A two-stage case-based matching method for civil law article recommendation

Abstract

The application discloses a civil law article recommendation method based on two-stage case matching, and steps are as follows, Step 1: taking the legal information cited by each civil judgment case as a feature, the case is divided into a case subset; Step 2: identifying the feature words of each case subset, and performing synonym expansion on the feature words; Step 3: calculating the matching degree sim g of a new civil dispute situation and the feature words of each case subset g , and selecting the case subset with the maximum sim g ; Step 4: calculating the similarity of the new civil dispute situation and each case i in the case subset; Step 5: obtaining the preliminary recommendation result of the applicable law article of the new civil dispute situation according to the law article cited by the case in the similar case subset; Step 6: correcting P new , and obtaining the final law article recommendation result. The application can quickly and effectively find similar cases in a large number of cases, and performs more accurate law article recommendation based on the similar cases, and the process of the method has strong interpretability.

Description

Technical Field

[0001] This invention belongs to the fields of intelligent legal technology and case management, and relates to a method for recommending civil law provisions based on two-stage case matching. Background Technology

[0002] In the context of smart justice and smart social governance, the efficient resolution of civil disputes is of paramount importance. Intelligent recommendations of applicable legal provisions for new dispute scenarios can enhance the legal awareness of parties involved in civil disputes, encouraging them to resolve disputes independently. For disputes that have already entered the litigation and trial stage, intelligent legal provision recommendations can assist judicial personnel in adjudicating cases more efficiently.

[0003] Drawing experience from similar cases to inform judgments in new situations is a highly interpretable approach. The core issues to be addressed are how to quickly and effectively identify cases similar to new dispute scenarios, and how to more accurately recommend legal provisions based on similar cases. Summary of the Invention

[0004] Based on the technical problems existing in the background technology, this invention proposes a civil law provision recommendation method based on two-stage case matching. This method is designed to address new civil dispute scenarios. new To obtain a recommendation for the applicable legal provisions, follow these steps:

[0005] Step 1: Using the legal information cited in each civil judgment case as a feature, the spectral clustering algorithm is used to cluster the cases, dividing a large number of cases into k case subsets;

[0006] Step 2: For each case subset g, use the TGF-IGF algorithm to identify the feature words KEY of each case subset. g The feature words are then expanded using the Word2Vec algorithm to obtain KEY_MORE. g ;

[0007] Step 3: Calculate the new civil dispute scenario. new The degree of matching between the feature words and each subset of cases (sim) g Select the one that makes sim g The subset of cases AS with the largest value is used as the case set for subsequent similar case matching;

[0008] Step 4: Calculate the similarity sim(new,i) between the new civil dispute scenario and each case i in the case subset, and obtain the final similar case set AS based on the sim(new,i) value. match ;

[0009] Step 5: Based on the similar case set ASmatch Based on the legal provisions cited in the cases, preliminary recommendations for the applicable legal provisions in new civil dispute scenarios are obtained. new ;

[0010] Step 6: Use the Apriori algorithm to mine the set of legal citation association rules R in AS, and use the association rule set to analyze P. new After making corrections, the final legal provision recommendation result P is obtained. new .

[0011] Furthermore, Step 1 specifically includes:

[0012] Step 1-1: Suppose that the legal information cited in a case includes two parts: legal unit information and legal provision information. Here, "legal unit" is a modification of the concept of "law," aiming to provide a more granular division and representation of laws with strong integration. The specific definition of a "legal unit" is as follows: each part and chapter of the *Civil Code of the People's Republic of China* (hereinafter referred to as the *Civil Code*) is considered a legal unit, and each other law as a whole is considered a legal unit. That is, for a certain law l, its corresponding legal unit l' is:

[0013]

[0014] Each legal unit consists of several legal provisions. Therefore, the legal unit cited in a case can be represented as L = {l1', l2', ..., l m The cited legal provisions can be represented as P = {p1, p2, ..., p}. r ,p r+1 ,p r+2 ,...,p s ,...,p t+1 ,,p t+2 ,...,p y};

[0015] Step 1-2: Use the Jaccard coefficient to measure the differences in case a i And Case a j Similarity between legal units and legal provisions J(L) i ,L j ) and J(P i ,P j ),Right now:

[0016]

[0017]

[0018] Based on this, the overall similarity J(a) of the cases i ,aj This can be represented as:

[0019] J(a i ,a j )=α*J(L i ,L j )+β*J(P i ,P j )

[0020] Where α and β are coefficients, and α + β = 1;

[0021] Steps 1-3: Based on Jaccard similarity, the first stage of clustering is performed using the spectral clustering algorithm for the entire case set A.

[0022] At this point, we will equally focus on the differences in legal units and laws across different cases, therefore we set α = 0.5 and β = 0.5, that is:

[0023] J(a i ,a j )=0.5*J(L i ,L j )+0.5*J(P i ,P j )

[0024] The entire set of cases is clustered according to the process of the spectral clustering algorithm. The silhouette coefficient is used as the evaluation index of the clustering effect under different numbers of categories, and the cases are divided into k_1 first-level case subsets.

[0025] Steps 1-4: For each subset of primary cases, a second-stage clustering algorithm is used. At this stage, the focus is on the legal differences between cases; therefore, α = 0.2 and β = 0.8 are set, i.e.:

[0026] J(a i ,a j )=0.2*J(L i ,L j )+0.8*J(P i ,P j )

[0027] Following the spectral clustering algorithm, cases in the first-level case subsets are clustered. The silhouette coefficient is used as the evaluation metric for clustering performance under different numbers of categories. Each first-level case subset is divided into k_2 second-level subsets. Therefore, for the entire case set A, it is divided into k case subsets through two-stage clustering, i.e.:

[0028] A = {AS1,AS2,...,AS} k}

[0029] Furthermore, Step 2 specifically includes:

[0030] Step 2-1: For each case, extract the plaintiff's claims and the facts found by the court using regular expressions. Then, based on the legal provision numbers cited in the case and combined with publicly available legal provision data, obtain the specific text of the cited legal provision. In other words, for each case, obtain three aspects of information: the plaintiff's claims, the facts found by the court, and the text of the cited legal provision. These three aspects of information can serve as the core textual information of this case.

[0031] Step 2-2: Use the TGF-IGF algorithm to obtain the plaintiff's claim keyword (KEY) for the case subset g. g 1 The TGF-IGF algorithm is a novel algorithm proposed in this invention as an improvement on the existing TF-IDF algorithm. Specifically, for each word i in each class, the final weight of that word in a particular class is obtained by simultaneously measuring the term-group frequency (TGF) of documents containing word i in each class and the inverse group frequency (IGF) of word i in other classes. Specifically, for word i, its TGF value in the g-th class can be obtained by dividing the number of cases containing word i in each class by the total number of cases in that class, calculated as follows:

[0032]

[0033] Where, d ig d is the number of cases in class g where word i appears. g This represents the total number of cases in category g. The IGF value of word i is obtained by taking the reciprocal of the average percentage of cases containing that word across all other categories, followed by the logarithm. The calculation method is as follows:

[0034]

[0035] Where N is the total number of case categories, and N-1 represents the total number of case categories other than the g-th case. This represents the percentage of cases containing word i within a certain category. To prevent the denominator from being 0, an extra 0.01 is added to the denominator of IGF. Based on the above, the formula for calculating TGF-IGF can be derived as follows:

[0036]

[0037] For the plaintiff's claims in the case subset g, the TGF-IGF algorithm is used to obtain the score of each word, and then the Top-N words are selected as the feature word set for this part;

[0038] Step 2-3: Use the TGF-IGF algorithm to obtain the legal factual feature words (KEY) of the case subset g. g 2 ;

[0039] Step 2-4: Use the TGF-IGF algorithm to obtain the key words (KEYs) of the cited legal provisions text in the case subset g. g 3 ;

[0040] Steps 2-5: Merge the words from the three parts to obtain the final word set KEY. g 3 ,Right now:

[0041]

[0042] Step 2-6: For the feature word set KEY g Each feature word key gw The Word2Vec model is used to obtain the five most similar words to the feature word, forming an expanded word set KEY_MORE. gw ,Right now:

[0043] KEY_MORE gw ={key gw ,key gw1 ,key gw2 ,...,key gw5}

[0044] Then the expanded vocabulary set of all feature words is:

[0045] KEY_MORE g ={KEY_MORE g1 ,KEY_MORE g2 ,...,KEY_MORE gn}

[0046] Furthermore, Step 3 specifically includes:

[0047] Step 3-1: For new civil dispute scenarios (ak) new After word segmentation and stop word removal, a new contextual word set KEY is obtained. new ;

[0048] Step 3-2: For each Calculate the new contextual word set KEY new With the expanded feature word set KEY_MORE g The relevance. First, calculate the key. newWith the expanded word set key_more for each feature word in the feature word set gw similarity sim gw The calculation formula is as follows:

[0049]

[0050] Among them, |KEY new ∩KE_Y MO g R w KE, a collection of situational vocabulary n Y ew With each feature word's expanded vocabulary set KEY_MORE gw The total number of words in the vocabulary;

[0051] Step 3-3: Based on this, calculate the contextual word set KEY. new KEY_MORE, an expanded vocabulary of all feature words g similarity sim g The specific calculation formula is as follows:

[0052]

[0053] This result represents a new scenario for civil disputes. new and case subset AS g The correlation between sim and sim is selected. g The subset of cases with the largest value AS match_1 This will serve as the dataset for subsequent similar case matching.

[0054] Furthermore, Step 4 specifically includes:

[0055] Step 4-1: For the subset AS of cases match_1 Each case a in i The Roformer-Sim was used to measure the impact of new input scenarios on the plaintiff's litigation, the court's determination of facts, and Case a. i The text similarity sim1(new,i) and sim2(new,i) of the corresponding parts are calculated, and the similarity sets SIM1(new) and SIM2(new) are obtained respectively.

[0056] Step 4-2: Based on the idea of ​​entropy weighting, calculate the information entropy e1 and e2 of SIM1(new) and SIM1(new) respectively, and then derive the weights w1 and w2. Therefore, in case a... new and a i The formula for calculating the similarity sim(new,i) is:

[0057] sim(new,i)=w1*sim1(new,i)+w2*sim2(new,i)

[0058] Step 4-3: Through calculation, a set of sim(new,i) is obtained. The similarity threshold for the two-stage matching is set to θ. All cases with sim(new,i) values ​​greater than or equal to θ are obtained. If the number of cases meeting the requirement exceeds 10, the 10 cases with the highest sim(new,i) values ​​are selected and included in the final similar case set AS. match If the number does not exceed a certain threshold, then all numbers will be included.

[0059] Furthermore, step 5 specifically includes:

[0060] Step 5-1: Transfer AS match The cases are traversed sequentially in descending order of sim(new,i): the first case traversed is as1, whose referenced legal provisions are set P1, so the recommendation index of all legal provisions in P1 is recorded as 1, and the legal provisions are added to the recommendation set; the second case traversed is as2, whose referenced legal provision P2 has a recommendation index of 0.9, the legal provision is added to the recommendation set, and the recommendation indices of the same legal provision are added together. This process continues, with the recommendation index decreasing by 0.1 each time, but never falling below 0.1.

[0061] Step 5-2: After the traversal is complete, filter out the legal provisions with a recommendation index greater than or equal to 2; if the number of legal provisions that meet the requirements exceeds n, then take the n legal provisions with the highest recommendation index to form a preliminary set of recommended legal provisions.

[0062] Furthermore, step 6 specifically includes:

[0063] Step 6-1: For the subset of cases AS match_1 The set of cited legal provisions in the dataset is used to obtain all legal provision association rules r(sup) that meet the requirements using an Apriori-based association rule algorithm. r >= minsup and con r >= mincon), and construct a set of legal provision association rules R;

[0064] Step 6-2: Sort the rules according to their confidence level, and iterate through the rules in R. If the legal provisions in the preceding terms of a rule are... In the middle, it is determined whether the legal provision following the rule is in If any legal provision in the following item is not present... In the middle, add it to In the middle; if the preceding legal provision is not present. In the case of [the specific rule], this rule is ignored;

[0065] Step 6-3: After traversal is complete, perform statistics. If the number of legal provisions that meet the requirements exceeds n, then the n legal provisions with the highest recommendation index are selected to form the final set of recommended legal provisions P. new This concludes the legal provision recommendation process.

[0066] The beneficial effects of this invention are as follows:

[0067] This invention can quickly and effectively find similar cases from a large number of cases, and make relatively accurate legal provisions recommendations based on similar cases. The process of the method has strong interpretability. Attached Figure Description

[0068] Figure 1 This is a schematic diagram illustrating the steps of the civil law provision recommendation method based on two-stage case matching in this invention;

[0069] Figure 2 This is a flowchart illustrating the civil law provision recommendation method based on two-stage case matching according to the present invention.

[0070] Figure 3 This is a structural diagram of the apparatus for the civil law provision recommendation method based on two-stage case matching of the present invention;

[0071] Figure 4 This is a schematic diagram of the structure of the civil law provision recommendation method based on two-stage case matching of the present invention. Detailed Implementation

[0072] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention.

[0073] The flowchart of the method of the present invention is as follows:

[0074] like Figure 3 As shown, Embodiment 1 of the present invention provides an application environment for a civil law provision recommendation system based on two-stage case matching, including a user terminal 10, a server terminal 20, and a legal case database 30. The user terminal 10 submits a legal provision recommendation request to the server terminal; upon receiving the request, the server terminal 20 executes a computer program to determine the recommended legal provisions by calculating the similarity between the user's case and cases in the legal case database 30, and transmits the result back to the user terminal 10.

[0075] like Figure 4 As shown, Embodiment 2 of the present invention provides a computer device, specifically including: a processor 1001, a memory 1002, a communication module 1003, an input unit 1004, an audio processor 1005, an output unit 1006, and a power supply 1007.

[0076] The computer device 1000 may include one or more processors 1001, such as one or more central processing units (CPUs) or graphics processing units (GPUs), or other general-purpose processors, application-specific integrated circuits, digital signal processors, etc. Each processing unit may implement one or more hardware threads. Related computer programs are stored in memory, and when the processor runs the computer programs and related data in memory, the various functions of the legal provision recommendation device of this invention, which integrates similar cases and legal provision association rules, are realized.

[0077] The memory 1002 can be used to store any kind of information, such as code, settings, and data, and is mainly divided into a data storage area and a program storage area (used to store application programs related to functions such as image and sound playback). In the computer device 1000, the memory 1002 is coupled to the central processing unit 1001. The central processing unit 1001 can execute the programs stored in the memory 1002 to perform operations such as information processing and / or storage.

[0078] The communication module 1003 is the transmitter / receiver 1003, and the computer device 1000 may include one or more communication modules. It transmits and receives signals via the antenna 1008 and exchanges data with other components and devices through one or more communication links.

[0079] The input unit 1004 in the computer device 1000 is used to receive various inputs, and the output unit 1006 is used to present output results such as text and images.

[0080] The audio processor 1005 is coupled to the speaker 1009 for audio output and to the microphone 1010 for audio input, enabling typical telecommunications functions.

[0081] Those skilled in the art will understand that the computer device 1000 shown in Embodiment 2 is merely a portion of the device components related to this invention and does not constitute a limitation on the computer device to which the present application is applied. Those skilled in the art can select and combine some of the components shown in Embodiment 2, or add more components to expand the device, depending on the specific hardware and software conditions.

[0082] In one specific embodiment, the memory 1002 can be a non-volatile memory, such as read-only memory (ROM) or flash memory; or it can be a volatile memory, such as random access memory (RAM). The memory 1002 mainly includes four functional modules: a buffer 1021, an application / function storage module 1022, a data storage module 1023, and a driver storage module 1024. The application / function storage module 1022 stores application and function programs; the data storage module 1023 stores various types of data used by the computer device; and the driver storage module 1024 stores various drivers to execute various functions of the computer device.

[0083] The input unit 1004 can be a button, trackball, or touchpad set on the casing of a computer device, or it can be an external keyboard, touchpad, or mouse, or it can be a touch layer covering the display screen.

[0084] The output unit 1006 is mainly a display screen, which can be either an LCD screen or an e-ink screen.

[0085] The present invention also provides a computer-readable storage medium. The computer program instructions stored in this medium are capable of directing a computer or other processor to execute in a specific manner, thereby realizing the proposed legal provision recommendation method based on the fusion of similar cases and legal provision association rules.

[0086] References to memory, storage, databases, or other media in the embodiments provided in this invention may include non-volatile and / or volatile memory. Non-volatile memory may be read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may be random access memory (RAM) or external cache memory. It should be noted that RAM exists in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), etc. The technical features in the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of technical features are described in detail. However, in the absence of contradictions, any combination of these technical features should be considered within the scope of this specification.

[0087] The above description is merely a specific method for implementing the present invention, and is not intended to limit the scope of protection of the present invention. Those skilled in the art, having understood the inventive concept of the present invention, can easily modify or replace these embodiments; therefore, all related modifications and variations of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for recommending civil law provisions based on two-stage case matching, characterized in that, The steps are as follows: Step 1: Using the legal information cited in each civil judgment case as a feature, a spectral clustering algorithm is used to cluster the cases, dividing them into groups. k A subset of cases; Step 2: For each subset of cases g The TGF-IGF algorithm was used to identify the feature words of each subset of cases. The Word2Vec algorithm is used to expand the feature words with synonyms, resulting in an expanded feature word set. ; The TGF-IGF algorithm is applied to each word in each class of cases. By simultaneously measuring the words appearing in each type of case The document frequency (TGF) and vocabulary. The final weight of a word in a particular class is obtained by using the Inverse Group Frequency (IGF) of other class cases; where, for a word... Then in the first The calculation method for similar cases is as follows: in, It is the first Vocabulary appearing in class cases The number of cases It is the first Total number of cases of the same type; words The IGF value is obtained by taking the reciprocal of the logarithm of the average percentage of cases containing the word across all other categories. The calculation method is as follows: in, The total number of case categories. Representatives except for the first The total number of case categories other than the first case category. For words appearing in a certain type of case The percentage of cases; the calculation formula for TGF-IGF is as follows: For the plaintiff's claims in the case subset g, the TGF-IGF algorithm is used to obtain the score of each word, and then the Top-N words are selected as the feature word set for this part; Step 3: Calculate the new civil dispute scenario Expand the feature word set with each subset of cases similarity Select the envoy The subset of cases with the largest value This serves as a case set for subsequent matching of similar cases; Step 4: Calculate the new civil dispute scenario and each case in the case subset. i similarity ,in accordance with Value to obtain the final set of similar cases ; Step 5: Based on similar case studies Based on the legal provisions cited in the cases, preliminary recommendations for the applicable legal provisions in new civil dispute scenarios can be obtained. ; Step 6: Use the Apriori algorithm to mine... The set of rules for citing legal provisions in China R Using association rule sets to The revisions were made to obtain the final recommended legal provisions. .

2. The method for recommending civil law provisions based on two-stage case matching as described in claim 1, characterized in that, Step 1 specifically involves: Step 1-1: Suppose that the legal information cited in a case includes two parts: legal unit information and legal provision information. Here, "legal unit" is a modification of the concept of "law." The specific definition of "legal unit" is as follows: each part and chapter of the Civil Code is considered a legal unit, and every other law as a whole is considered a legal unit; that is, for a certain law... Its corresponding legal unit for: Each legal unit consists of several legal provisions; therefore, the legal unit cited in a case is represented as follows: The cited legal provisions can be represented as ; Step 1-2: Use the Jaccard coefficient to measure different cases and Case Studies Similarity between legal units and legal provisions and ,Right now: Based on this, the overall similarity of the cases It can be represented as: in, , The coefficient is , and ; Steps 1-3: Based on Jaccard similarity, spectral clustering algorithm is used for the first stage of clustering for the entire case set A; at this stage, the differences in legal units and laws among different cases are given equal attention, therefore, the following settings are used: , ,Right now: The entire set of cases is clustered according to the process of the spectral clustering algorithm. The silhouette coefficient is used as the evaluation index of the clustering effect under different numbers of categories, and the cases are divided into k_1 first-level case subsets. Steps 1-4: For each subset of primary cases, a second-stage clustering algorithm is used; at this stage, the focus is on the legal differences between different cases, therefore, the following settings are made: , ,Right now: Following the spectral clustering algorithm, cases in the first-level case subsets are clustered. The silhouette coefficient is used as the evaluation metric for clustering performance under different numbers of categories. Each first-level case subset is divided into k_2 second-level subsets. Therefore, for the complete case set... It was divided into two stages of clustering. A subset of cases, namely: 。 3. A method for recommending civil law provisions based on two-stage case matching as described in claim 1 or 2, characterized in that, Step 2 specifically involves: Step 2-1: For each case, extract the plaintiff's claims and the facts found by the court using regular expressions, and obtain the specific text of the cited legal provisions based on the legal provision numbers cited in the case and in combination with publicly available legal provision data; Step 2-2: Use the TGF-IGF algorithm to obtain the characteristic words of the plaintiff's claims in the case subset g. ; Step 2-3: Use the TGF-IGF algorithm to obtain the legal factual feature words of the case subset g. ; Step 2-4: Use the TGF-IGF algorithm to obtain the feature words of the cited legal provisions in the case subset g. ; Steps 2-5: Merge the words from the three parts to obtain the final word set. ,Right now: Step 2-6: For the feature word set Each feature word in The Word2Vec model is used to obtain the five most similar words to the feature word, forming an expanded word set for that feature word. ,Right now: Then the expanded vocabulary set of all feature words is: 。 4. A method for recommending civil law provisions based on two-stage case matching as described in claim 1 or 2, characterized in that, Step 3 specifically involves: Step 3-1: For new civil dispute scenarios After word segmentation and stop word removal, a new contextual word set is obtained. ; Step 3-2: For each Calculate the new contextual vocabulary set With expanded feature word set The correlation; first calculate The expanded word set of each feature word in the feature word set similarity The calculation formula is as follows: in, A collection of situational vocabulary With each feature word's expanded vocabulary The total number of words in the vocabulary; Step 3-3: Based on this, calculate the contextual vocabulary set. With all feature words, the expanded vocabulary set similarity The specific calculation formula is as follows: This result constitutes a new scenario for civil disputes. and case subset The correlation, select the ones that make The subset of cases with the largest value This will serve as the dataset for subsequent similar case matching.

5. The method for recommending civil law provisions based on two-stage case matching as described in claim 3, characterized in that, Step 3 specifically involves: Step 3-1: For new civil dispute scenarios After word segmentation and stop word removal, a new contextual word set is obtained. ; Step 3-2: For each Calculate the new contextual vocabulary set With expanded feature word set The correlation; first calculate The expanded word set of each feature word in the feature word set similarity The calculation formula is as follows: in, A collection of situational vocabulary With each feature word's expanded vocabulary The total number of words in the vocabulary; Step 3-3: Based on this, calculate the contextual vocabulary set. With all feature words, the expanded vocabulary set similarity The specific calculation formula is as follows: This result constitutes a new scenario for civil disputes. and case subset The correlation, select the ones that make The subset of cases with the largest value This will serve as the dataset for subsequent similar case matching.

6. A method for recommending civil law provisions based on two-stage case matching as described in claim 1, 2, or 5, characterized in that, Step 4 specifically involves: Step 4-1: For the subset of cases Each case in The Roformer-Sim was used to measure the impact of new input scenarios on the plaintiff's litigation, the court's determination of facts, and the case. Text similarity of corresponding parts and And obtain the similarity sets respectively. and ; Step 4-2: Based on the idea of ​​the entropy weight method, calculate respectively... and Information entropy and Thus, the weights are derived. and Therefore, the case and similarity The calculation formula is: Step 4-3: Through calculation, a set of results is obtained. The set, setting the similarity threshold for two-stage matching to... , obtain Value greater than or equal to If the number of eligible cases exceeds 10, then [the list of eligible cases] will be used. The top 10 will be included in the final set of similar cases. ; If it does not exceed the limit, then all will be included.

7. A method for recommending civil law provisions based on two-stage case matching as described in claim 1, 2, or 5, characterized in that, Step 5 specifically involves: Step 5-1: ... The cases in the middle are according to Traverse in descending order: the first case is... The set of legal provisions it cites is Then The recommendation index of all legal provisions is recorded as 1, and the legal provisions are included in the recommendation set; The second traversal case is The legal provisions it cites The recommendation index is recorded as 0.

9. The legal provisions are included in the recommendation set, and the recommendation indices of the same legal provisions are superimposed. This process is repeated, with the recommendation index decreasing by 0.1 each time, but never lower than 0.

1. Step 5-2: After traversal, filter out the legal provisions with a recommendation index greater than or equal to 2; if the number of legal provisions meeting the requirements exceeds... Then choose the one with the highest recommendation index. These legal provisions form a preliminary set of recommended legal provisions. .

8. A method for recommending civil law provisions based on two-stage case matching as described in claim 1, 2, or 5, characterized in that, Step 6 specifically involves: Step 6-1: For the subset of cases The set of cited legal provisions is used to obtain all legal provision association rules that meet the requirements using an Apriori-based association rule algorithm. And construct a set of rules for associating legal provisions. ; Step 6-2: Sort the rules according to their confidence level and iterate through them. The rules in the text, if the legal provisions in the preceding clause of the rule are... In the middle, it is determined whether the legal provision following the rule is in If any legal provision in the following item is not present... In the middle, add it to In the middle; if the preceding legal provision is not present. In the case of [the specific rule], this rule is ignored; Step 6-3: After traversal is complete, perform statistics. The number of legal provisions in the text, if the number of legal provisions that meet the requirements exceeds... Then choose the one with the highest recommendation index. These legal provisions form the final set of recommended legal provisions. ; At this point, the process of recommending legal provisions has been completed.

9. A method for recommending civil law provisions based on two-stage case matching as described in any one of claims 3 or 4, characterized in that, Step 6 specifically involves: Step 6-1: For the subset of cases The set of cited legal provisions is used to obtain all legal provision association rules that meet the requirements using an Apriori-based association rule algorithm. And construct a set of rules for associating legal provisions. ; Step 6-2: Sort the rules according to their confidence level and iterate through them. The rules in the text, if the legal provisions in the preceding clause of the rule are... In the middle, it is determined whether the legal provision following the rule is in If any legal provision in the following item is not present... In the middle, add it to In the middle; if the preceding legal provision is not present. In the case of [the specific rule], this rule is ignored; Step 6-3: After traversal is complete, perform statistics. The number of legal provisions in the text, if the number of legal provisions that meet the requirements exceeds... Then choose the one with the highest recommendation index. These legal provisions form the final set of recommended legal provisions. ; At this point, the process of recommending legal provisions has been completed.

Images