Intelligent similar case recommendation method and system

Abstract

The present invention relates to an intelligent similar case recommendation method and system, which relates to the field of information processing technology, and includes: collecting detection label categories; when no text appears in a preset text box, finding a benchmark case from a preset case library in response to the detection label category; obtaining the benchmark label quantity and the total number of benchmark labels in response to the benchmark case and the detection label category; obtaining the detection label quantity in response to the detection label category; obtaining the case overlap in response to the benchmark label quantity, the total number of benchmark labels, and the detection label quantity; obtaining a recommended case ranking in response to the case overlap and the benchmark case, and outputting the recommended case ranking to a preset display area. The present application has the effect of improving the accuracy of similar case recommendations.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to an intelligent similar case recommendation method and system. Background Art

[0002] Information processing is the process of collecting, storing, converting, analyzing and disseminating information.

[0003] In the process of case information processing, the text content in the case is analyzed and matched according to pre-set label categories to obtain cases marked with multiple labels. When users need to query cases, they usually need to manually enter keywords to query the text of the cases stored in the system, so as to obtain the cases required by the users for reference.

[0004] When users search for cases, there are many cases with the same keywords, and the text content corresponding to each case still has deviations, resulting in low accuracy in recommending similar cases. Summary of the Invention

[0005] In order to improve the accuracy of similar case recommendations, the present invention provides an intelligent similar case recommendation method and system.

[0006] In a first aspect, the present invention provides an intelligent similar case recommendation method, which adopts the following technical solution:

[0007] An intelligent similar case recommendation method, comprising:

[0008] S100: Collect and detect label categories;

[0009] S101: When no text appears in the preset text box, searching for a reference case from a preset case library in response to the detection tag category;

[0010] S102: Obtaining a number of benchmark tags and a total number of benchmark tags in response to the benchmark case and the detection tag category;

[0011] S103: Obtaining the number of detection tags in response to the detection tag category;

[0012] S104: Obtaining case coincidence in response to the number of reference tags, the total number of reference tags, and the number of detection tags;

[0013] S105: Obtaining a recommended case ranking in response to the case overlap degree and the benchmark case, and outputting the recommended case ranking to a preset display area.

[0014] By adopting the above technical solution, the benchmark cases and detection label categories are analyzed to obtain the number of benchmark labels, the total number of benchmark labels and the number of detection labels, and then the case overlap is obtained by the number of benchmark labels, the total number of benchmark labels and the number of detection labels. The benchmark cases are sorted according to the case overlap to obtain the recommended case ranking and output to the display area, so that similar cases can be recommended through the labels in the system to improve the accuracy of similar case recommendations.

[0015] Optionally, also include:

[0016] S200: When text appears in a preset text box, collect and detect the query text;

[0017] S201: Responding to the detection query text to obtain detection query semantics;

[0018] S202: Responding to the detection query semantics to obtain a detection unprocessed text and a reference processed text;

[0019] S203: In response to the detecting of the unprocessed text and the preset reference label category to obtain an unprocessed weight;

[0020] S204: Obtaining non-overlapping labels in response to the unprocessed weights and the detected label categories;

[0021] S205: Obtain the recommended case ranking by using the non-overlapping labels, the benchmark processing text, and the detection label category.

[0022] Optionally, the method for obtaining the recommended case ranking includes:

[0023] S300: Obtaining a marked processed text and other processed texts by using the non-overlapping labels and the reference processed text;

[0024] S301: Responding to the tag processing text and the detection tag category to obtain a tag processing weight;

[0025] S302: defining the other processed text corresponding to the marked processed weight consistent with the unprocessed weight as the detected processed text;

[0026] S303: processing the text and the non-overlapping labels in response to the detection to obtain a non-overlapping weight;

[0027] S304: Responding to the benchmark case and the detection label category to obtain a benchmark selection case;

[0028] S305: Selecting cases in response to the detection label categories, the non-overlapping weights, and the benchmark to obtain the recommended case ranking.

[0029] Optionally, the method for obtaining the recommended case ranking includes:

[0030] S400: Responding to the detection tag category to obtain a detection selection order;

[0031] S401: Obtaining an adjustment coefficient and a number of detection selections in response to the detection selection order;

[0032] S402: Obtaining a weight ratio by using the adjustment coefficient, the number of detection selections, and the non-overlapping weight;

[0033] S403: Selecting cases in response to the benchmark to obtain a benchmark weight ratio;

[0034] S404: Obtaining the recommended case ranking in response to the weight ratio and the benchmark weight ratio.

[0035] Optionally, after obtaining the recommended case ranking, the method further includes:

[0036] S500: Responding to the detection query text and the reference tag category to obtain a detection query tag;

[0037] S501: Retrieving a search tag, a subject tag, and other query tags in response to the detection query tag;

[0038] S502: Responding to the search tag and the benchmark case to obtain similar cases;

[0039] S503: Responding to the similar cases and the subject tags to obtain labeled similar cases;

[0040] S504: Obtain the benchmark selected case in response to the marked similar case, the search tag, and the subject tag.

[0041] Optionally, the method for obtaining the benchmark selection case includes:

[0042] S600: Retrieving texts that are not corresponding to the search tag and the subject tag from the similarly marked cases as similarly marked texts;

[0043] S601: Responding to the marked similar text to obtain a marked similar tag;

[0044] S602: defining the text of the similarly marked case whose similarity tag is consistent with the other query tag as the benchmark selected case.

[0045] Optionally, the method for obtaining the detected unprocessed text includes:

[0046] S700: Responding to the detection query semantics to obtain unprocessed text;

[0047] S701: When the number of the unprocessed texts is not 1, obtaining a detection query vocabulary in response to the detection query semantics and the reference tag category;

[0048] S702: Obtain the detected unprocessed text in response to the detected query term, the unprocessed text, and the reference tag category.

[0049] Optionally, the method for obtaining the detected unprocessed text includes:

[0050] S800: generating the detection integrated vocabulary in response to the detection query vocabulary;

[0051] S801: integrating vocabulary through the detection to obtain an integrated relevance coefficient;

[0052] S802: Obtaining an unprocessed coefficient in response to the unprocessed text and the reference label category;

[0053] S803: When the unprocessed coefficient is greater than the integrated correlation coefficient, defining the unprocessed text corresponding to the unprocessed coefficient greater than the integrated correlation coefficient as a marked text;

[0054] S804: When the number of the marked texts is not 1, the marked text with the largest unprocessed coefficient is defined as the detected unprocessed text.

[0055] Optionally, after collecting the detection query text, the method further includes:

[0056] S900: Collect detection query images and detection query voice;

[0057] S901: Responding to the detection query image and the preset detection features to obtain a marked detection image;

[0058] S902: generating a marked conversion text in response to the marked detection picture;

[0059] S903: generating a detection conversion text in response to the detection query voice;

[0060] S904: Responding to the detected transformed text and the marked transformed text to update the detected query text.

[0061] In a second aspect, the present application provides an intelligent similar case recommendation system, which adopts the following technical solutions:

[0062] An intelligent similar case recommendation system, comprising:

[0063] Acquisition module, used to obtain detection label categories;

[0064] A memory for storing a program for an intelligent similar case recommendation method;

[0065] The processor is configured to load, execute, and implement the program stored in the memory.

[0066] In summary, this application includes at least one of the following beneficial technical effects:

[0067] 1. By analyzing the text in the text box, the benchmark case, and the detection label category, the recommended case ranking is obtained and output to the display area. This allows the system to recommend similar cases based on the labels in the system, thereby improving the accuracy of similar case recommendations;

[0068] 2. By analyzing the user's input detection query text and the click order of the detection tag category to obtain the weight ratio, the recommended cases are ranked according to the weight ratio, thereby improving the accuracy of similar case recommendations;

[0069] 3. By analyzing the detection query labels and detection query words to obtain benchmark selection cases, the number of recommended cases in the recommended case ranking is increased and re-ranked, thereby improving the accuracy of similar case recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is a method flow of an intelligent case recommendation method according to an embodiment of the present invention. Figure 1 ;

[0071] Figure 2 This is a method flow of an intelligent case recommendation method according to an embodiment of the present invention. Figure 2 ;

[0072] Figure 3 This is the method flow for obtaining the recommended case ranking according to an embodiment of the present invention. Figure 1 ;

[0073] Figure 4 This is the method flow for obtaining the recommended case ranking according to an embodiment of the present invention. Figure 2 ;

[0074] Figure 5 This is a flow chart of the method after obtaining the recommended case ranking according to an embodiment of the present invention. DETAILED DESCRIPTION

[0075] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0076] Reference Figure 1 , the embodiment of the present application discloses an intelligent similar case recommendation method, comprising the following steps:

[0077] S100: Collect and detect label categories.

[0078] The case recommendation platform is a platform set up by technical personnel for case inquiries and case recommendations. The case recommendation platform has a dialog box for users to enter text to search for similar cases.

[0079] Detection tag categories refer to the types of case tags clicked by users on the case recommendation platform. Detection tag categories can be retrieved from the case recommendation platform. These categories include number of individuals, execution methods, execution methods, location, age, and case outcomes. The names of these categories are pre-defined by those skilled in the art and are not detailed here.

[0080] S101: When no text appears in a preset text box, a reference case is searched from a preset case library in response to a detected label category.

[0081] The case library is a database set up by technicians to store cases. A benchmark case is one that has a specific detection label category. If no text appears in the text box, it indicates that only the selected detection label category is needed to recommend similar cases. In this case, benchmark cases are searched from the case library based on the detection label category. A benchmark case must contain at least one detection label category.

[0082] S102: Obtaining the number of benchmark labels and the total number of benchmark labels in response to the benchmark case and the detection label category.

[0083] The total number of benchmark labels refers to the total number of label categories in the benchmark case. The total number of label categories is obtained from the benchmark case as the total number of benchmark labels. The number of benchmark labels refers to the number of detection label categories in the benchmark case. The number of detection label categories is obtained from the benchmark case as the number of benchmark labels.

[0084] S103: Obtain the number of detection tags in response to the detection tag category.

[0085] The number of detection labels refers to the number of labels in the detection label category, and the number of labels retrieved from the detection label category is used as the number of detection labels.

[0086] S104: Obtain case coincidence in response to the number of reference tags, the total number of reference tags, and the number of detection tags.

[0087] Case overlap refers to the overlap between the tag clicked by the user and the benchmark case. The case overlap is obtained by querying the preset overlap comparison table based on the number of benchmark tags, the total number of benchmark tags, and the number of detection tags. The overlap comparison table stores the case overlap corresponding to different numbers of benchmark tags, the total number of benchmark tags, and the number of detection tags. The fewer the total number of benchmark tags and the smaller the difference between the number of benchmark tags and the number of detection tags, the greater the case overlap. The parameters in the overlap comparison table are set by those skilled in the art in advance based on actual experiments and will not be elaborated here.

[0088] S105: Obtaining a recommended case ranking in response to the case overlap and the benchmark case, and outputting the recommended case ranking to a preset display area.

[0089] The display area is the area on the case recommendation platform set by technical personnel for displaying similar cases.

[0090] Recommended case ranking refers to the ranking of cases that are recommended to users based on similar cases. The ranking is achieved by arranging the benchmark cases from large to small based on the degree of overlap between the cases. The ranked cases are used as the recommended case ranking, and the recommended case ranking is output to the display area.

[0091] Reference Figure 2 , an intelligent similar case recommendation method, further comprising:

[0092] S200: When text appears in a preset text box, the query text is collected and detected.

[0093] The detection query text refers to the text entered by the user in the text box. When text appears in the text box, it means that the user needs to recommend similar cases by detecting the selection of the label category and the text entered by the user. The detection query text is then retrieved from the text box of the case recommendation platform.

[0094] S201: Responding to a detection query text to obtain detection query semantics.

[0095] Detecting query semantics refers to detecting the linguistic meaning expressed by the query text, and obtaining the query semantics by analyzing the query text. Methods for analyzing text semantics are common knowledge among those skilled in the art and will not be described in detail here.

[0096] S202: Responding to the detection query semantics to obtain the detection unprocessed text and the reference processed text.

[0097] The detected unprocessed text refers to the case text that has not been processed in the case library, and the benchmark processed text refers to the case text that has been processed in the case library. By detecting the query semantics, the semantically similar benchmark processed text and unprocessed text are matched from the case library, and the unprocessed text is analyzed to obtain the detected unprocessed text.

[0098] S203: Obtaining an unprocessed weight in response to detecting the unprocessed text and a preset reference tag category.

[0099] The benchmark tag categories are defined by technical personnel for user selection within the case recommendation platform. The unprocessed weight refers to the weighted proportion of vocabulary belonging to the benchmark tag category present in the unprocessed text. The unprocessed weight is determined by analyzing the unprocessed text and the benchmark tag categories. The method for determining the unprocessed weight is well-known to those skilled in the art and will not be elaborated upon here.

[0100] In this embodiment, the unprocessed weight represents the importance of the reference tag category in the unprocessed text. The larger the value of the unprocessed weight, the more important the reference tag category is in the unprocessed text.

[0101] S204: Obtain non-overlapping labels in response to the unprocessed weights and the detected label categories.

[0102] Non-overlapping labels refer to labels in the labels corresponding to the unprocessed weights that do not overlap with the detection label categories. The unprocessed labels with weights not equal to 0 are retrieved from the unprocessed weights, and the labels in the detection label categories that do not overlap with the unprocessed labels are used as non-overlapping labels.

[0103] S205: Processing text with non-overlapping labels and benchmarks and detecting label categories to obtain a ranking of recommended cases.

[0104] The recommended case ranking is obtained by analyzing the non-overlapping labels, benchmark processed texts, and detection label categories.

[0105] Reference Figure 3 , the method for obtaining the recommended case ranking includes:

[0106] S300: Obtaining marked processed text and other processed text by processing the text with non-overlapping labels and the reference.

[0107] Marked text refers to the text in the baseline processed text that does not contain overlapping tags, while other processed text refers to the text in the baseline processed text that contains overlapping tags. The marked text and other processed text are obtained by splitting and analyzing the overlapping tags and the baseline processed text.

[0108] S301: Responding to the label processing text and the detection label category to obtain the label processing weight.

[0109] The tagging weight refers to the weight of the detection tag category that appears in the tagging text. The tagging weight is obtained by performing a weight analysis on the tagging text and the detection tag category. The method for determining the tagging weight is common knowledge among those skilled in the art and will not be detailed here.

[0110] S302: defining other processed texts corresponding to the marked processed weights that are consistent with the unprocessed weights as detected processed texts.

[0111] Other processed texts corresponding to the marked processed weights that are consistent with the unprocessed weights are defined as detected processed texts.

[0112] S303: Processing the text and the non-overlapping labels in response to the detection to obtain a non-overlapping weight.

[0113] The non-overlap weight refers to the weight ratio of the non-overlapping labels in the detection processing text. The non-overlapping weight is obtained by performing a weight analysis on the detection processing text and the non-overlapping labels.

[0114] S304: Obtain a benchmark selected case in response to the benchmark case and the detection label category.

[0115] The benchmark selected cases are benchmark cases that contain a detection label category. The benchmark selected cases are selected from the benchmark cases that contain a detection label category. The benchmark selected cases contain at least one detection label category.

[0116] S305: Select cases in response to the detected label categories, the non-overlapping weights, and the benchmark to obtain a recommended case ranking.

[0117] The recommended case ranking is obtained by analyzing the detection label categories, non-overlapping weights, and benchmark selection cases.

[0118] Reference Figure 4 , the method for obtaining the recommended case ranking includes:

[0119] S400: Responding to the detection tag category to obtain a detection selection order.

[0120] The detection selection order refers to the order in which users click on the detection label categories, and the detection selection order is retrieved from the case recommendation platform.

[0121] S401: Obtaining an adjustment coefficient and a number of detection selections in response to a detection selection order.

[0122] The adjustment coefficient is the coefficient used to adjust the weight of the detection tag categories corresponding to each detection selection order. The adjustment coefficient is calculated by analyzing the detection selection order. The later the detection tag category is in the detection selection order, the larger the adjustment coefficient. For example, the first, second, and third detection tag categories correspond to adjustment coefficients of 1, 2, and 3, respectively.

[0123] The number of detection selections refers to the number of detection label categories in the detection selection order, and the number of detection selections is obtained by retrieving the number of detection label categories in the detection selection order from the case recommendation platform.

[0124] S402: Obtain weight proportions by adjusting coefficients, detecting the number of selections, and non-overlapping weights.

[0125] The weight percentage refers to the weight of the case that the user needs to search for. The weight percentage is obtained by querying the preset weight lookup table based on the adjustment coefficient, the number of detections selected, and the non-overlapping weight. The weight lookup table stores the weight percentages corresponding to different adjustment coefficients, the number of detections selected, and the non-overlapping weight.

[0126] The parameters in the weight query table are set by those skilled in the art based on actual conditions and will not be described in detail here. When the number of selected tests and the weight of non-overlapping remain unchanged, the larger the adjustment coefficient, the smaller the weight ratio.

[0127] S403: Select cases in response to the benchmark to obtain a benchmark weight ratio.

[0128] The benchmark weight ratio refers to the weight ratio of the detection label category in the benchmark selected cases. The benchmark weight ratio is obtained by analyzing the detection label category in the benchmark selected cases. The method for obtaining the benchmark weight ratio is common knowledge to those skilled in the art and will not be detailed here.

[0129] S404: Obtaining a ranking of recommended cases in response to the weight ratio and the benchmark weight ratio.

[0130] The recommended case ranking is obtained by matching and sorting the difference between the weight ratio of the corresponding detection label category and the benchmark weight ratio. The smaller the difference between the weight ratio and the benchmark weight ratio, the higher the corresponding case ranking.

[0131] Reference Figure 5 , after obtaining the recommended case ranking, the method further includes:

[0132] S500: Obtain a detection query label in response to a detection query text and a reference label category.

[0133] The detection query tag refers to the tag contained in the detection query text, which is obtained by analyzing the detection query text and the reference tag category. The analysis method of the detection query tag is common knowledge among those skilled in the art and will not be described in detail here.

[0134] S501: Retrieve a search tag, a subject tag, and other query tags in response to detecting a query tag.

[0135] Search tags refer to tags used for auxiliary searches within a test query tag. Subject tags refer to the vocabulary used as a benchmark when using search tags for auxiliary searches within a test query tag. Other query tags refer to tags within a test query tag that are not associated with search tags or subject tags. Search tags, subject tags, and other query tags are obtained by analyzing the test query tag and the test query vocabulary.

[0136] Example: The detection query tags are A person, B location, and C execution method. Location B in the detection query tag is a specific location. In this case, location B is expanded to include location B1 as the actual search tag. A person is the subject tag, which remains unchanged. C execution method is another query tag.

[0137] S502: Responding to the search tags and the benchmark cases to obtain similar cases.

[0138] Similar cases refer to benchmark cases that include search tags, and benchmark cases that include search tags are found from benchmark cases as similar cases.

[0139] S503: Responding to similar cases and subject labels to obtain labeled similar cases.

[0140] Marked similar cases refer to similar cases that contain subject tags. Similar cases that contain subject tags are found from similar cases as marked similar cases.

[0141] S504: In response to marking similar cases, searching tags and subject tags to obtain benchmark selected cases.

[0142] New benchmark cases are selected by analyzing the marked similar cases, retrieval tags and subject tags.

[0143] Methods for obtaining benchmark selection cases include:

[0144] S600: Retrieving texts that are not corresponding to the search tag and the subject tag from the marked similar cases as marked similar texts.

[0145] Marked similar text refers to text that does not correspond to the search label and the subject label in the marked similar case. The marked similar text is obtained by retrieving text that does not correspond to the search label and the subject label from the text of the marked similar case.

[0146] S601: Responding to marking similar texts to obtain marking similar tags.

[0147] The similar tags are tags in similar texts, and are obtained by performing tag analysis on the similar texts. The analysis method of similar tags is well known to those skilled in the art and will not be described in detail here.

[0148] S602: Define the text of the marked similar case whose marked similar tags are consistent with other query tags as the benchmark selected case.

[0149] Cases are selected by defining texts with similar labeled cases that are consistent with other query labels as benchmark cases.

[0150] Methods for getting unprocessed text for detection include:

[0151] S700: Responding to detecting query semantics to obtain unprocessed text.

[0152] Unprocessed text refers to text in the case database that has not been processed. Refer to S202 to obtain the unprocessed text.

[0153] S701: When the number of unprocessed texts is not 1, obtain a detection query vocabulary in response to the detection query semantics and the reference tag category.

[0154] When the number of unprocessed texts is 1, the unprocessed text is used as the detected unprocessed text.

[0155] The detection query vocabulary refers to the vocabulary corresponding to the benchmark label category in the detection query semantics. When the number of unprocessed texts is not 1, it means that there are multiple unprocessed texts. The detection query vocabulary is obtained by analyzing the detection query semantics and the benchmark label category. The analysis method of the detection query vocabulary is common knowledge among technicians in this field and will not be elaborated here.

[0156] S702: Responding to the detection query words, the unprocessed text and the reference tag category to obtain the detected unprocessed text.

[0157] The detection unprocessed text is obtained by analyzing the detection query words, unprocessed text and benchmark label categories.

[0158] Methods for getting unprocessed text for detection include:

[0159] S800: generating a detection integrated vocabulary in response to a detection query vocabulary.

[0160] The detection integration vocabulary refers to the category name vocabulary that classifies the detection query vocabulary. The detection integration vocabulary is obtained by analyzing the detection query vocabulary. For example, the names of various characters are detection query words, so the names are used as detection integration vocabulary.

[0161] S801: Detecting the integrated vocabulary to obtain the integrated association coefficient.

[0162] The integrated correlation coefficient refers to the correlation coefficient between each detection integration word within a case. Each detection integration word is searched for the integrated correlation coefficient in a preset correlation table. The correlation table stores the corresponding correlation coefficients between different detection integration words. The parameters in the correlation table are set by those skilled in the art based on actual experimental conditions and are not detailed here.

[0163] S802: Obtaining unprocessed coefficients in response to unprocessed text and reference label categories.

[0164] The unprocessed coefficient refers to the correlation coefficient between the words corresponding to the benchmark label category in the unprocessed text. The unprocessed coefficient is obtained by retrieving the integrated words of the benchmark label category from the unprocessed text and searching the retrieved words in the correlation comparison table.

[0165] S803: When the unprocessed coefficient is greater than the integrated correlation coefficient, the unprocessed text corresponding to the unprocessed coefficient greater than the integrated correlation coefficient is defined as a marked text.

[0166] When the unprocessed coefficient is greater than the integrated relevance coefficient, it indicates that the unprocessed text is semantically relevant to the detection query, and the unprocessed text corresponding to the unprocessed coefficient greater than the integrated relevance coefficient is defined as the marked text.

[0167] S804: When the number of marked texts is not 1, the marked text with the largest unprocessed coefficient is defined as the detected unprocessed text.

[0168] When the number of marked texts is not 1, it means that there are still multiple marked texts, and the marked text with the largest unprocessed coefficient is defined as the detected unprocessed text.

[0169] The method after collecting the detection query text also includes:

[0170] S900: Collecting detection query images and detection query voice.

[0171] The detection query image refers to the image input by the user into the case recommendation platform, and the detection query voice refers to the voice input by the user into the case recommendation platform. The image input into the case recommendation platform and used for similar case retrieval is used as the detection query image, and the corresponding voice is used as the detection query voice.

[0172] S901: Responding to detecting a query image and a preset detection feature to obtain a marked detection image.

[0173] Detection features are the features set by technicians for case detection. Marking detection images refers to detecting images within the query image that the user is investigating. For example, if there are footprints in an image, the footprints are typically framed with a red or white line in the marking detection image. The image containing the footprints is then used as the marked detection image.

[0174] S902: Detecting the image in response to the mark to generate mark-converted text.

[0175] Marked-to-text refers to the text information expressed by the marked-to-detect image, and is obtained by analyzing the marked-to-detect image. The image-to-text method is well known to those skilled in the art and will not be described in detail here.

[0176] S903: Generate a detection conversion text in response to the detection query voice.

[0177] Detecting the converted text refers to detecting the text expressed by the query voice, converting the sound in the detected query voice into text, and using the text as the detected converted text. The method of converting sound into text is common knowledge among those skilled in the art and will not be described in detail here.

[0178] S904: In response to detecting the transformed text and marking the transformed text, update the detected query text.

[0179] The detection query text is regenerated by adding the detection conversion text and the markup conversion text to the detection query text. The method of adding multiple texts to generate the final text is common knowledge to those skilled in the art and will not be described in detail here.

[0180] Based on the same inventive concept, an embodiment of the present invention provides an intelligent similar case recommendation system, including:

[0181] The acquisition module is used to obtain the detection label category, detection query text, detection query image and detection query voice;

[0182] A memory for storing a program for an intelligent similar case recommendation method;

[0183] The processor is configured to load, execute, and implement the program stored in the memory.

[0184] Those skilled in the art will clearly understand that for the sake of convenience and brevity, the division of the above-mentioned functional modules is only used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working processes of the above-mentioned systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0185] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An intelligent similar case recommendation method, characterized in that: include: S100: Collect and detect label categories; S101: When no text appears in the preset text box, searching for a reference case from a preset case library in response to the detection tag category; S102: Obtaining a number of benchmark tags and a total number of benchmark tags in response to the benchmark case and the detection tag category; S103: Obtaining the number of detection tags in response to the detection tag category; S104: Obtaining case coincidence in response to the number of reference tags, the total number of reference tags, and the number of detection tags; S105: obtaining a recommended case ranking in response to the case overlap degree and the benchmark case, and outputting the recommended case ranking to a preset display area; Also includes: S200: When text appears in a preset text box, collect and detect the query text; S201: Responding to the detection query text to obtain detection query semantics; S202: Responding to the detection query semantics to obtain a detection unprocessed text and a reference processed text; S203: In response to the detecting of the unprocessed text and the preset reference label category to obtain an unprocessed weight; S204: Obtaining non-overlapping labels in response to the unprocessed weights and the detected label categories; S205: Obtain the recommended case ranking by using the non-overlapping labels, the benchmark processing text, and the detection label category.

2. The intelligent similar case recommendation method according to claim 1, characterized in that: The method for obtaining the recommended case ranking includes: S300: Obtaining a marked processed text and other processed texts by using the non-overlapping labels and the reference processed text; S301: Responding to the tag processing text and the detection tag category to obtain a tag processing weight; S302: defining the other processed text corresponding to the marked processed weight consistent with the unprocessed weight as the detected processed text; S303: processing the text and the non-overlapping labels in response to the detection to obtain a non-overlapping weight; S304: Responding to the benchmark case and the detection label category to obtain a benchmark selection case; S305: Selecting cases in response to the detection label categories, the non-overlapping weights, and the benchmark to obtain the recommended case ranking.

3. The intelligent similar case recommendation method according to claim 2, characterized in that: The method for obtaining the recommended case ranking includes: S400: Responding to the detection tag category to obtain a detection selection order; S401: Obtaining an adjustment coefficient and a number of detection selections in response to the detection selection order; S402: Obtaining a weight ratio by using the adjustment coefficient, the number of detection selections, and the non-overlapping weight; S403: Selecting cases in response to the benchmark to obtain a benchmark weight ratio; S404: Obtaining the recommended case ranking in response to the weight ratio and the benchmark weight ratio.

4. The intelligent similar case recommendation method according to claim 2, characterized in that: After obtaining the recommended case ranking, the method further includes: S500: Responding to the detection query text and the reference tag category to obtain a detection query tag; S501: Retrieving a search tag, a subject tag, and other query tags in response to the detection query tag; S502: Responding to the search tag and the benchmark case to obtain similar cases; S503: Responding to the similar cases and the subject tags to obtain labeled similar cases; S504: Obtain the benchmark selected case in response to the marked similar case, the search tag, and the subject tag.

5. The intelligent similar case recommendation method according to claim 4, characterized in that: Methods for obtaining the benchmark selection cases include: S600: Retrieving texts that are not corresponding to the search tag and the subject tag from the similarly marked cases as similarly marked texts; S601: Responding to the marked similar text to obtain a marked similar tag; S602: defining the text of the similarly marked case whose similarity tag is consistent with the other query tag as the benchmark selected case.

6. The intelligent similar case recommendation method according to claim 1, characterized in that: The method for obtaining the detected unprocessed text includes: S700: Responding to the detection query semantics to obtain unprocessed text; S701: When the number of the unprocessed texts is not 1, obtaining a detection query vocabulary in response to the detection query semantics and the reference tag category; S702: Obtain the detected unprocessed text in response to the detected query term, the unprocessed text, and the reference tag category.

7. The intelligent similar case recommendation method according to claim 6, characterized in that: The method for obtaining the detected unprocessed text includes: S800: generating the detection integrated vocabulary in response to the detection query vocabulary; S801: integrating vocabulary through the detection to obtain an integrated relevance coefficient; S802: Obtaining an unprocessed coefficient in response to the unprocessed text and the reference label category; S803: When the unprocessed coefficient is greater than the integrated correlation coefficient, defining the unprocessed text corresponding to the unprocessed coefficient greater than the integrated correlation coefficient as a marked text; S804: When the number of the marked texts is not 1, the marked text with the largest unprocessed coefficient is defined as the detected unprocessed text.

8. The intelligent similar case recommendation method according to claim 1, characterized in that: After collecting the detection query text, the method further includes: S900: Collect detection query images and detection query voice; S901: Responding to the detection query image and the preset detection features to obtain a marked detection image; S902: generating a marked conversion text in response to the marked detection picture; S903: generating a detection conversion text in response to the detection query voice; S904: Responding to the detected transformed text and the marked transformed text to update the detected query text.

9. An intelligent similar case recommendation system, characterized in that: include: Acquisition module, used to obtain detection label categories; A memory for storing a program of an intelligent similar case recommendation method according to any one of claims 1 to 8; The processor is configured to load, execute, and implement the program stored in the memory.

Images