's search preferences, such as user's favorite topics, keywords, etc. Based on the user's search behavior preference data, a user's behavior profile is constructed. This may include the user's interest areas, search habits, preference tags, etc. Machine learning algorithms and natural language processing technologies can be used to analyze and model the search behavior data to generate a user's behavior profile. The user's behavior profile is modeled and analyzed using algorithms such as deep neural networks (DNN) to understand the user's search preferences and behavior patterns. Then, based on the user's search behavior profile and the current search environment, a recommendation algorithm is used to generate personalized search recommendation results for the user. These recommendation results may be articles, products, services, etc. related to the user's interests. The generated recommendation result set is used to intelligently filter the target search data. According to the user's search preferences and behavior profiles, search results with high relevance to the user's interests are selected for display. This can improve the relevance of search results and user satisfaction.

[0090] Preferably, step S1 comprises the following steps:

[0091] Step S11: obtaining an original search data set;

[0092] Step S12: performing data cleaning on the original search data set to generate an original search cleaning data set; performing data filtering on the original search cleaning data set to generate an original search filtering data set; performing data format recognition on the original search filtering data set, and performing data format standardization on the recognized original search filtering data set to generate a search format data set;

[0093] Step S13: classifying the search format data set for data duplication to obtain search non-duplicate data and search duplicate data and remove the search duplicate data;

[0094] Step S14: performing data missing value detection on the search non-duplicate data to generate a search non-duplicate data missing detection result; performing data repair on the search non-duplicate data using the search non-duplicate data missing detection result to generate a standard search data set;

[0095] Step S15: construct a database based on the standard search data set to obtain a search database.

[0096] In an embodiment of the present invention, search data is collected from various sources. The data can come from website logs, user behavior records, API call records, etc. The collected data includes information such as search keywords, search time, search results, etc. The data is cleaned to remove or repair erroneous, inconsistent or invalid data. For example, remove duplicate records, repair format errors, process missing values, etc. Data filtering is to filter out data that meets specific conditions or rules, including deleting irrelevant data, removing noise data, filtering out outliers, etc. Convert the data into a unified format for subsequent processing and analysis, including converting data from different data types (such as text, date, number, etc.) into a unified format, unifying the naming conventions of data fields, etc. Perform duplicate detection and classification on the cleaned and filtered data. Duplicate data is identified and classified as duplicate data sets, while non-duplicate data is retained. Perform missing value detection on non-duplicate data sets to identify the location and type of missing values. According to the missing value detection results, an appropriate method is used to repair or fill in the missing values. Common methods include using the mean, median, interpolation, etc. to fill in missing values. Based on the processed standard search data set, the search database is constructed using appropriate database technology, which involves operations such as creating database table structure, defining indexes, setting data relationships, etc. to support efficient storage, retrieval and query of data.

[0097] Preferably, step S14 comprises the following steps:

[0098] Step S141: using the conditional random field to perform data missing value detection on the search non-repeated data to obtain the search non-repeated data missing detection result;

[0099] Step S142: filling missing values ​​in the preliminary detected abnormal data set according to the search non-repetitive data missing detection result to generate a search filled data set; performing data consistency detection on the search filled data to obtain a data consistency detection result; performing data anomaly identification on the search filled data set according to the data consistency detection result to generate a data anomaly type identification result;

[0100] Step S143: When it is confirmed that the data anomaly type identification result is a semantic anomaly, a knowledge graph method is used to perform semantic repair on the search filling data set to generate a semantic repair data set;

[0101] Step S144: when it is confirmed that the data anomaly type identification result is a statistical anomaly, Bayesian inference is used to statistically repair the search filling data set to generate a statistical repair data set;

[0102] Step S145: Perform repair verification on the semantic repair data set and the statistical repair data set to generate a standard search data set.

[0103] In an embodiment of the present invention, a conditional random field (CRF) model is used to detect missing values ​​in search non-repeated data. CRF is a probabilistic graph model commonly used in sequence labeling tasks. Here, the search data can be regarded as a sequence, and the missing values ​​in the sequence are detected using the CRF model to obtain the missing detection result of the search non-repeated data. According to the missing detection result of the search non-repeated data, the missing values ​​of the abnormal data set preliminarily detected are filled to generate a search filling data set. The filled data is subjected to consistency detection to ensure the internal consistency of the data, for example, to ensure the association relationship between different fields and the consistency of the data format. If the data consistency detection result confirms that there is a semantic anomaly, that is, the data does not conform to the expected semantic rules, the search filling data set can be semantically repaired using the knowledge graph method. The knowledge graph can provide connection information between semantic relations and entities, which is used to repair semantic problems in the data and generate a semantic repair data set. If the data consistency detection result confirms that there is a statistical anomaly, that is, the data does not statistically conform to the expected distribution law, the Bayesian inference method can be used to statistically repair the search filling data set. Bayesian inference can infer the value of missing values ​​by modeling the probability distribution of data, so as to fill in the missing values ​​and generate a statistical repair data set. The repair verification is performed on the semantically repaired and statistically repaired data set to ensure that the repaired data meets the expected quality standards and generate the final standard search data set.

[0104] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0105] Step S21: Obtain user input data;

[0106] Step S22: performing input type identification on the user input data to obtain user input type data; performing keyword analysis on the user input data based on the user input type data to generate user input keyword data;

[0107] Step S23: performing search association on the search database according to the keyword data input by the user to generate a first search association data set; performing association weight calculation on the first search association data set to obtain a first search association data weight;

[0108] Step S24: Compare the first search association data weight with the preset search association weight threshold. When the first search association data weight is greater than or equal to the preset search association weight threshold, sort and display the first search association data set corresponding to the first search association data weight to generate first search association sorting data.

[0109] In the embodiment of the present invention, input data is collected from the user, specifically data in the form of text, voice, image, etc. According to the characteristics or metadata of the data, the system uses a machine learning algorithm or a rule engine to identify the data type. For example, a text classification model can be used to identify text data, a voice recognition model can be used to identify voice data, and an image classification model can be used to identify image data. For text data, the system can use a word segmenter to split the text into words or phrases, and perform processing such as part-of-speech tagging and named entity recognition to identify keywords. For voice data, the system can use voice recognition technology to convert the voice into text form, and then extract keywords. For image data, the system can use image processing and feature extraction technology to identify key information in the image, such as identifying objects in the image through a target detection algorithm. Search association is performed in the search database according to the keyword data input by the user. The search results can be sorted according to factors such as keyword matching degree and relevance, and the association weight of each search result is calculated. The association weight in the search result is compared with a preset search association weight threshold. If the association weight of a search result is greater than or equal to the preset search association weight threshold, the search result is added to the final search result set, and the search results are sorted and displayed according to the association weight. You can sort in descending order based on the association weight to ensure that users first see the results most relevant to their input. For example, if a user enters text data in the search box: "car maintenance", the type of the input text data is identified to determine that it is a text type, and the text data is decomposed into keywords: "car", "maintenance" using a word segmenter. The search database is searched for relevant data containing the keywords "car" and "maintenance". The association weight is calculated based on the relevance of the search results. Assume that the search results are as follows:

[0110] , Result 1: Title: "Car Maintenance Guide", Relevance: 0.8, Result 2: Title: "Car Maintenance FAQ", Relevance: 0.7, Result 3: Title: "Car Repair Tips", Relevance: 0.5 For each result, a weight is given according to the relevance. For example, the weight of Result 1 is 0.8, the weight of Result 2 is 0.7, and the weight of Result 3 is 0.5. If the preset search association weight threshold is 0.6, the weights of Result 1 and Result 2 meet the conditions and can be added to the final search result set.

[0111] Preferably, step S22 includes the following steps:

[0112] Step S221: performing input type recognition on the user input data to obtain user input type data, wherein the input type recognition includes text recognition, voice recognition, image recognition and video recognition;

[0113] Step S222: when it is confirmed that the user input type data is text, a text semantic analysis is performed on the user input data to generate text semantic analysis data; keywords are extracted from the user input data according to the text semantic analysis data to generate text keyword data;

[0114] Step S223: When it is confirmed that the user input type data is audio, Fourier transform is performed on the user input data to obtain a user input spectrum graph; Mel frequency cepstral coefficients are extracted from the user input spectrum graph to obtain user input audio features; text transcription is performed on the user input audio features to generate audio conversion keyword data;

[0115] Step S224: when it is confirmed that the user input type data is an image, image content recognition is performed on the user input data to generate image content recognition data, wherein the image content recognition includes image object recognition and image scene recognition; image content description is performed on the image content recognition data to generate image conversion keyword data;

[0116] Step S225: when it is confirmed that the user input type data is a video, extracting video frames from the user input data based on a preset time interval to obtain a user input video frame set; performing frame set change screening on the user input video frame set to obtain a user input change frame set; performing frame set image content recognition on the user input change frame set to generate frame set image content data; performing frame set image content description on the frame set image content data to generate video conversion keyword data;

[0117] Step S226: Comprehensively extract the text keyword data, the audio conversion keyword data, the image conversion keyword data, and the video conversion keyword data to obtain the user input keyword data.

[0118] In the embodiment of the present invention, the input type is identified according to the features and metadata of the user input data. Specific input types include text, voice, image and video. If the user input type data is confirmed to be text, the system performs the following processing: Text semantic analysis: semantic analysis of the text is performed to understand the meaning and meaning of the text. Keyword extraction: based on the text semantic analysis results, keywords are extracted as the keyword data input by the user. If the user input type data is confirmed to be audio, the system performs the following processing: Fourier transform: Fourier transform the audio data to convert the audio signal into a spectrogram. Mel frequency cepstral coefficient extraction: Mel frequency cepstral coefficients are extracted from the spectrogram as audio features. Text transcription: audio is transcribed into text to extract keywords as the keyword data input by the user. If the user input type data is confirmed to be an image, the system performs the following processing: Image content recognition: image object recognition and image scene recognition technology are used to recognize the image content. Image content description: the image content is described according to the recognition result, and keywords are extracted as the keyword data input by the user. If the user input type data is confirmed to be video, the system performs the following processing: Video frame extraction: Frame extraction of video data based on a preset time interval to obtain a video frame set. Frame set change screening: Screen out frames that have changed in the video frame set as the changed frame set input by the user. Frame set image content recognition: Image content recognition is performed on each frame in the changed frame set, and keywords are extracted as the keyword data input by the user. Frame set image content description: Each frame in the frame set is described according to the recognition result, and keywords are extracted as the keyword data input by the user. Comprehensive keyword extraction is performed on text keyword data, audio conversion keyword data, image conversion keyword data, and video conversion keyword data to obtain the keyword data input by the user.

[0119] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0120] Step S31: collecting user feedback data on the first search related ranking data to obtain first search user feedback data, wherein the feedback data collection includes user click frequency collection and user stay time collection;

[0121] Step S32: allocating attention weights to the first search related ranking data according to the first search user feedback data to generate first search related data attention weights;

[0122] Step S33: comparing the attention weight of the first search-related data with a preset standard data attention weight threshold, and when the attention weight of the first search-related data is greater than or equal to the preset standard data attention weight threshold, performing high search level screening on the first search user feedback data to obtain high search level screening data;

[0123] Step S34: perform similarity calculation on the high search level screening data using a search similarity calculation formula to obtain high search level similarity; perform secondary screening on the high search level screening data according to the high search level similarity to obtain second search related data.

[0124] In an embodiment of the present invention, feedback information of the user on the first search related ranking data is collected, including the user click frequency and the user stay time, for example, the user click frequency data: {result 1: 10 times, result 2: 5 times, result 3: 3 times}, the user stay time data: {result 1: 30 seconds, result 2: 20 seconds, result 3: 15 seconds}. According to the first search user feedback data, the system assigns an attention weight to each search result to reflect the user's attention to the search result, wherein the attention weight is assigned as a weighted calculation based on the user click frequency and the stay time. According to the attention weight of the first search related data, the system screens out search results with higher attention for further processing, for example: standard data attention weight threshold: 0.5, high search level screening data: {result 1: attention weight 0.7, result 2: attention weight 0.6}. The similarity between the search results is calculated based on the high search level screening data, and secondary screening is performed. A search similarity calculation formula such as cosine similarity or other similarity measurement methods is used.

[0125] Preferably, the search similarity calculation formula in step S34 is as follows:

[0126]

[0127] In the formula, S(Q,D) represents the similarity between query Q and document D, a represents the lower limit of search time, b represents the upper limit of search time, n represents the number of queries, and w represents the number of queries. i Expressed as the weight of the i-th screening feature, λ i Expressed as the attenuation coefficient of the i-th screening feature, q i Represented as the value of the i-th feature of the query Q, d i It is represented as the value of the i-th feature of document D, exp(-λ i (q i -d i ) 2 ) represents the similarity between the i-th features, g(t) represents the contribution adjustment coefficient that changes with time point t, and σ represents the standard deviation of the normal distribution.

[0128] The present invention analyzes and integrates a search similarity calculation formula. The integral term in the formula is used to accumulate similarity values ​​in the interval [a, b], which can capture the similarity changes between queries and documents in different feature spaces and time dimensions. Through integration, dynamic or time-varying similarity calculations can be processed to adapt to different application scenarios. Sum all n features and comprehensively consider the contribution of each feature to the similarity. Ensure that the similarity calculation covers all relevant features and provide a comprehensive similarity assessment. Weight w i Represents the weight of each feature i, adjusting the importance of the feature in the similarity calculation. By adjusting the weight, important features can be given priority, the influence of unimportant features can be ignored or reduced, and the accuracy and relevance of the calculation can be improved. Gaussian function exp(-λ i (q i -d i ) 2 )) represents the similarity between the i-th features, as q i and d i The parameter λ decreases exponentially with the increase of the difference. i Control the influence of feature differences and provide adjustments for the sensitivity of different features. The form of the Gaussian function ensures that the similarity is high when the feature difference is small, and the similarity drops rapidly when the difference is large, which is in line with the actual situation. The weight function g(t) can capture the change of feature similarity over time and adapt to the processing of dynamic data and time series data. Normalization constant Ensure the normalization of the similarity calculation results so that they conform to the normal distribution and are easy to explain and compare. Standardization reduces possible deviations in the calculation process and provides stable similarity calculation results. When using the conventional search similarity calculation formula in the field, the similarity between the query Q and the document D can be obtained. By applying the search similarity calculation formula provided by the present invention, the similarity between the query Q and the document D can be calculated more accurately. The formula takes into account multiple features and their weights, and integrates them through Gaussian functions and integrals to provide a comprehensive similarity evaluation. The use of normalization constants and Gaussian functions ensures the stability and robustness of the similarity calculation and reduces the influence of noise and outliers. Through the synergistic effect of multiple parameters, accurate, comprehensive and stable similarity evaluation can be provided in high-search-level data processing to adapt to different application scenarios and needs.

[0129] As an example of the present invention, refer to Figure 4 As shown, in this example, step S4 includes:

[0130] Step S41: using blockchain to perform distributed and trusted storage of the second search-related data to generate second search-related storage data;

[0131] Step S42: performing a user search behavior preference analysis on the user input data according to the second search-related storage data to generate user search behavior preference data;

[0132] Step S43: construct an intelligent user behavior profile for the user search behavior preference data to generate a user search behavior profile; use a deep neural network algorithm to construct a personalized prediction model for the user search behavior profile to generate a personalized intelligent search recommendation model;

[0133] Step S44: Perform intelligent search recommendation on the user input data based on the personalized intelligent search recommendation model, generate a recommendation result set, and perform intelligent screening of the target search data.

[0134] In the embodiment of the present invention, the second search-related data (including search records, click records, etc.) is stored through blockchain technology by using a search engine. This means that the data is stored in a dispersed manner on multiple nodes, and each node has a complete copy of the data, thereby ensuring the security and credibility of the data. The immutability and decentralization characteristics of the blockchain can prevent data tampering and single point failure. The search engine analyzes the second search-related data stored on the blockchain, including information such as the user's search keywords, clicked links, and search time. Then, through data analysis and mining technology, a preference analysis is performed on the user's search behavior, such as determining the user's preference for certain topics, commonly used search terms, etc., and the results are generated as user search behavior preference data. The search engine uses the user search behavior preference data to build a user's behavior profile, including the user's interests, preferences, behavioral habits, etc. Then, a personalized prediction model is constructed using machine learning algorithms such as deep neural networks. The model can predict the content of interest to the user based on the user's behavior profile, and generate a personalized intelligent search recommendation model. Based on the personalized intelligent search recommendation model constructed above, the search data entered by the user is intelligently recommended. By analyzing the user's behavior profile and preferences, search results related to the user's interests are recommended, thereby providing users with a more intelligent and personalized search experience.

[0135] Preferably, step S43 includes the following steps:

[0136] Step S431: performing context information aggregation on the user search behavior preference data to generate a user context information data set;

[0137] Step S432: performing time series analysis on the user context information data set to generate user time series feature data;

[0138] Step S433: Performing user behavior analysis on the user context information data set according to the user time series feature data to generate user behavior data; constructing a user profile based on the user behavior data to generate a user search behavior profile;

[0139] Step S434: performing multimodal feature fusion on the user context information dataset, the user time series feature data and the user search behavior profile to generate a fused feature set; performing dataset division on the fused feature set to generate a model training set and a model test set;

[0140] Step S435: Use the DNN neural network algorithm to perform model training on the model training set to generate a personalized intelligent search recommendation training model; perform model testing iteration on the personalized intelligent search recommendation training model through the model test set to generate a personalized intelligent search recommendation model.

[0141] In an embodiment of the present invention, the context information of the user's search behavior preference data is aggregated, including the time, location, device information, etc. of the user's search, to generate the user's context information data set. This context information can help better understand the background and intention behind the user's search behavior. The user's context information data set is subjected to time series analysis, such as the changing trend of the user's search frequency over time, to generate the user's time series feature data. These feature data can reflect the time correlation and changing law of the user's search behavior. The user's context information data set is further analyzed according to the user's time series feature data to generate the user's behavior data. Then, based on these behavior data, a user's search behavior profile is constructed, including the user's interests, preferences, behavior habits, etc. The user's context information data set, time series feature data, and search behavior profile are multimodally fused to generate a fused feature set. Then, the fused feature set is divided into a model training set and a model test set for use in training and testing a personalized intelligent search recommendation model. The model training set is trained using algorithms such as a deep neural network (DNN) to generate a training model for personalized intelligent search recommendation. Then, the training model is tested and iterated through the model test set to optimize the performance of the model, and finally a personalized intelligent search recommendation model is generated to provide users with personalized search recommendation results. More specifically, for example, a user performs multiple searches in a week, and the search results on weekdays are different from those on weekends. By aggregating these search behaviors and adding contextual information (such as search time, search location, search device, etc.), a data set containing the background information of the user's search behavior can be generated. The user's search behavior within a week is analyzed in time series, such as analyzing whether the search frequency is significantly different on weekdays and weekends, or whether there is a trend of increasing or decreasing user search behavior in a specific time period. These analyses can generate the user's time series feature data. If it is found that the user is more inclined to search for leisure and entertainment content on weekends, this behavior pattern can be included in the user's search behavior profile. The user's search behavior profile is associated with contextual information and time series features. Then, the fused feature set is divided into training and testing data sets for use in building a personalized intelligent search recommendation model. The training set is trained using algorithms such as deep neural networks (DNN) to generate a training model for personalized intelligent search recommendation. Then, the training model is tested and iterated through the test set to optimize the model performance. The final generated model can be used to provide users with personalized search recommendation results.

[0142] In this specification, an intelligent screening system based on target search data is provided, which is used to execute the above-mentioned intelligent screening method based on target search data. The intelligent screening system based on target search data includes:

[0143] The search library integration module is used to obtain the original search data set; perform data preprocessing on the original search data set to generate a standard search data set; and construct a database based on the standard search data set to obtain a search database;

[0144] A shallow search association module is used to obtain user input data; perform keyword analysis on the user input data to generate user input keyword data; perform search association on the search database according to the user input keyword data to generate a first search association data set; compare the first search association data weight with a preset search association weight threshold, and when the first search association data weight is greater than or equal to the preset search association weight threshold, sort and display the first search association data set corresponding to the first search association data weight to generate first search association sorting data;

[0145] A deep search association module is used to assign attention weights to the first search association ranking data to generate first search association data attention weights; compare the first search association data attention weights with a preset standard data attention weight threshold to obtain high search level screening data; perform similarity calculation on the high search level screening data to obtain high search level similarity; perform secondary screening on the high search level screening data according to the high search level similarity to obtain second search association data;

[0146] The intelligent recommendation module is used to use the blockchain to perform user search behavior preference analysis on the second search related data to generate user search behavior preference data; to construct an intelligent user behavior portrait for the user search behavior preference data to generate a user search behavior portrait; to use a deep neural network algorithm to perform intelligent search recommendations on the user search behavior portrait to generate a recommendation result set to perform intelligent screening of target search data.

[0147] The beneficial effect of the present invention is to generate a standard search data set by obtaining and preprocessing the original search data set. This step can help extract and clean the original data to meet the needs of subsequent processing and ensure the accuracy and consistency of the data. After the user input data is parsed by keywords, it will be used for associated search in the search database to generate a first search associated data set. According to the preset search associated weight threshold, the first search associated data is sorted and screened to generate the first search associated sorted data. This step helps to match and sort the user input with the content in the search database and provide search results related to the user query. By assigning attention weights to the first search associated sorted data, high search level screening data can be obtained. Then, similarity calculation is performed on the high search level screening data to further screen out the second search associated data with high similarity. This step helps to provide more relevant and accurate search results. By performing user search behavior preference analysis on the second search associated data, user search behavior preference data can be generated. Then, a deep neural network algorithm can be used to construct a user search behavior portrait, and intelligent search recommendation is performed based on the portrait to generate a recommended result set. This step helps to provide users with personalized search recommendation results that meet their preferences and needs, which helps to improve the search effect and user experience of the search engine. By processing and analyzing search data, and understanding and modeling user behavior, more accurate and personalized search results can be provided. Therefore, the present invention improves the accuracy and intelligence of search results through data preprocessing, keyword analysis, attention weight allocation and intelligent recommendation.

[0148] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0149] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An intelligent screening method based on target search data, characterized in that: The following steps are involved: Step S1: obtaining an original search data set; performing data preprocessing on the original search data set to generate a standard search data set; constructing a database based on the standard search data set to obtain a search database; Step S2: obtaining user input data; performing keyword analysis on the user input data to generate user input keyword data; performing search association on the search database according to the user input keyword data to generate a first search association data set; comparing the first search association data weight with a preset search association weight threshold, and when the first search association data weight is greater than or equal to the preset search association weight threshold, sorting and displaying the first search association data set corresponding to the first search association data weight to generate first search association sorting data; Step S3: performing attention weight allocation on the first search related ranking data to generate the first search related data attention weight; Comparing the attention weight of the first search-related data with a preset standard data attention weight threshold to obtain high-search-level screening data; performing similarity calculation on the high-search-level screening data to obtain high-search-level similarity; performing secondary screening on the high-search-level screening data according to the high-search-level similarity to obtain second-search-related data; Step S4: Utilizing blockchain to perform user search behavior preference analysis on the second search-related data to generate user search behavior preference data; Build intelligent user behavior profiles based on user search behavior preference data to generate user search behavior profiles; Use deep neural network algorithms to perform intelligent search recommendations on user search behavior profiles and generate recommendation result sets to perform intelligent screening of target search data.

2. The intelligent screening method based on target search data according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: obtaining an original search data set; Step S12: performing data cleaning on the original search data set to generate an original search cleaning data set; performing data filtering on the original search cleaning data set to generate an original search filtering data set; performing data format recognition on the original search filtering data set, and performing data format standardization on the recognized original search filtering data set to generate a search format data set; Step S13: classifying the search format data set for data duplication, obtaining search non-duplicate data and search duplicate data, and removing the search duplicate data; Step S14: performing data missing value detection on the search non-duplicate data to generate a search non-duplicate data missing detection result; performing data repair on the search non-duplicate data using the search non-duplicate data missing detection result to generate a standard search data set; Step S15: construct a database based on the standard search data set to obtain a search database.

3. The intelligent screening method based on target search data according to claim 2 is characterized in that: Step S14 includes the following steps: Step S141: using the conditional random field to perform data missing value detection on the search non-repeated data to obtain the search non-repeated data missing detection result; Step S142: filling missing values ​​in the preliminary detected abnormal data set according to the search non-repetitive data missing detection result to generate a search filled data set; performing data consistency detection on the search filled data to obtain a data consistency detection result; performing data anomaly identification on the search filled data set according to the data consistency detection result to generate a data anomaly type identification result; Step S143: When it is confirmed that the data anomaly type identification result is a semantic anomaly, a knowledge graph method is used to perform semantic repair on the search filling data set to generate a semantic repair data set; Step S144: when it is confirmed that the data anomaly type identification result is a statistical anomaly, Bayesian inference is used to statistically repair the search filling data set to generate a statistical repair data set; Step S145: Perform repair verification on the semantic repair data set and the statistical repair data set to generate a standard search data set.

4. The intelligent screening method based on target search data according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Obtain user input data; Step S22: performing input type identification on the user input data to obtain user input type data; performing keyword analysis on the user input data based on the user input type data to generate user input keyword data; Step S23: performing search association on the search database according to the keyword data input by the user to generate a first search association data set; performing association weight calculation on the first search association data set to obtain a first search association data weight; Step S24: Compare the first search association data weight with the preset search association weight threshold. When the first search association data weight is greater than or equal to the preset search association weight threshold, sort and display the first search association data set corresponding to the first search association data weight to generate first search association sorting data.

5. The intelligent screening method based on target search data according to claim 4 is characterized in that: Step S22 includes the following steps: Step S221: performing input type recognition on the user input data to obtain user input type data, wherein the input type recognition includes text recognition, voice recognition, image recognition and video recognition; Step S222: when it is confirmed that the user input type data is text, a text semantic analysis is performed on the user input data to generate text semantic analysis data; keywords are extracted from the user input data according to the text semantic analysis data to generate text keyword data; Step S223: When it is confirmed that the user input type data is audio, Fourier transform is performed on the user input data to obtain a user input spectrum graph; Mel frequency cepstral coefficients are extracted from the user input spectrum graph to obtain user input audio features; text transcription is performed on the user input audio features to generate audio conversion keyword data; Step S224: when it is confirmed that the user input type data is an image, image content recognition is performed on the user input data to generate image content recognition data, wherein the image content recognition includes image object recognition and image scene recognition; image content description is performed on the image content recognition data to generate image conversion keyword data; Step S225: when it is confirmed that the user input type data is a video, extracting video frames from the user input data based on a preset time interval to obtain a user input video frame set; performing frame set change screening on the user input video frame set to obtain a user input change frame set; performing frame set image content recognition on the user input change frame set to generate frame set image content data; performing frame set image content description on the frame set image content data to generate video conversion keyword data; Step S226: Comprehensively extract the text keyword data, the audio conversion keyword data, the image conversion keyword data, and the video conversion keyword data to obtain the user input keyword data.

6. The intelligent screening method based on target search data according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: collecting user feedback data on the first search related ranking data to obtain first search user feedback data, wherein the feedback data collection includes user click frequency collection and user stay time collection; Step S32: allocating attention weights to the first search related ranking data according to the first search user feedback data to generate first search related data attention weights; Step S33: comparing the attention weight of the first search-related data with a preset standard data attention weight threshold, and when the attention weight of the first search-related data is greater than or equal to the preset standard data attention weight threshold, performing high search level screening on the first search user feedback data to obtain high search level screening data; Step S34: perform similarity calculation on the high search level screening data using a search similarity calculation formula to obtain high search level similarity; perform secondary screening on the high search level screening data according to the high search level similarity to obtain second search related data.

7. The intelligent screening method based on target search data according to claim 6 is characterized in that: The search similarity calculation formula in step S34 is as follows: In the formula, S(Q,D) represents the similarity between query Q and document D, a represents the lower limit of search time, b represents the upper limit of search time, n represents the number of queries, and w represents the number of queries. i Expressed as the weight of the i-th screening feature, λ i Expressed as the attenuation coefficient of the i-th screening feature, q i Represented as the value of the i-th feature of the query Q, d i It is represented as the value of the i-th feature of document D, exp(-λ i (q i -d i ) 2 ) represents the similarity between the i-th features, g(t) represents the contribution adjustment coefficient that changes with time point t, and σ represents the standard deviation of the normal distribution.

8. The intelligent screening method based on target search data according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: using blockchain to perform distributed and trusted storage of the second search-related data to generate second search-related storage data; Step S42: performing a user search behavior preference analysis on the user input data according to the second search-related storage data to generate user search behavior preference data; Step S43: construct an intelligent user behavior profile for the user search behavior preference data to generate a user search behavior profile; use a deep neural network algorithm to construct a personalized prediction model for the user search behavior profile to generate a personalized intelligent search recommendation model; Step S44: Perform intelligent search recommendation on the user input data based on the personalized intelligent search recommendation model, generate a recommendation result set, and perform intelligent screening of the target search data.

9. The intelligent screening method based on target search data according to claim 8, characterized in that: Step S43 includes the following steps: Step S431: performing context information aggregation on the user search behavior preference data to generate a user context information data set; Step S432: performing time series analysis on the user context information data set to generate user time series feature data; Step S433: Performing user behavior analysis on the user context information data set according to the user time series feature data to generate user behavior data; constructing a user profile based on the user behavior data to generate a user search behavior profile; Step S434: performing multimodal feature fusion on the user context information dataset, the user time series feature data and the user search behavior profile to generate a fused feature set; performing dataset division on the fused feature set to generate a model training set and a model test set; Step S435: Use the DNN neural network algorithm to perform model training on the model training set to generate a personalized intelligent search recommendation training model; perform model testing iteration on the personalized intelligent search recommendation training model through the model test set to generate a personalized intelligent search recommendation model.

10. An intelligent screening system based on target search data, characterized in that: Used to execute the intelligent screening method based on target search data as claimed in claim 1, the intelligent screening system based on target search data comprises: The search library integration module is used to obtain the original search data set; perform data preprocessing on the original search data set to generate a standard search data set; and construct a database based on the standard search data set to obtain a search database; A shallow search association module is used to obtain user input data; perform keyword analysis on the user input data to generate user input keyword data; perform search association on the search database according to the user input keyword data to generate a first search association data set; compare the first search association data weight with a preset search association weight threshold, and when the first search association data weight is greater than or equal to the preset search association weight threshold, sort and display the first search association data set corresponding to the first search association data weight to generate first search association sorting data; A deep search association module is used to assign attention weights to the first search association ranking data to generate first search association data attention weights; compare the first search association data attention weights with a preset standard data attention weight threshold to obtain high search level screening data; perform similarity calculation on the high search level screening data to obtain high search level similarity; perform secondary screening on the high search level screening data according to the high search level similarity to obtain second search association data; The intelligent recommendation module is used to use the blockchain to perform user search behavior preference analysis on the second search related data to generate user search behavior preference data; to construct an intelligent user behavior portrait for the user search behavior preference data to generate a user search behavior portrait; to use a deep neural network algorithm to perform intelligent search recommendations on the user search behavior portrait to generate a recommendation result set to perform intelligent screening of target search data.

Images