A computer information data quick query system

Abstract

The present invention relates to the field of data processing technology, and in particular to a computer information data rapid query system, which is responsible for recording the start and end time of a user's query through a query behavior capture module. The real-time query analysis module determines the query behavior type by calculating the residence time and score of the query result. The index management module generates a temporary index based on the overlap between the query timeline and the request string, and monitors its usage popularity to dynamically adjust the index status. The result output module traverses the temporary index and the regular index according to the query request. It is achieved that only a small amount of fuzzy search is needed to obtain the most relevant query results. The speed and quality of information retrieval are significantly improved, the efficiency of fuzzy query in a big data environment is effectively improved, and a more accurate and fast search experience is provided to users.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a computer information data rapid query system. Background Art

[0002] Retrieving information quickly and accurately from a large amount of data when the known information is inaccurate or ambiguous is a common challenge. With the rapid growth of the amount of information stored on the Internet and in databases, when users are looking for specific information when the known information is inaccurate or ambiguous, they need to go through multiple queries and often need to browse a large amount of irrelevant data, which is inefficient and increases the burden on the server and wastes data.

[0003] Some existing technologies introduce fuzzy matching to address the retrieval challenge when the known information is inaccurate or fuzzy. Fuzzy matching technology allows the system to not only rely on precise keyword matching when processing query requests, but also to identify and respond to data that is similar to or partially matches the query conditions. The introduction of this technology greatly expands the scope of search results and improves the flexibility and accuracy of the system in processing fuzzy queries. Intelligent search algorithms, such as search optimization based on machine learning, further improve the search system's ability to handle fuzzy information. The application of natural language processing (NLP) technology enables the search system to better understand the semantics of user queries and perform effective information retrieval even when the information provided by the user is incomplete or ambiguous. However, although these existing technologies have made progress in processing fuzzy information, they still have some shortcomings. First, fuzzy matching and intelligent search algorithms often require processing large amounts of data and complex calculations, which not only consumes a lot of computing resources, but also may affect the response speed of the search. Second, although machine learning and NLP technologies can improve the accuracy of search, their effectiveness depends largely on the quality and coverage of training data, which may make the system unable to provide satisfactory results when facing new or rare queries. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a computer information data rapid query system.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A computer information data rapid query system, comprising a query behavior capture module, a real-time query analysis module, an index management module and a result output module, wherein the query behavior capture module, the real-time query analysis module, the index management module and the result output module are connected in sequence;

[0007] The query behavior capturing module is used to capture the query behavior and obtain the query start time and query end time;

[0008] The real-time query analysis module is used to obtain a query timeline through the query start time and the query end time, calculate the residence time of each query result on the query timeline and perform normalization calculation through a function to obtain the score of each query result, and determine the query behavior type according to the scores of several query results; the query behavior type includes single-target query and non-single-target query;

[0009] The index management module is used to trace back the query timeline, collect query request strings, calculate the overlap of several strings and generate a temporary index based on the overlap, and then associate the temporary index with the query result according to the query behavior type; the index management module is also used to monitor the usage popularity of the temporary index in real time, determine the usage popularity trend of the temporary index based on the time series algorithm and suspend the temporary index;

[0010] The result output module is used to traverse the query results associated with the temporary index and the regular index according to several query requests and calculate the triggering times of the same query result, and output the final query result according to the triggering times.

[0011] Furthermore, capturing the query behavior and obtaining the query start time and the query end time includes:

[0012] Compare and analyze the query requests in the query behavior to obtain a similarity score, standardize the query request time interval and perform a weighted average with the similarity score to obtain a correlation score;

[0013] According to the correlation scores, query behaviors are correlated with each other to obtain query combinations;

[0014] The time when the earliest query behavior in the query combination occurs is taken as the query start time, and the time when the query result of the latest query behavior in the query combination is obtained is taken as the query end time.

[0015] Furthermore, the stay time of the query result is determined starting from the time point when the user obtains the query result and ending with the time point when the user leaves the query result.

[0016] Furthermore, the calculation method for determining the query behavior type according to the scores of several query results is as follows:

[0017]

[0018] Among them, SD is the standard deviation of the query result score, S i The calculation result of the residence time of the i-th query result after normalization by the function, n represents the total number of query results in the query combination, and the calculation formula of Mean is as follows:

[0019]

[0020] If SD is less than the first preset value and max(S i ) is greater than or equal to the second preset value, it is determined to be a single target query;

[0021] If SD is less than the first preset value and max(S i ) is less than a second preset value, it is determined to be a non-single target query;

[0022] If SD is greater than or equal to the first preset value, it is determined to be a non-single target query.

[0023] Furthermore, the calculating the overlap degree of a plurality of character strings and generating a temporary index according to the overlap degree includes:

[0024] Evaluate the semantic overlap between query request strings based on semantic similarity algorithm;

[0025] Determine the text overlap between query request strings through cosine similarity;

[0026] Temporary indexes are generated and associated according to the semantic overlap and text overlap.

[0027] Furthermore, associating the temporary index with the query result according to the query behavior type includes:

[0028] If the query behavior type is a single-target query, the associated temporary index is associated with the single query result;

[0029] If the query behavior type is a non-single-target query, the associated temporary index is associated with several query results whose similarities reach a preset standard value.

[0030] Furthermore, the real-time monitoring of the usage popularity of the temporary index includes collecting the access frequency and query times of the temporary index and calculating the popularity score by weighted average.

[0031] Furthermore, judging the usage popularity trend of the temporary index based on the time series algorithm and pausing the temporary index includes:

[0032] Use ARIMA to perform trend analysis on the heat score. If the heat score is lower than the preset threshold and the heat trend is negative, the use of the associated temporary index is suspended; if the heat score is lower than the preset threshold and the heat trend is positive, the use of the associated temporary index is retained; if the heat score is higher than the preset threshold, the use of the associated temporary index is retained.

[0033] The beneficial effect of the present invention is that the present invention provides the system with basic data for in-depth understanding of the user's query intention by capturing the user's query behavior, including the start and end time of the query and the query content. The system is allowed to capture the query iteration process when the user is dissatisfied with the preliminary search results, and to separate the non-identical query purposes before and after. The real-time query analysis module obtains the query timeline through the query start time and the query end time, calculates the residence time of each query result on the query timeline, and obtains the score of each query result by normalization calculation through the function to determine the query behavior type. The index management module generates a temporary index according to the overlap between the query timeline backtracking and the query request string, and associates these temporary indexes with the query results through the query behavior type, so that when the temporary index is triggered multiple times, the result output module can directly output the query result with a higher correct probability according to the number of triggers of the common possible query results pointed to by the temporary index and the conventional index according to several query requests. In addition, by real-time monitoring of the use heat of the temporary index and dynamically adjusting the state of the temporary index based on the time series algorithm, the module significantly improves the retrieval efficiency and reduces the waste of resources. The result output module traverses the temporary index and the conventional index according to several query requests and outputs the query result. The next user only needs to conduct a few fuzzy searches to quickly get the corresponding single-target results and non-single-target results by touching the associated temporary index. This avoids browsing a large amount of irrelevant data due to fuzzy search conditions, improves efficiency, and reduces server burden and data waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a structural schematic diagram of a computer information data rapid query system in the present invention.

[0035] Figure 2 It is a flow chart of the steps of capturing query behavior and obtaining query start time and query end time in the present invention. DETAILED DESCRIPTION

[0036] See also Figure 1-2 As shown, the present invention relates to a computer information data rapid query system, comprising a query behavior capture module, a real-time query analysis module, an index management module and a result output module, wherein the query behavior capture module, the real-time query analysis module, the index management module and the result output module are connected in sequence;

[0037] The query behavior capturing module is used to capture the query behavior and obtain the query start time and query end time;

[0038] The real-time query analysis module is used to obtain a query timeline through the query start time and the query end time, calculate the residence time of each query result on the query timeline and perform normalization calculation through a function to obtain the score of each query result, and determine the query behavior type according to the scores of several query results; the query behavior type includes single-target query and non-single-target query;

[0039] The index management module is used to trace back the query timeline, collect query request strings, calculate the overlap of several strings and generate a temporary index based on the overlap, and then associate the temporary index with the query result according to the query behavior type; the index management module is also used to monitor the usage popularity of the temporary index in real time, determine the usage popularity trend of the temporary index based on the time series algorithm and suspend the temporary index;

[0040] The result output module is used to traverse the query results associated with the temporary index and the regular index according to several query requests and calculate the triggering times of the same query result, and output the final query result according to the triggering times.

[0041] Specifically, the query behavior capture module is responsible for real-time monitoring and recording of user query behavior, including the start and end time of the query. When a user enters a search term and starts searching, the module records the time point of this request; when the user stops browsing search results or starts a new search, this time point is also recorded. These timestamp data provide the basis for subsequent query behavior analysis. Using the time data provided by the query behavior capture module, the real-time query analysis module constructs a query timeline and calculates the user's stay time on each query result. Through specific algorithms, such as normalization, the stay time is quantified, so that a score is assigned to each query result. Based on these scores, the module can determine whether the user's query behavior type is a single-target query or a non-single-target query. For example, if the user stays on a query result much longer than other results, the system may determine that the query is a single-target query. The index management module generates a temporary index based on the user's query behavior by tracing back the query timeline and analyzing the overlap between query request strings. For example, if the keywords of two queries are highly similar in semantics, the system may associate them with the same temporary index. In addition, the module monitors the usage frequency of these temporary indexes, analyzes their usage trends through time series algorithms, and dynamically adjusts the status of the indexes based on the analysis results, such as suspending the use of indexes whose popularity continues to decline.

[0042] When a user initiates a query request, the result output module traverses the temporary index and the regular index to find and output the most relevant query results. During this process, the module will comprehensively consider the relevance of the previous query and the current query, and give priority to displaying the most matching results between the two queries through the temporary index. If the user's query behavior is analyzed as a single-target query, the system will give priority to displaying a single result that is highly relevant to the temporary index; for non-single-target queries, a series of highly relevant results will be displayed. In general, the computer information data rapid query system significantly improves the speed and accuracy of data retrieval through real-time capture and in-depth analysis of user query behavior, as well as intelligent management and dynamic adjustment of indexes. The system can not only effectively cope with the challenges of inaccurate or fuzzy query information, but also reduce the burden on the server and data waste.

[0043] Furthermore, capturing the query behavior and obtaining the query start time and the query end time includes:

[0044] Compare and analyze the query requests in the query behavior to obtain a similarity score, standardize the query request time interval and perform a weighted average with the similarity score to obtain a correlation score;

[0045] According to the correlation scores, query behaviors are correlated with each other to obtain query combinations;

[0046] The time when the earliest query behavior in the query combination occurs is taken as the query start time, and the time when the query result of the latest query behavior in the query combination is obtained is taken as the query end time.

[0047] It should be noted that the purpose is to identify and quantify the similarity between different query requests by comparing and analyzing them. In this process, each query request is first analyzed to extract its key features, including the keywords used in the query, the semantic content of the query, and other relevant attributes. Next, a specific algorithm, such as a text similarity algorithm (such as cosine similarity, Jaccard similarity, etc.), is used to calculate the similarity of each pair of query requests to obtain a similarity score. This similarity score is a quantitative indicator that indicates the degree of similarity between two query requests in terms of content. Its value is usually between 0 and 1, and the closer the value is to 1, the higher the similarity. It also includes standardizing the time intervals between query requests. Standardization refers to converting the time interval into a unified metric to facilitate comparison and analysis between different query requests. For example, the time interval can be converted into a percentage relative to a certain starting time point in a day, or into a proportion relative to the duration of the entire query session. After obtaining the similarity score and the standardized time interval of each pair of query requests, these two types of information are further combined by weighted average to calculate a correlation score. Weighted average can adjust the weight distribution according to the needs of the actual application scenario to reflect the relative importance of similarity and time interval in the calculation of the correlation score. For example, if the system attaches more importance to the similarity of the content, a higher weight can be assigned to the similarity score; conversely, if the temporal proximity of the query request is more important, a higher weight can be assigned to the time interval. In this way, the content similarity and time interval of the query request are effectively integrated into a comprehensive correlation score, which provides a basis for subsequent query behavior analysis and query combination generation. For example, if a user conducts several consecutive searches on "natural language processing" in a short period of time, the similarity scores between these query requests will be high and the time interval will be short. Therefore, the correlation score obtained by weighted average will also be correspondingly high, reflecting the strong correlation between this series of query requests.

[0048] Next, the system compares the correlation scores between the query requests and pairs the query requests with high correlation to form a query combination. The core of this process is to identify those query requests that are closely related in content or time, and regard them as different expressions or attempts of the same information needs, so as to provide users with more accurate and comprehensive search results. Specifically, the system first traverses all query requests and compares the correlation scores between each pair of query requests. Based on a preset threshold, when the correlation scores of two query requests exceed the threshold, the system determines that the two query requests have a high correlation in the user's information retrieval process, so they are associated with each other to form a query combination. For example, if the user searches for "Basics of Machine Learning" and "Introduction to Deep Learning" successively, and the two queries are highly related in content and have a short time interval, then their correlation scores will be high, and the system will associate the two query requests into a combination. This pairwise association method allows the system to flexibly identify and organize the complex or changing information needs expressed by users in a series of queries. Each query combination essentially represents a series of exploration behaviors of the user on a specific topic or question. By aggregating and analyzing these behaviors, the system can understand the user's query intention more comprehensively.

[0049] The earliest query behavior time point in the query combination is identified, and this time point is regarded as the start time of the entire query activity. Subsequently, this step identifies the time when the result of the latest query behavior in the query combination is obtained, and determines it as the end time of the query activity. These two time points together define the entire time range that users spend looking for specific information, providing the system with an important time background for judging users' query intentions and optimizing query results. Suppose a user is conducting market research. They start searching for information about market trends at 9 am, and after continuously modifying and refining query keywords, they obtain the last batch of query results at 10 am. In this case, 9 am is determined as the query start time, and the query result acquisition time at 10 am is the query end time. This time period indicates that the user has been actively looking for and evaluating relevant information during this hour. By accurately identifying the start and end times of queries, the system can better understand the user's query behavior patterns, such as identifying the user's continued interest in a topic or the urgent need for information within a specific time. This helps the system to more accurately adjust the information push strategy in subsequent query responses.

[0050] Furthermore, the stay time of the query result is determined starting from the time point when the user obtains the query result and ending with the time point when the user leaves the query result;

[0051] It should be noted that in the computer information data fast query system, determining the dwell time of query results is a key step, which involves accurately measuring the time users spend on specific query results. The process starts from the time when the user obtains (or clicks into) the query result and ends at the time when the user leaves (or returns to the result list) the query result. The length of this period is regarded as a direct indicator of the user's interest or satisfaction with the query result. Specifically, when a user selects and clicks a query result on the search results page, the system records the exact time when this behavior occurs, marked as the dwell start time. Then, the user browses the content of the result, during which time he may read information, click links, or perform other interactions. When the user finishes viewing the result, whether by clicking the back button to return to the search results page, closing the browser tab, or entering a new search condition to conduct a new query, the system records the time when this behavior occurs again, which is marked as the dwell end time. It is also possible to track the user's cursor and typing behavior to achieve real-time user activity in the search results. Accurate measurement of dwell time provides valuable data for the system, enabling it to evaluate the relevance and quality of query results. Generally speaking, the longer a user stays on a query result, the more the result can meet the user's needs. Otherwise, it may indicate that the result is not relevant enough or the user has not found the required information in the result.

[0052] Furthermore, the calculation method for determining the query behavior type according to the scores of several query results is as follows:

[0053]

[0054] Among them, SD is the standard deviation of the query result score, S i The calculation result of the residence time of the i-th query result after normalization by the function, n represents the total number of query results in the query combination, and the calculation formula of Mean is as follows:

[0055]

[0056] If SD is less than the first preset value and max(S i ) is greater than or equal to the second preset value, it is determined to be a single target query;

[0057] If SD is less than the first preset value and max(S i ) is less than a second preset value, it is determined to be a non-single target query;

[0058] If SD is greater than or equal to the first preset value, it is determined to be a non-single target query.

[0059] Specifically, in the computer information data fast query system, the query behavior type is determined by comprehensively analyzing the user's stay time on each query result. This analysis process covers the normalization of the stay time, calculating the score of the query result, and obtaining the standard deviation (SD) and mean (Mean) of these scores, and finally determining the user's query behavior tendency based on these statistics. First, the system can normalize the stay time of each query result through the log function to obtain the score of each query result. Normalization refers to converting the original stay time into a relative measurement value, usually between 0 and 1, to eliminate the difference in the absolute value of the stay time between different query results, so that the results can be compared. For example, if the stay time of a query result is much longer than that of other results, then the normalized score of this result will be relatively high. Next, the system calculates the standard deviation (SD) and mean (Mean) of the normalized scores of all query results. The standard deviation SD is used to measure the degree of dispersion between the scores of each query result, and the mean Mean reflects the average attention of the user to all query results. Specifically, the calculation formula of Mean is the sum of the scores of all query results divided by the total number of query results n. Based on the obtained SD and the maximum query result score (max()), the system further determines the type of query behavior. If SD is less than the first preset value, determined by extracting the corresponding mean from the labeled sample set pair, and the score of any query result is greater than or equal to the second preset value, determined by extracting the corresponding mean from the labeled sample set pair, this indicates that the user has a clear preference for a specific query result, so the query behavior is determined to be a single-target query. This situation usually occurs when the user has found specific information that meets his needs. On the contrary, if SD is less than the first preset value, but the maximum score of all query results is less than the second preset value, or SD is greater than or equal to the first preset value, this indicates that the user's query behavior is more dispersed and does not show a clear preference for any one query result. In this case, the query behavior is determined to be a non-single-target query, which means that the user may still be in the exploration stage or is equally interested in multiple query results. For example, suppose a user browses a series of query results related to "machine learning", and the dwell time on a specific tutorial page is significantly longer than that on other pages, and the normalized score of this page is much higher than that of any other query results, and meets the judgment criteria of a single-target query. In contrast, if the user evenly distributes the reading time among multiple related paper abstract pages, and no single page stands out significantly, then this will be judged as a non-single-target query.

[0060] Furthermore, the step of calculating the overlap of a plurality of character strings and generating a temporary index according to the overlap includes:

[0061] Evaluate the semantic overlap between query request strings based on semantic similarity algorithm;

[0062] Determine the text overlap between query request strings through cosine similarity;

[0063] Temporary indexes are generated and associated according to the semantic overlap and text overlap.

[0064] In some embodiments, the query request string is first preprocessed, including lexical analysis, removal of stop words, stem extraction, etc., to extract key information from the string. Then, natural language processing (NLP) technology, such as word embedding (Word Embedding) model (e.g., Word2Vec, GloVe, etc.), is used to convert the words in the string into vector form. The position of these vectors in the multidimensional space reflects the semantic attributes of the words, so that the distance between semantically similar words in the vector space is also similar. Subsequently, the algorithm calculates the similarity of the word vectors in the two query request strings, usually using metrics such as cosine similarity to evaluate the overlap between these word vector sets. Ultimately, each pair of query request strings is assigned a semantic overlap score based on their semantic content. The higher this score, the closer the two strings are semantically, and vice versa. For example, suppose there are two query request strings: "Basics of Machine Learning" and "Introduction to Deep Learning". Although the two strings are not similar in literal sense, through the analysis of semantic similarity algorithm, the system can identify the close semantic connection between "machine learning" and "deep learning", so the two query requests will get a higher score in terms of semantic overlap.

[0065] Next, the query request string is segmented to extract the keywords. Subsequently, these keywords are weighted to obtain vectors representing each string. The dimensions of these vectors correspond to the words in the system vocabulary, and the values ​​in the vectors reflect the weights of the corresponding words in the original string (for example, calculated by TF-IDF). After completing this transformation, each query request string is mapped to the position of a point in a multidimensional space. By calculating the cosine similarity between these vectors, the text overlap between two query request strings can be quantified. The value of cosine similarity ranges from -1 to 1, where 1 means that the two vectors are exactly the same in direction, 0 means that the two vectors are orthogonal and have no overlap, and -1 means that the two vectors are completely opposite in direction. In the scenario of query request strings, only positive values ​​are considered because the vectors of text content will not have negative similarities. For example, suppose there are two query requests: "the future of artificial intelligence" and "the trend of artificial intelligence". After word segmentation and TF-IDF processing, the two query requests are converted into vectors. Although the two strings are not exactly the same, they share some keywords, such as "artificial intelligence", which will cause their vectors to have high cosine similarity values ​​in multidimensional space, indicating that they have a high degree of overlap at the text level.

[0066] Taking into account the semantic overlap and text overlap calculated in the previous steps, temporary indexes are generated based on these two overlaps, and these indexes are associated with specific query results. This step is a key step in achieving efficient and accurate information retrieval. It allows the system to flexibly respond to user query needs and provide accurate search results even in the face of inaccurate or ambiguous information. In the process of generating temporary indexes, the system first creates a set of temporary indexes based on the semantic overlap between query request strings. These indexes reflect the similarity and relevance of query requests at the semantic level, so that even if the query requests are not exactly the same on the surface of the text, as long as they are closely related in meaning, the system can associate them through these temporary indexes. For example, for the query requests "latest research in machine learning" and "innovative applications of deep learning", although their text expressions are different, because they are semantically related to research and application in the field of artificial intelligence, the system will generate and associate them with corresponding temporary indexes based on semantic overlap. At the same time, the system also generates another set of temporary indexes based on text overlap, which focus more on capturing the text-level similarity between query requests. This means that even if the semantic association of query requests is not very obvious, as long as they have enough overlap in keywords or expressions, the system can still effectively associate them. This is especially important for processing query requests with very clear keywords but diverse expressions. After generating temporary indexes based on semantics and text overlap, the system associates these indexes. This process not only considers the similarity between query requests, but also the relevance and quality of each query result, ensuring that end users can receive the most accurate and relevant information.

[0067] Furthermore, associating the temporary index with the query result according to the query behavior type includes:

[0068] If the query behavior type is a single-target query, the associated temporary index is associated with the single query result;

[0069] If the query behavior type is a non-single-target query, the associated temporary index is associated with several query results whose similarities reach a preset standard value.

[0070] Specifically, for cases classified as single-target queries, the system recognizes that the user's query intent is focused on finding a specific piece of information or answer. In this case, the user's need for specific information is clear, so the system associates the associated temporary index with the single most relevant query result. This means that the system will analyze all possible query results under the temporary index, and select a result that best meets the user's query needs based on indicators such as relevance score, user feedback, and click-through rate, and directly associate this result with the temporary index. For example, if a user queries "Who is the pioneer of quantum computing", after the system recognizes this query as a single-target query, it will focus on finding and associating a high-quality content page that clearly answers this question.

[0071] In contrast, for non-single-target queries, the system believes that the user's query intent may be more extensive or exploratory, and the user may be looking for multi-faceted information about a topic. In this case, the system associates the associated temporary index with multiple query results whose similarity reaches a preset standard value. The preset standard value is determined by artificial historical experience. This approach allows the system to present a diverse set of query results, covering different aspects that the user may be interested in, and enhancing the depth and breadth of information exploration. For example, for the query "Introduction to Quantum Computing", the system may associate multiple query results containing basic tutorials, important concepts, historical developments, etc., to meet the user's needs for a comprehensive understanding of the topic.

[0072] Furthermore, the real-time monitoring of the usage popularity of the temporary index includes collecting the access frequency and query times of the temporary index and calculating the popularity score by weighted average.

[0073] In some embodiments, the system continuously tracks and records the frequency with which each temporary index is accessed and queried, including the number of times a user clicks on the query results under a specific index and the total number of searches for the index. This data collection process provides the system with real-time, quantified user behavior information, reflecting the actual usage of different indexes in user query activities. In order to ensure that the heat score accurately reflects the current usage of the temporary index, the system uses a weighted average method to calculate this score. This method takes into account the relative importance of index accesses in different time periods, allowing the system to give higher weights to recent query activities. For example, the system may give greater weight to query activities within the past 24 hours, and less weight to earlier activities. In this way, the heat score calculated by weighted average not only reflects the overall usage of the index, but also highlights the trend of recent user behavior.

[0074] Furthermore, judging the usage popularity trend of the temporary index based on the time series algorithm and pausing the temporary index includes:

[0075] In some embodiments, the system uses ARIMA to perform trend analysis on the heat score. If the heat score is lower than a preset threshold and the heat trend is negative, the use of the associated temporary index is suspended; if the heat score is lower than a preset threshold and the heat trend is positive, the use of the associated temporary index is retained; if the heat score is higher than the preset threshold, the use of the associated temporary index is retained. The ARIMA model is a powerful tool widely used in time series data analysis, which can simulate and predict the trend and pattern of data changing over time. In this system, the ARIMA model is used to analyze the changing trend of the access frequency and query number of temporary indexes over time, that is, the time series data of the use heat. By modeling the past use heat data, ARIMA can predict the use heat trend of temporary indexes in the future. If the ARIMA model analysis shows that the use heat trend of a temporary index is negative, that is, the use heat is gradually decreasing, and the current heat score is already lower than the preset threshold set by the system, the system will decide to suspend the use of the temporary index. This is based on the recognition that the index may no longer reflect the interests or needs of the current user, and continuing to maintain its active state will be an unnecessary resource consumption. For those temporary indexes whose popularity scores are lower than the preset threshold but whose usage popularity trends are positive, the system chooses to keep them in use. This situation indicates that although the current usage popularity of the index is not high, its popularity is rising and it may soon become the focus of users. For those temporary indexes whose popularity scores are higher than the preset threshold, the system will keep them in use regardless of their popularity trends. This reflects that these indexes are currently highly relevant and popular and are hot content in user search activities. In this way, the system can ensure that the dynamic management of temporary indexes is closely linked to the actual needs of users, which can not only respond to emerging query trends in a timely manner, but also effectively reduce the resource investment in content that is no longer popular or outdated. This data-driven strategy not only improves the overall efficiency of the search service and user satisfaction, but also optimizes the use of system resources and ensures the efficient operation of the search platform.

[0076] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A computer information data rapid query system, characterized in that: It includes a query behavior capture module, a real-time query analysis module, an index management module and a result output module, wherein the query behavior capture module, the real-time query analysis module, the index management module and the result output module are connected in sequence; The query behavior capturing module is used to capture the query behavior and obtain the query start time and query end time; The real-time query analysis module is used to obtain a query timeline through the query start time and the query end time, calculate the residence time of each query result on the query timeline and perform normalization calculation through a function to obtain a score for each query result, and determine the query behavior type according to the scores of several query results; The query behavior types include single-target query and non-single-target query; The index management module is used to trace back the query timeline, collect query request strings, calculate the overlap of several strings and generate a temporary index based on the overlap, and then associate the temporary index with the query result according to the query behavior type; the index management module is also used to monitor the usage popularity of the temporary index in real time, determine the usage popularity trend of the temporary index based on the time series algorithm and suspend the temporary index; The result output module is used to traverse the query results associated with the temporary index and the regular index according to several query requests and calculate the triggering times of the same query result, and output the final query result according to the triggering times; The real-time monitoring of the usage popularity of the temporary index includes collecting the access frequency and query times of the temporary index and calculating the popularity score by weighted average; The method of judging the usage popularity trend of the temporary index based on the time series algorithm and pausing the temporary index includes: Use ARIMA to perform trend analysis on the heat score. If the heat score is lower than the preset threshold and the heat trend is negative, the use of the associated temporary index is suspended; if the heat score is lower than the preset threshold and the heat trend is positive, the use of the associated temporary index is retained; if the heat score is higher than the preset threshold, the use of the associated temporary index is retained.

2. A computer information data rapid query system according to claim 1, characterized in that: Capturing the query behavior and obtaining the query start time and query end time includes: Compare and analyze the query requests in the query behavior to obtain a similarity score, standardize the query request time interval and perform a weighted average with the similarity score to obtain a correlation score; According to the correlation scores, query behaviors are correlated with each other to obtain query combinations; The time when the earliest query behavior in the query combination occurs is taken as the query start time, and the time when the query result of the latest query behavior in the query combination is obtained is taken as the query end time.

3. A computer information data rapid query system according to claim 1, characterized in that: The residence time of the query result is determined from the time point when the user obtains the query result and ends when the user leaves the query result.

4. A computer information data rapid query system according to claim 1, characterized in that: The calculation method for determining the query behavior type based on the scores of several query results is as follows: ， Among them, SD is the standard deviation of the query result score, The calculation result of the residence time of the i-th query result after normalization by the function, n represents the total number of query results in the query combination, and the calculation formula of Mean is as follows: ， If SD is less than the first preset value and max ( ) is greater than or equal to the second preset value, it is determined to be a single target query; If SD is less than the first preset value and max ( ) is less than the second preset value, it is determined to be a non-single target query; If SD is greater than or equal to the first preset value, it is determined to be a non-single target query.

5. A computer information data rapid query system according to claim 1, characterized in that: The step of calculating the overlap of a plurality of character strings and generating a temporary index according to the overlap comprises: Evaluate the semantic overlap between query request strings based on semantic similarity algorithm; Determine the text overlap between query request strings through cosine similarity; Temporary indexes are generated and associated according to the semantic overlap and text overlap.

6. A computer information data rapid query system according to claim 5, characterized in that: The associating the temporary index with the query result according to the query behavior type includes: If the query behavior type is a single-target query, the associated temporary index is associated with the single query result; If the query behavior type is a non-single-target query, the associated temporary index is associated with several query results whose similarities reach a preset standard value.

Images