An information management method based on intelligent classification and efficient retrieval

Abstract

The present invention discloses an information management method based on intelligent classification and efficient retrieval, which relates to the field of information management technology and solves the technical problem that the lack of in-depth understanding of information semantics and content leads to inaccurate retrieval results. The present invention extracts keywords by combining a machine learning algorithm for preliminary classification, and then performs secondary classification based on semantics, and determines standard classification results through an integration and screening mechanism, thereby improving the accuracy of information classification and reflecting the core content of the information. A statistical model is used for word segmentation to determine a retrieval header, and retrieval headers without retrieval content are eliminated to improve retrieval efficiency. Content to be analyzed is screened by calculating the proportion of quantity, and retrieval content is screened according to scientific quantitative standards, so that information highly related to user needs can be quickly located. The correlation strength between the retrieval header and the retrieval information is accurately analyzed by assigning values to the grammatical positions of pre-selected retrieval headers, calculating cosine similarity, and the like, and distinguishing between strongly correlated and weakly correlated retrieval headers.

Description

Technical Field

[0001] The present invention relates to the technical field of information management, and specifically provides an information management method based on intelligent classification and efficient retrieval. Background Art

[0002] With the rapid development of information technology, the quantity of information has grown explosively. Whether in academic research, business operations, or daily life, people are facing the challenge of quickly and accurately obtaining the required content from a vast amount of information.

[0003] According to the patent application with the application number CN202410441933.8, an information management method based on intelligent classification and efficient retrieval is disclosed. By collecting the data that needs to be managed and processed and performing preprocessing for subsequent classification and retrieval operations, the burden of manual operations is reduced; feature extraction and text representation are performed on the data; a classification model is constructed to perform intelligent classification and organization on the data for subsequent efficient retrieval and query. Intelligent classification and efficient retrieval can help users quickly and accurately find the required information and improve work efficiency; according to the user's search conditions and query requirements, the data is quickly located and retrieved through the classification model; according to the user's interests and preferences, personalized data recommendations are provided so that users can find the required data. By analyzing the user's behavior and interests, personalized recommendations and search results are provided to meet the specific needs of users; through interaction and feedback with users, the results of classification and retrieval are continuously optimized to improve the accuracy of retrieval.

[0004] Traditional information retrieval and management methods often rely on simple keyword matching, lacking a deep understanding of information semantics and content, resulting in inaccurate retrieval results and imprecise classification, and being unable to meet the growing demand of users for efficient and accurate information acquisition. When facing complex retrieval requirements, existing information management systems are difficult to quickly locate and provide information that meets user expectations, causing waste of time and resources. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides an information management method based on intelligent classification and efficient retrieval, which solves the problem of inaccurate retrieval results caused by the lack of a deep understanding of information semantics and content.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: An information management method based on intelligent classification and efficient retrieval, which specifically includes the following steps:

[0007] Obtain the retrieval information, extract keywords, first perform a preliminary classification, then perform a secondary classification based on semantics, integrate the classification results. If the classifications are the same, mark it as the standard classification result; if different, screen the keywords to filter out the standard classification result;

[0008] Determine the retrieval direction based on the standard classification results. After word segmentation, generate word segmentation results using a statistical model. Use its index as the retrieval header. After retrieval, eliminate the retrieval headers without content to obtain preliminary retrieval headers;

[0009] Obtain the preliminary retrieval headers and their retrieval content, extract the common semantics as the content to be analyzed, calculate the proportion of the types of retrieval headers and the number of preliminary retrieval headers in the content to be analyzed, compare with the preset proportion, and retain the content greater than the preset proportion to obtain the retained content;

[0010] Extract the preliminary retrieval headers from the retained content, assign values according to the grammatical position, calculate the cosine similarity with other preliminary retrieval headers, and mark those greater than or equal to the preset similarity as associated preliminary retrieval headers;

[0011] Obtain the number of all preliminary retrieval headers and their associated preliminary retrieval headers, calculate the index association strength value corresponding to the preliminary retrieval headers, and compare with the preset comparison value. Mark those greater than the preset comparison value as strongly associated retrieval headers, and mark those less than as weakly associated retrieval headers;

[0012] Obtain the strongly and weakly associated retrieval headers and the retained content, analyze the weak associations, screen out the content with a proportion less than the first preset value, analyze the strong associations, eliminate the part where the number of weak association headers exceeds the second preset value, summarize the screened content, sort it in descending order according to the number of word segmentation results, and generate and display the retrieval results.

[0013] As a further solution of the present invention, the specific method for screening out the standard classification results is as follows:

[0014] Obtain the retrieval information corresponding to the retrieval user, extract the keywords, and at the same time perform preliminary classification based on the keywords to obtain the preliminary classification results. Then, obtain the semantics of the retrieval information and perform secondary classification based on the semantics of the retrieval information to obtain the secondary classification results;

[0015] Integrate and judge the preliminary classification results and the secondary classification results, obtain the common classification results and mark them as the standard classification results. Conversely, if there are no identical classification results in the two, perform secondary analysis and processing;

[0016] Obtain all the keywords and their corresponding semantics, match and screen the keywords and semantics to obtain the preliminary keywords, and at the same time screen the preliminary classification results and the secondary classification results with the preliminary keywords as the standard to obtain the standard classification results.

[0017] As a further solution of the present invention, the specific method for obtaining the preliminary retrieval headers is as follows:

[0018] Obtain the standard classification results corresponding to the retrieval information, and at the same time determine the retrieval direction. Then, perform word segmentation on the retrieval information according to the retrieval direction to obtain the word segmentation results. Denote the obtained index as the retrieval header and the label as i, and i = 1, 2, …, j, where j represents the number of retrieval headers. Then, perform retrievals respectively with the retrieval headers as the criteria to obtain the corresponding retrieval contents, denoted as Ki, and K = A, B, …, where A and B represent the corresponding retrieval content information. And based on the retrieval contents, eliminate the retrieval header i to obtain the preselected retrieval headers a, and a = 1, 2, …, b.

[0019] As a further solution of the present invention, the specific method for retaining the content with a proportion greater than the preset proportion to obtain the retained content is:

[0020] Obtain the preselected retrieval headers a and the corresponding retrieval contents Ka. At the same time, based on the semantics of the retrieval contents, obtain the common semantic content existing among the retrieval contents Ka, denoted as the content to be analyzed n, and n = 1, 2, …, m, where m represents the number of contents to be analyzed;

[0021] Obtain the types of retrieval headers in the content to be analyzed n as the retrieval headers to be analyzed. At the same time, calculate the proportion of the number of retrieval headers to be analyzed and the preselected retrieval headers, and compare the proportion with the preset proportion. Then, retain the corresponding content to be analyzed to obtain the retained content. On the contrary, if the proportion is less than the preset proportion, then eliminate the corresponding content to be analyzed.

[0022] As a further solution of the present invention, the specific screening method for associating the preselected retrieval headers is:

[0023] Obtain the grammatical positions corresponding to the preselected retrieval headers a, perform assignment processing based on different grammatical positions and denote it as Ca, calculate the similarity between the preselected retrieval headers a, and compare it with the preset similarity. The specific value of the preset similarity is set by the operator, and screen the preselected retrieval headers with similarity greater than the preset similarity and mark them as associated preselected retrieval headers.

[0024] As a further solution of the present invention, the specific method for obtaining the strongly associated retrieval headers and weakly associated retrieval headers is:

[0025] Obtain the number La of all preselected retrieval headers a and the corresponding associated preselected retrieval headers. Then, according to the formula Calculate the index association strength value Sa corresponding to the preselected retrieval header a, where q is the index adjustment factor, and compare the obtained index association strength value Sa with the preset comparison value Sy;

[0026] Mark the preselected retrieval header a with the index association strength value Sa greater than the preset comparison value Sy as the strongly associated retrieval header, and mark the preselected retrieval header a with the index association strength Sa less than the preset comparison value Sy as the weakly associated retrieval header.

[0027] As a further solution of the present invention, the specific manner of generating and displaying the retrieval results is as follows:

[0028] Obtain the quantity corresponding to the weak association retrieval head in the retained content, calculate the proportion of the weak association quantity corresponding to the weak association retrieval head at the same time, compare it with the first preset value, and select the retained content corresponding to the proportion of the weak association quantity less than the first preset value, which is denoted as the screened content;

[0029] Obtain the quantity of the corresponding weak association retrieval head in the retained content, eliminate the retained content with the quantity of the corresponding weak association retrieval head greater than the second preset value, denote the remaining retained content as the screened content, and display and analyze the screened content at the same time.

[0030] As a further solution of the present invention, the specific manner of displaying and analyzing the screened content is as follows:

[0031] Obtain all the screened content, obtain the quantity of the corresponding word segmentation results in the screened content at the same time, sort them from large to small according to the quantity of the word segmentation results, generate the retrieval results, and then display the obtained retrieval results to the corresponding retrieval user.

[0032] The present invention provides an information management method based on intelligent classification and efficient retrieval. Compared with the prior art, it has the following beneficial effects:

[0033] The present invention combines machine learning algorithms to extract keywords for preliminary classification, and then performs secondary classification based on semantics. The standard classification results are determined through an integration and screening mechanism, improving the accuracy of information classification and being able to more accurately reflect the core content of the information. A statistical model is used to perform word segmentation to determine the retrieval head, and the retrieval heads without retrieval content are eliminated, reducing invalid retrievals and improving retrieval efficiency. By calculating the proportion of quantities to screen the content to be analyzed and screening the retrieval content according to scientific quantification criteria, information highly relevant to the user's needs can be quickly located. By assigning grammatical positions to the preselected retrieval heads, calculating cosine similarity, etc., the association strength between the retrieval head and the retrieval information is accurately analyzed, distinguishing strong association and weak association retrieval heads, providing more accurate and demand-compliant retrieval results for users, and improving the efficiency and quality of users' information acquisition. Description of the Drawings

[0034] Figure 1 It is a flowchart of the steps of the present invention. Detailed Embodiment

[0035] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0036] Embodiment 1. Please refer to Figure 1 , this application provides an information management method based on intelligent classification and efficient retrieval. The method specifically includes the following steps:

[0037] Step1. Obtain the corresponding retrieval information, and at the same time obtain the keywords in the retrieval information, and perform a preliminary classification on the retrieval information based on the keywords to generate a preliminary classification result. The specific classification method is as follows:

[0038] First, obtain the retrieval information corresponding to the retrieval user, and extract the keywords in the retrieval information. Here, the extraction is based on the corresponding classification system, such as using machine learning algorithms for extraction, such as decision trees, support vector machines, neural networks, to learn the labeled information data, extract features and patterns, build a classification model, and at the same time perform a preliminary classification on the retrieval information according to the obtained keywords to obtain a preliminary classification result. Here, the obtained preliminary classification result may include multiple types. Then, obtain the semantics of the retrieval information, and perform a secondary classification based on the semantics of the retrieval information to obtain a secondary classification result;

[0039] Integrate and judge the obtained preliminary classification result and secondary classification result. If there is the same classification result in the two, obtain the common classification result and mark it as the standard classification result. Otherwise, if there is no same classification result in the two, perform secondary analysis processing;

[0040] Obtain all the keywords and their corresponding semantics, match and screen the keywords and semantics to obtain preselected keywords, and at the same time screen the preliminary classification result and secondary classification result based on the preselected keywords to obtain the standard classification result. Here, it is default that there is only one standard classification result.

[0041] Step 2. Obtain the standard classification results corresponding to the retrieval information, and at the same time determine the retrieval direction. Then, perform word segmentation on the retrieval information according to the retrieval direction to obtain the word segmentation result. Specifically, use a statistical model to model the character sequence in the retrieval information, calculate the probability of each character belonging to different words, so as to determine the word segmentation result. At the same time, record the obtained index as the retrieval header, and the label is denoted as i, and i = 1, 2, …, j, where j represents the number of retrieval headers. Then, perform retrieval respectively with the retrieval headers as the criteria to obtain the corresponding retrieval content, which is denoted as Ki, and K = A, B, …, where A and B represent the corresponding retrieval content information. And based on the retrieval content, eliminate the retrieval header i to obtain the preselected retrieval header a, and a = 1, 2, …, b. The specific elimination method is: eliminate the retrieval headers without retrieval content and retain the retrieval headers with retrieval content;

[0042] For example, the retrieval information is "Application of Artificial Intelligence in Medical Image Diagnosis". According to the standard classification system of this database, the corresponding standard classification results may be determined as "Computer Science - Artificial Intelligence" and "Medicine - Medical Image Diagnosis", and the retrieval direction is determined as the relevant research on the application of artificial intelligence in the medical field;

[0043] Use a statistical model (such as a conditional random field model) to model the retrieval information "Application of Artificial Intelligence in Medical Image Diagnosis". By calculating the probability of each character belonging to different words, the word segmentation result is obtained as "artificial intelligence", "in", "medical", "image", "diagnosis", "in", "the", "application";

[0044] Furthermore, obtain the corresponding retrieval headers. Retrieval header 1 (i = 1): "artificial intelligence", retrieval header 2 (i = 2): "medical", retrieval header 3 (i = 3): "image", retrieval header 4 (i = 4): "diagnosis", retrieval header 5 (i = 5): "application". At this time, j = 5 (the number of retrieval headers). Check the retrieval content corresponding to these retrieval headers. In this database, when using "in", "in", "the" as retrieval headers for retrieval, no relevant academic literature is found (that is, there is no retrieval content), while "artificial intelligence", "medical", "image", "diagnosis", "application" all have corresponding retrieval content. According to the elimination method, eliminate the retrieval headers "in", "in", "the" without retrieval content.

[0045] Step 3. Obtain the preselected search head a and the corresponding search content Ka. At the same time, based on the semantics of the search content, obtain the common semantic content existing among the search contents Ka, denoted as the content to be analyzed, and label it as n, where n = 1, 2, …, m, and m represents the number of contents to be analyzed. Then, obtain the types of search heads in the content to be analyzed n, denoted as the search head to be analyzed. At the same time, calculate the proportion of the number of the search head to be analyzed to the preselected search head. Then, sort the content to be analyzed n from largest to smallest according to the proportion;

[0046] Filter the content to be analyzed n according to the proportion, and the specific filtering method is as follows: Compare the proportion with the preset proportion. The specific value of the preset proportion is set by the operator, specifically set according to the number of all search heads, and takes a value of one-fourth in this application. If the proportion is greater than the preset proportion, and this includes the case of being greater than the preset proportion, then retain the corresponding content to be analyzed. On the contrary, if the proportion is less than the preset proportion, then eliminate the corresponding content to be analyzed;

[0047] Step 4. Obtain all the retained content to be analyzed, denoted as the retained content. Then, obtain the corresponding preselected search head a in the retained content. At the same time, analyze the correlation strength between the preselected search head a and the search information. The specific analysis method is as follows:

[0048] Obtain the grammatical position corresponding to the preselected search head a. Here, the grammatical position mainly represents the position borne, such as the subject, predicate, object, or adverbial, etc. At the same time, perform different assignment processing based on different grammatical positions, denoted as Ca. Here, the assignment is carried out separately according to the subject, predicate, object, and adverbial. For example, the subject is assigned 3, the predicate is assigned 2, and the object and adverbial are assigned 1. Then, obtain the associated preselected search head of the preselected search head a. Here, when obtaining the associated preselected search head, calculate the corresponding similarity. Specifically, take one of the preselected search heads as the analysis object, then perform quantization processing on the analysis object to obtain the corresponding vector. By analogy, perform quantization processing on all preselected search heads and obtain the corresponding vectors. Then, calculate the cosine similarity between the analysis object and all preselected search head vectors, and compare the obtained similarity with the preset similarity. The specific value of the preset similarity is set by the operator;

[0049] If the similarity is greater than the preset similarity, and this includes the case of being equal to the preset similarity, then mark the corresponding preselected search head as the associated preselected search head. On the contrary, if the similarity is less than the preset similarity, then no processing is performed;

[0050] Obtain all the preselected search heads a, and at the same time, obtain the number of the corresponding associated preselected search heads, denoted as La. Then, according to the formula Calculate and obtain the index association strength value Sa corresponding to the pre-selected search head a, where q is the index adjustment factor, and the specific value is adjusted according to the actual situation to highlight or weaken the influence of the quantitative value of the association feature on the association strength, and compare the obtained index association strength value Sa with the preset comparison value Sy;

[0051] If the index association strength value Sa is greater than the preset comparison value Sy, and here includes the case where it is equal to the preset comparison value, the corresponding pre-selected search head a is marked as a strongly associated search head, otherwise if the index association strength Sa is less than the preset comparison value Sy, the corresponding pre-selected search head a is marked as a weakly associated search head;

[0052] Step 5, then obtain all the strongly associated search headers and weakly associated search headers, and obtain the corresponding retained content at the same time, and the retained content here is the same as the retained content analyzed in Step 3, and the corresponding retained content of the two is analyzed respectively;

[0053] Analyze the reserved content corresponding to the weakly associated search header, obtain the number of weakly associated search headers in the reserved content, and calculate the proportion of the weakly associated number of the weakly associated search headers, and compare the obtained proportion of the weakly associated number with a first preset value, and the specific value of the first preset value is set by the operator, select and filter the reserved content corresponding to the weakly associated number proportion less than the first preset value, and record it as the filtered content, and the case where it is equal to the first preset value is not included here;

[0054] Analyze the reserved content corresponding to the strongly associated search headers, obtain the number of corresponding weakly associated search headers in the reserved content, and remove the reserved content whose number of corresponding weakly associated search headers is greater than a second preset value, and record the remaining reserved content as the screened content;

[0055] Then, all the filtered contents are obtained, and the corresponding word segmentation result quantity in the filtered contents is obtained, and the word segmentation result quantity is sorted from large to small, and the search results are generated, and then the obtained search results are displayed to the corresponding search users.

[0056] Some of the data in the above formulas are calculated by taking their numerical values, and are not substituted into parameter units for calculation. At the same time, the contents not described in detail in this specification belong to the existing technologies known to those skilled in the art.

[0057] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An information management method based on intelligent classification and efficient retrieval, characterized in that The method specifically includes the following steps: Obtain retrieval information, extract keywords, first conduct a preliminary classification, then perform a secondary classification based on semantics, integrate the classification results. If the classifications are the same, mark them as the standard classification results; if different, screen the keywords to filter out the standard classification results in this way; Determine the retrieval direction according to the standard classification results, perform word segmentation on the retrieval information, and generate word segmentation results using a statistical model. Use its index as the retrieval header, and after retrieval, eliminate the retrieval headers without content to obtain the preselected retrieval headers; Obtain the preselected retrieval headers and their retrieval contents, extract the common semantics as the content to be analyzed, calculate the proportion of the types of retrieval headers in the content to be analyzed and the number of preselected retrieval headers, compare it with the preset proportion, and retain the content with a proportion greater than the preset proportion to obtain the retained content; Extract the preselected retrieval headers from the retained content, assign values according to the grammatical positions, calculate the cosine similarity with other preselected retrieval headers, and mark those with a cosine similarity greater than or equal to the preset similarity as associated preselected retrieval headers; Obtain all preselected search headers a and their associated preselected search header quantities La, and according to the formula calculate the exponential association strength value Sa corresponding to the preselected search header a, compare Sa with the preset comparison value Sy, mark the preselected search header a with Sa greater than Sy as a strongly associated search header, and mark the one with Sa less than Sy as a weakly associated search header; Obtain the number of weakly associated retrieval headers in the retained content, calculate the proportion of the corresponding weakly associated quantities of the weakly associated retrieval headers, and compare it with the first preset value. Take the retained content with a proportion less than the first preset value as the screened content. Obtain the number of weakly associated retrieval headers in the retained content, eliminate the retained content with a quantity greater than the second preset value, and record the remaining part as the screened content, and conduct display analysis; Collect all the screened content, obtain the number of word segmentation results among them, sort them from largest to smallest in terms of quantity to generate the retrieval results, and present them to the retrieval user.

2. The information management method based on intelligent classification and efficient retrieval according to claim 1, wherein, The specific method for filtering out the standard classification results is as follows: Obtain the retrieval information corresponding to the retrieval user, extract the keywords, and at the same time conduct a preliminary classification based on the keywords to obtain the preliminary classification results. Then, obtain the semantics of the retrieval information and conduct a secondary classification based on the semantics of the retrieval information to obtain the secondary classification results; Integrate and judge the preliminary classification results and the secondary classification results, obtain the common classification results and mark them as the standard classification results. Conversely, if there are no identical classification results in the two, perform secondary analysis and processing; Obtain all the keywords and their corresponding semantics, match and screen the keywords and semantics to obtain the preselected keywords, and at the same time use the preselected keywords as the standard to screen the preliminary classification results and the secondary classification results to obtain the standard classification results.

3. An information management method based on intelligent classification and efficient retrieval according to claim 1, characterized in that, The specific method for obtaining the preselected retrieval headers is as follows: Obtain the standard classification results corresponding to the retrieval information, and at the same time determine the retrieval direction. Then, perform word segmentation on the retrieval information according to the retrieval direction to obtain the word segmentation results. Denote the obtained index as the retrieval header, and denote the label as i, and i = 1, 2,..., j, where j represents the number of retrieval headers. Then, conduct retrievals respectively with the retrieval headers as the standard to obtain the corresponding retrieval contents, and denote them as Ki, and K = A, B,..., where A, B represent the corresponding retrieval content information, and eliminate the retrieval header i based on the retrieval content to obtain the preselected retrieval header a, and a = 1, 2,..., b.

4. An information management method based on intelligent classification and efficient retrieval according to claim 1, characterized in that, The specific method for retaining the content with a proportion greater than the preset proportion to obtain the retained content is as follows: Obtain the preselected retrieval head a and the corresponding retrieval content Ka. At the same time, based on the semantics of the retrieval content, obtain the common semantic content existing among the retrieval contents Ka, denoted as the content to be analyzed n, and n = 1, 2, …, m, where m represents the number of contents to be analyzed; Obtain the types of retrieval heads in the content to be analyzed n, denoted as the retrieval heads to be analyzed. At the same time, calculate the quantity ratio of the retrieval heads to be analyzed and the preselected retrieval heads, and compare the quantity ratio with the preset ratio. Then retain the corresponding content to be analyzed to obtain the retained content. On the contrary, if the quantity ratio is less than the preset ratio, the corresponding content to be analyzed will be eliminated.

5. An information management method based on intelligent classification and efficient retrieval according to claim 1, characterized in that, The specific screening method for the associated preselected retrieval heads is as follows: Obtain the grammatical positions corresponding to the preselected retrieval head a, perform assignment processing based on different grammatical positions, denoted as Ca, calculate the similarity between the preselected retrieval heads a, and compare it with the preset similarity. The specific value of the preset similarity is set by the operator. Screen the preselected retrieval heads with similarity greater than the preset similarity and mark them as associated preselected retrieval heads.

Images