Law case retrieval method based on digital intelligence

Abstract

The invention relates to the technical field of information retrieval, and particularly discloses a law case retrieval method based on digital intelligence, and the method comprises the steps: 1, obtaining a retrieval demand text of a user, sorting the retrieval demand text based on A I, and obtaining a case element extraction table; step 2, retrieving in a database according to a plurality of preset retrieval strategies on the basis of the case element extraction table, and respectively obtaining first N matching cases under each preset retrieval strategy; and 3, calculating matching values of all matched cases, displaying the cases for the user in sequence according to the matching values, and dynamically adjusting the display sequence of the cases according to the historically selected cases of the user. According to the method, the display sequence is dynamically adjusted according to the cases historically selected by the user, the display sequence of subsequent cases can be adjusted according to the tendency of the identity of the user, different requirements of different users for legal cases are met, and the retrieval efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of information retrieval technology, and in particular to a legal case retrieval method based on digital intelligence. Background Art

[0002] With the rapid development of big data on the Internet, legal practitioners have found it increasingly convenient to obtain legal cases. By searching for required cases in databases, users can handle cases, learn relevant knowledge, write relevant papers, etc. However, with the accumulation and expansion of cases in various official and commercial databases, it has become increasingly difficult to obtain required cases from massive databases. Therefore, it is necessary to set various search conditions for retrieval.

[0003] Existing intelligent legal case retrieval methods mainly use machine learning and AI technologies to assist in completing retrieval content. First, they are trained through a large number of samples, and then the machine learning model or AI model is used to identify the user's needs, obtain the various elements of the case, and search through the elements of the case, thereby helping users obtain the legal cases they need.

[0004] Although the existing intelligent legal case retrieval system can effectively assist users in the process of organizing the required information, due to the different identities of users, they also have differentiated needs for retrieval results. For example, lawyer users prioritize displaying the defense strategies in losing cases, while judge users highlight the judgment rules of similar cases by higher courts. Therefore, unified retrieval results cannot meet the needs of different user identities. At the same time, when users fail to retrieve the required cases, the repeated retrieval process will waste a lot of time. Therefore, how to meet the different needs of different users for legal cases and improve the efficiency of retrieval is the fundamental problem to be solved by the present invention. Summary of the Invention

[0005] The purpose of this invention is to provide a digital legal case retrieval method to solve the following technical problems:

[0006] How to meet the different needs of different users for legal cases and improve retrieval efficiency.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A digitally intelligent legal case retrieval method, comprising:

[0009] Step 1: Obtain the user's search requirement text, organize the search requirement text based on AI, and obtain the case element extraction table;

[0010] Step 2: Search the database according to several preset search strategies based on the case element extraction table, and obtain the top N matching cases under each preset search strategy;

[0011] Step 3: Calculate the matching values of all matching cases, display cases to users in order based on the matching values, and dynamically adjust the display order of cases based on the user's historical selections.

[0012] Through the above technical solution, various preset search strategies can be combined to obtain cases that are more closely matched with user needs for display. More importantly, the display order can be dynamically adjusted based on the user's historically selected cases. In this process, on the one hand, it can reduce the time required for users to re-search. At the same time, by comparing the user's historically selected cases with cases obtained by different preset search strategies, the order of subsequent case displays can be adjusted according to the user's identity tendencies, thereby meeting the different needs of different users for legal cases and improving retrieval efficiency.

[0013] Furthermore, the preset search strategy includes a judge preset search strategy, a lawyer preset search strategy, a scholar preset search strategy and a basic preset search strategy.

[0014] Through the above technical solution, it is possible to perform searches simultaneously according to the above four preset search strategies, obtain matching cases preferred by users with different identities, and then compare the user's historical selection cases with the cases obtained by different preset search strategies. The order of subsequent case display is adjusted according to the user's identity preference, thereby meeting the different needs of different users for legal cases and improving the efficiency of retrieval.

[0015] Furthermore, the matching value calculation process includes:

[0016] De-duplicate all matching cases;

[0017] By formula Calculate the matching value v of all matching cases after deduplication match , display cases for users in descending order of matching values;

[0018] Where i = 1, 2, 3, 4, represents the order of the preset search strategy, λ i is the preset weight of the i-th preset search strategy, m is the matching item in the case element extraction table, j = 1, 2, ..., m; p j is the judgment factor. When the matching result of the jth case element extraction table is consistent, p j =1, when the matching result of the jth case element extraction table matching item is not consistent, p j =0, when the matching result is similar, p j =0.5; χij is the influence coefficient of the matching item in the jth case element extraction table of the i-th preset retrieval strategy, f i (x) is the judgment function. When the current matching case appears in the first N matching cases obtained by the i-th preset search strategy, f i (x) = x, otherwise, f i (x)=0.

[0019] Through the above technical solution, it is possible to display more comprehensive search results according to the needs of users with different identities in the initial search results, so that users with different identities can obtain corresponding required cases in the search results, which provides a basis for dynamically adjusting the display order of cases according to the user's historical selection of cases in the subsequent process, thereby meeting the different needs of different users for legal cases and improving the efficiency of retrieval.

[0020] Furthermore, the process of dynamically adjusting the case display order includes:

[0021] Calculate the correlation coefficient between the user's historically selected cases and each preset retrieval strategy;

[0022] The preset weight of each preset search strategy is adjusted according to the size of the correlation coefficient, the browsed cases are removed, and the matching values of the remaining matching cases after adjustment are calculated; and the cases are displayed according to the matching values of the remaining matching cases after adjustment.

[0023] Through the above technical solution, by calculating the correlation coefficient between the cases historically selected by the user and each preset search strategy, the stronger the correlation between the cases historically selected by the user and a certain preset search strategy, the closer the user's needs are to the identity corresponding to the preset search strategy, thereby realizing the judgment of the user's demand tendency for legal cases and the user's identity, and then adjusting the preset weight of each preset search strategy according to the size of the correlation coefficient, so that the subsequently displayed search cases are more in line with the user's needs and identity, and by removing the browsed cases, calculating the matching values of the remaining matching cases after adjustment; and displaying the cases according to the matching values of the remaining matching cases after adjustment, through the above process, the subsequently displayed legal cases can continue to approach the direction of user needs, thereby improving the pertinence of the search process.

[0024] Furthermore, the calculation process of the correlation coefficient includes:

[0025] By formula Calculate the correlation coefficient R between the user and the i-th preset search strategy i ;

[0026] According to the correlation coefficient R i The preset weight of each preset retrieval strategy is adjusted based on the size of .

[0027] Through the above technical solution, the preset weight of each preset search strategy is adjusted according to the size of the correlation coefficient, so that the subsequently displayed legal cases continue to move closer to the user's needs, thereby improving the pertinence of the search process.

[0028] Furthermore, the process of adjusting the preset weight of each preset search strategy includes:

[0029] By formula Calculate the adjusted preset weight λt of the i-th preset retrieval strategy i ;

[0030] According to the adjusted preset weight λt i Calculate the matching value for the remaining matched cases after adjustment.

[0031] Through the above technical solution, the greater the correlation coefficient between the user and the i-th preset search strategy, the greater the adjusted preset weight of the i-th preset search strategy will be, thereby realizing the process of adjusting the preset weight of each preset search strategy according to the size of the correlation coefficient.

[0032] Furthermore, the retrieval process in step 2 adopts a hybrid retrieval mode, which includes semantic retrieval and logical retrieval, and searches are performed according to semantic retrieval and logical retrieval respectively to obtain corresponding retrieval results, and the retrieval results are integrated according to the preset retrieval weights;

[0033] Adjust the selection weights of semantic retrieval and logical retrieval based on the user's historical retrieval data.

[0034] Through the above technical solution, the advantages of semantic retrieval and logical retrieval can be combined to improve the comprehensiveness of the retrieval process. At the same time, the selection weights of semantic retrieval and logical retrieval can be adjusted according to the user's historical retrieval data, and the selection weights of semantic retrieval and logical retrieval can be dynamically adjusted, thereby further improving the accuracy of retrieval case results.

[0035] Furthermore, the process of adjusting the weights of semantic retrieval and logical retrieval selection includes:

[0036] Compare the user's historical search data with the search results corresponding to semantic search and logical search respectively, and use the formula Calculate the priority value of each search method; where Q is the number of search results corresponding to the user's historical search data in the corresponding search mode, k = 1, 2, ..., Q;

[0037] Determine the priority values of semantic retrieval and logical retrieval, and adjust the selection weight toward the side with the larger priority value. The adjustment amount is obtained by comparing the interval range of the difference in priority values.

[0038] Through the above technical solution, it is possible to determine by priority value which retrieval mode obtains results that are more in line with user needs, and then dynamically adjust the selection weights of semantic retrieval and logical retrieval. By judging the size of the priority values of semantic retrieval and logical retrieval, the selection weight is adjusted to the side with a larger priority value, thereby adjusting the accuracy of the retrieval results.

[0039] Beneficial effects of the present invention:

[0040] (1) The present invention detects through a variety of preset search strategies, and can display more comprehensive search results according to the needs of users with different identities in the display of the initial search results, so that users with different identities can obtain corresponding demand cases in the search results. By calculating the correlation coefficient between the user's historically selected cases and each preset search strategy, the user's demand tendency for legal cases and the user's identity are judged. Then, the preset weight of each preset search strategy is adjusted according to the size of the correlation coefficient, so that the subsequent displayed search cases are more matched with the user's needs and identity, and the subsequent displayed legal cases can continue to move closer to the user's needs, thereby improving the pertinence of the search process; at the same time, the display order is dynamically adjusted according to the user's historically selected cases. On the one hand, it can reduce the time required for the user to re-search. At the same time, by comparing the user's historically selected cases with the cases obtained by different preset search strategies, the order of subsequent case display can be adjusted according to the user's identity tendency, thereby meeting the different needs of different users for legal cases and improving the efficiency of the search.

[0041] (2) The present invention adopts a hybrid retrieval mode and adjusts the selection weights of semantic retrieval and logical retrieval according to the user's historical retrieval data. It can combine the advantages of semantic retrieval and logical retrieval to improve the comprehensiveness of the retrieval process. At the same time, it adjusts the selection weights of semantic retrieval and logical retrieval according to the user's historical retrieval data, and dynamically adjusts the selection weights of semantic retrieval and logical retrieval, thereby further improving the accuracy of the retrieval case results. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention will be further described below with reference to the accompanying drawings.

[0043] Figure 1 This is a flowchart of the steps of the digital legal case retrieval method of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0045] In one embodiment, a digital legal case retrieval method is provided. Figure 1 As shown, the method includes: step one, obtaining the user's search demand text, sorting the search demand text based on AI, and obtaining a case element extraction table. In step one, the search demand text is sorted through AI technology, which can be extracted from the user's search demand text to the case element extraction table. The case element extraction table can reflect the user's search needs. Through this process, the information sorting process can be more conveniently realized; step two, based on the case element extraction table, searching in the database according to several preset search strategies, respectively obtaining the top N matching cases under each preset search strategy; step two searches through several preset search strategies, which can meet the search content needs of different user identities. Then, step three is used to calculate the matching values of all matching cases, and the cases are displayed to the user in order according to the size of the matching values, and the display order of the cases is dynamically adjusted according to the user's historical selection of cases. Through the above process, various preset search strategies can be combined to obtain cases that are more closely matched with user needs for display. More importantly, the display order is dynamically adjusted according to the user's historical selection of cases. In this process, on the one hand, it can reduce the time required for the user to re-search, and at the same time, by comparing the user's historical selection of cases with cases obtained by different preset search strategies, the order of subsequent case displays can be adjusted according to the user's identity tendency, thereby meeting the different needs of different users for legal cases and improving the efficiency of retrieval.

[0046] In one embodiment, the preset search strategy includes a judge preset search strategy, a lawyer preset search strategy, a scholar preset search strategy and a basic preset search strategy, wherein the basic preset search strategy does not consider the search strategy set by the user identity, while the judge preset search strategy, the lawyer preset search strategy and the scholar preset search strategy are respectively set according to the common identities of the legal case inquirer (judge, lawyer, scholar). By searching simultaneously through the above four preset search strategies, matching cases preferred by users with different identities can be obtained, and then the user's historical selected cases are compared with the cases obtained by different preset search strategies, and the order of subsequent case display is adjusted according to the user's identity preference, thereby meeting the different needs of different users for legal cases and improving the efficiency of retrieval.

[0047] In one embodiment, a matching value calculation process is provided, including: first, deduplication of all matching cases. Since the retrieval cases obtained by different preset retrieval strategies may have the same situation, deduplication is used to reduce the subsequent calculation amount. Then, the formula Calculate the matching value v of all matching cases after deduplication match , display the cases to users in descending order of matching values; in this embodiment, there are 4 preset search strategies, so i = 1, 2, 3, 4, which represents the order of the preset search strategies, λ i is the preset weight of the i-th preset retrieval strategy, and the preset weight λ i It can be set according to the proportion of different user identities in the legal case database query data, m is the matching item of the case element extraction table, j = 1, 2, ..., m; p j is the judgment factor. When the matching result of the jth case element extraction table is consistent, p j =1, when the matching result of the jth case element extraction table matching item is not consistent, p j =0, when the matching result is similar, p j =0.5; It should be noted that the judgment process of the judgment factor is based on the logical rules set by the retrieval system and the assistance of the AI model. Therefore, while being able to judge whether the matching results are consistent, it can also judge similar matching results to improve the comprehensiveness of the retrieval. ij is the influence coefficient of the jth case element extraction table matching item of the i-th preset retrieval strategy. This parameter is set based on the empirical data fitting of different preset retrieval strategies. Therefore, the influence coefficients of different case element extraction table matching items in different preset retrieval strategies are different. i (x) is the judgment function. When the current matching case appears in the first N matching cases obtained by the i-th preset search strategy, f i (x) = x, otherwise, f i (x)=0. Through the calculation process of the matching value in the above technical solution, a more comprehensive search result can be displayed in the initial search result display according to the needs of users with different identities, so that users with different identities can obtain corresponding required cases in the search results. This provides a basis for dynamically adjusting the display order of cases according to the user's historical selection of cases in the subsequent process, thereby meeting the different needs of different users for legal cases and improving the efficiency of retrieval.

[0048] In one embodiment, a process for dynamically adjusting the order of case display is provided, including: calculating the correlation coefficient between the cases historically selected by the user and each preset search strategy; adjusting the preset weight of each preset search strategy according to the size of the correlation coefficient, removing the browsed cases, and calculating the matching value of the remaining matching cases after the adjustment; and displaying the cases according to the matching value of the remaining matching cases after the adjustment. Through the above technical solution, by calculating the correlation coefficient between the cases historically selected by the user and each preset search strategy, the stronger the correlation between the cases historically selected by the user and a certain preset search strategy, the closer the user's needs are to the identity corresponding to the preset search strategy, thereby achieving a judgment of the user's demand tendency for legal cases and the user's identity. Thereafter, the preset weight of each preset search strategy is adjusted according to the size of the correlation coefficient, so that the subsequently displayed search cases are more consistent with the user's needs and identity. By removing the browsed cases, calculating the matching value of the remaining matching cases after the adjustment; and displaying the cases according to the matching value of the remaining matching cases after the adjustment. Through the above process, the subsequently displayed legal cases can be continuously closer to the user's needs, thereby improving the pertinence of the search process.

[0049] In one embodiment, a calculation process of the correlation coefficient is provided, including: Calculate the correlation coefficient R between the user and the i-th preset search strategy i ; According to the correlation coefficient R i The size of the preset weight of each preset search strategy is adjusted by the above correlation coefficient R i The calculation process can judge the proportion of the matching status of the user's browsed cases and the i-th preset search strategy to the average status of all preset search strategies. Obviously, when the correlation coefficient R i The larger the value, the stronger the correlation between the user and the i-th preset search strategy, and then the correlation coefficient R i The preset weight of each preset search strategy is adjusted based on the size of , so that the legal cases displayed subsequently are constantly closer to the user's needs, thereby improving the pertinence of the search process.

[0050] In one embodiment, a process for adjusting the preset weight of each preset search strategy is provided, including: Calculate the adjusted preset weight λt of the i-th preset retrieval strategy i ; According to the adjusted preset weight λt iCalculate the matching values of the remaining matching cases after adjustment. Through the above process, when the correlation coefficient between the user and the i-th preset search strategy is greater, the adjusted preset weight of the i-th preset search strategy will also be greater, thereby realizing the process of adjusting the preset weight of each preset search strategy according to the size of the correlation coefficient.

[0051] In one embodiment, the retrieval process in step 2 adopts a hybrid retrieval mode, which includes semantic retrieval and logical retrieval. The retrieval is performed according to the semantic retrieval and logical retrieval respectively to obtain corresponding retrieval results, and the retrieval results are integrated according to the preset retrieval weights, that is, the selection weights of the different retrieval methods are weighted and accumulated; the selection weights of the semantic retrieval and the logical retrieval are adjusted according to the user's historical retrieval data, wherein the semantic retrieval uses Legal-BERT vectorized query statements to calculate the cosine similarity with the case library, and the logical retrieval parses the natural language input by the user to generate Datal og rules, and searches according to the rules. Through the above technical solution, the advantages of semantic retrieval and logical retrieval can be combined to improve the accuracy of the retrieval process. At the same time, the selection weights of semantic retrieval and logical retrieval are adjusted according to the user's historical retrieval data, and the selection weights of semantic retrieval and logical retrieval are dynamically adjusted, thereby further improving the accuracy of the retrieval case results.

[0052] It should be noted that the various preset search strategies in the above scheme, when executed separately, can all be implemented through the mixed search mode in this embodiment. Therefore, the scheme of this embodiment is coordinated and independent with the overall technical scheme.

[0053] In one embodiment, a process for adjusting the selection weights of semantic retrieval and logical retrieval is provided, including comparing the user's historical retrieval data with the retrieval results corresponding to semantic retrieval and logical retrieval, and using the formula The priority value of each search method is calculated, where Q is the number of search results corresponding to the user's historical search data in the corresponding search mode, k = 1, 2, ..., Q; therefore, through the priority value calculation process, it is possible to determine which search mode obtains results that are more in line with the user's needs, and then dynamically adjust the selection weights of semantic retrieval and logical retrieval. By judging the size of the priority values of semantic retrieval and logical retrieval, the selection weight is adjusted to the side with a larger priority value. According to empirical data, the interval range of the difference between different priority values is divided and the corresponding adjustment ratio is set. Therefore, the adjustment amount can be obtained by comparing the interval range of the priority value difference.

[0054] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A digital-based legal case retrieval method, characterized by: The method comprises: Step 1: Obtain the user's search requirement text, organize the search requirement text based on AI, and obtain the case element extraction table; Step 2: Search the database according to several preset search strategies based on the case element extraction table, and obtain the top N matching cases under each preset search strategy; Step 3: Calculate the matching values of all matching cases, display cases to users in order based on the matching values, and dynamically adjust the display order of cases based on the user's historical selections.

2. The digitalized legal case retrieval method according to claim 1 is characterized in that: The preset search strategies include judge preset search strategies, lawyer preset search strategies, scholar preset search strategies and basic preset search strategies.

3. The digitalized legal case retrieval method according to claim 2 is characterized in that: The calculation process of the matching value includes: De-duplicate all matching cases; By formula Calculate the matching value v of all matching cases after deduplication match , display cases for users in descending order of matching values; Where i = 1, 2, 3, 4, represents the order of the preset search strategy, λ i is the preset weight of the i-th preset search strategy, m is the matching item in the case element extraction table, j = 1, 2, ..., m; p j is the judgment factor. When the matching result of the jth case element extraction table is consistent, p j =1, when the matching result of the jth case element extraction table matching item is not consistent, p j =0, when the matching result is similar, p j =0.5; χ ij is the influence coefficient of the matching item in the jth case element extraction table of the i-th preset retrieval strategy, f i (x) is the judgment function. When the current matching case appears in the first N matching cases obtained by the i-th preset search strategy, f i (x) = x, otherwise, f i (x)=0.

4. The digitalized legal case retrieval method according to claim 3 is characterized in that: The process of dynamically adjusting the case display order includes: Calculate the correlation coefficient between the user's historically selected cases and each preset retrieval strategy; The preset weight of each preset search strategy is adjusted according to the size of the correlation coefficient, the browsed cases are removed, the matching values of the remaining matching cases after adjustment are calculated, and the cases are displayed according to the matching values of the remaining matching cases after adjustment.

5. The digital intelligence-based legal case retrieval method according to claim 4 is characterized in that: The calculation process of the correlation coefficient includes: By formula Calculate the correlation coefficient R between the user and the i-th preset search strategy i ; According to the correlation coefficient R i The preset weight of each preset retrieval strategy is adjusted based on the size of .

6. The digitalized legal case retrieval method according to claim 5 is characterized in that: The process of adjusting the preset weights for each preset search strategy includes: By formula Calculate the adjusted preset weight λt of the i-th preset retrieval strategy i ; According to the adjusted preset weight λt i Calculate the matching value for the remaining matched cases after adjustment.

7. The digitalized legal case retrieval method according to claim 1 is characterized in that: The retrieval process in step 2 adopts a hybrid retrieval mode, which includes semantic retrieval and logical retrieval. The retrieval is performed according to the semantic retrieval and logical retrieval respectively to obtain corresponding retrieval results, and the retrieval results are integrated according to the preset retrieval weights. Adjust the selection weights of semantic retrieval and logical retrieval based on the user's historical retrieval data.

8. The digitalized legal case retrieval method according to claim 2 is characterized in that: The process of adjusting the weights of semantic and logical retrieval selections includes: Compare the user's historical search data with the search results corresponding to semantic search and logical search respectively, and use the formula Calculate the priority value of each search method; where Q is the number of search results corresponding to the user's historical search data in the corresponding search mode, k = 1, 2, ..., Q; Determine the priority values of semantic retrieval and logical retrieval, and adjust the selection weight toward the side with the larger priority value. The adjustment amount is obtained by comparing the interval range of the difference in priority values.

Images