A legal information interpretation optimization system and method based on intelligent interaction

Abstract

The present invention discloses a legal information interpretation optimization system and method based on intelligent interaction, which relates to the technical field of text analysis, marks the clause elements in the legal clause, obtains the matching records of the same legal clause and several legal texts, marks the text feature words corresponding to the clause elements in the legal clause in the legal text, collects the corpus of the clause elements in the legal clause, forms a text feature word sequence with the text feature words in the legal text matching the legal clause according to the arrangement order of the clause elements in the legal clause, collects the association degree of the text feature words in the text feature word sequence, obtains the logical sequence of the clause elements in the legal clause, obtains the clause elements in the target clause, matches the text feature words corresponding to each clause element, and forms a feature relationship pair with the clause element and the text feature word when the difference between the association degree and the association degree of the clause element corresponding to the logical sequence of the target clause is the smallest.

Description

Technical Field

[0001] The present invention relates to the technical field of text analysis, and in particular to a legal information interpretation optimization system and method based on intelligent interaction. Background Art

[0002] As the concept of the rule of law penetrates into all aspects of social life, in order to clearly express the responsibilities, rights, and obligations of all parties, professional legal texts are needed to explain. However, for non-professionals, there is a certain threshold for reading legal texts. It takes time and energy to query legal clauses, and it is necessary to consult relevant materials to understand the legal clauses. On the basis of consulting the legal clauses, it is also necessary to judge the application of the legal clauses in combination with the actual situation reflected in the legal texts. The above steps will take a lot of time for relevant personnel to consult and compare materials, making the efficiency of reading legal texts low. Summary of the invention

[0003] The purpose of the present invention is to provide a legal information interpretation optimization system and method based on intelligent interaction to solve the problems raised in the prior art.

[0004] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a legal information interpretation optimization method based on intelligent interaction, the method comprising:

[0005] Step S100: sort out the legal clause matching records of all historical legal texts, and build a matching model between legal texts and legal clauses through a deep learning algorithm;

[0006] Step S200: marking the clause elements in the legal clause, obtaining matching records of the same legal clause and several legal texts, marking the text feature words corresponding to the clause elements in the legal clause in the legal texts, and compiling a corpus of the clause elements in the same legal clause;

[0007] Step S300: according to the arrangement order of the clause elements in the legal clause, the text feature words in the legal text matched by the legal clause are grouped into a text feature word sequence, and the associations of the text feature words in the text feature word sequence are collected to obtain a logical sequence of the clause elements in the legal clause;

[0008] Step S400: a legal text input by the user into the large language model is set as the target text, the text title of the target text is obtained through the large language model, and words that are semantically related to the title feature words in the text title are obtained from the target text, and the words are recorded as related words;

[0009] Step S500: the user adjusts the text title, obtains the related words in the adjusted text title and the target text, and generates adjustment prompt words according to the difference between the related words before and after the adjustment;

[0010] Step S600: input the target text and the adjustment prompt words into the large language model to obtain summary information of the target text, and set the legal clauses that match the target text as target clauses through the matching model of the legal text and the legal clauses;

[0011] Step S700: Obtain the clause elements in the target clause, match the text feature words corresponding to each clause element in the summary information according to the corpus of the clause elements, obtain the logical sequence of the target clause, calculate the correlation between the text feature words corresponding to the clause elements according to the logical sequence order, and when the correlation is the smallest with the correlation between the clause elements corresponding to the logical sequence of the target clause, form a feature relationship pair with the clause elements and the text feature words;

[0012] Step S800: Collect all feature relationship pairs, and annotate the clause elements in the target clause through the text feature words in the feature relationship pairs.

[0013] Further, step S200 includes:

[0014] Step S201: Obtain a matching record (text, cla) of a certain legal text and legal clause, where text represents the legal text in the matching record, and cla represents the legal clause in the matching record;

[0015] Step S202: Mark the clause elements in cla, annotate the instances corresponding to the clause elements in text, and record the words corresponding to the instances as text feature words;

[0016] Step S203: Gather several legal texts matching the legal clause cla to form a text matching set corresponding to the legal clause cla, and separately gather text feature words corresponding to each clause element in the legal clause cla to obtain a corpus corresponding to each clause element.

[0017] Furthermore, step S300 includes:

[0018] Step S301: Obtain the i-th matching legal text text from the text matching set corresponding to the legal clause cla i , get the jth clause element in the legal clause cla in the legal text text i The corresponding text feature word v j，i , the j+1th clause element in the legal text i The corresponding text feature word v j+1，i ;

[0019] Step S302: Obtain text feature word v j , i and text feature word v j+1，i The correlation between them is denoted as rij ;

[0020] Step S303: Calculate the average value c of the correlation between the text feature words of the jth clause element and the j+1th clause element in the legal clause cla j , Where m represents the number of legal texts in the text matching set corresponding to the legal clause cla;

[0021] Step S304: Arrange the average values ​​of the correlations between the text feature words corresponding to the clause elements according to the order of the clause elements in the legal clause cla, and obtain the logical sequence L of the clause elements in the legal clause cla, L: c1→c2→c3→…→c k , where k represents the total number of clause elements in the legal clause cla;

[0022] Obtain different legal texts that match the same legal clause, extract legal terms in the legal clauses, and mark the instances corresponding to the legal terms in the legal texts;

[0023] By obtaining the correlation between the text feature words corresponding to the instance, the logical distance of the logical relationship of the text feature words corresponding to the instance is reflected, and the logical distances of all text feature words in the legal text are arranged in order. The sequence of logical relationships between instances in the legal text, when consistent with the logical relationship sequence, is the text feature word of the instance corresponding to the legal term matching.

[0024] Furthermore, step S500 includes:

[0025] Step S501: Obtain the text title output by the large language model. When the user modifies a word in the text title, it is recorded as a target modification record, wherein the word is recorded as a first target word, and after the user modifies the word, the word that replaces the word is recorded as a second target word;

[0026] Step S502: from the target text, collect the related words of the first target word to form a first word set, denoted as E1, obtain the semantic association between each related word in E1 and the first target word, collect the related words of the second target word to form a second word set, denoted as E2, and obtain the semantic association between each related word in E2 and the second target word;

[0027] Step S503: Obtain a word reference set E0, E0 = E1 ∪ E2, obtain any word w in E0, and record the semantic association between w and the first target word as α w , the semantic correlation between w and the second target word is recorded as β w , set the strengthening word set H1 and the weakening word set H2;

[0028] When β w >αw When w is recorded in the intensifier set H1, when α w >β w When , record w into the weakened word set H2;

[0029] Step S504: generating strengthening prompt words according to the words in the strengthening word set H1, generating weakening prompt words according to the words in the weakening word set H2, and the adjusted prompt words include strengthening prompt words and weakening prompt words;

[0030] By obtaining the user's modification records of the title and comparing the title words selected by the user with the title words provided by the large language model, we can obtain the content that the user attaches more importance to in the legal text. We can extract the content that the user attaches more importance to and obtain the prompt words applied to the large language model. Through the prompt words, we can better cover the content in the original legal text and generate a text summary that meets the user's needs.

[0031] Furthermore, step S700 includes:

[0032] Step S701: Obtain the logical sequence Lh of the target clause, Lh: ch1→ch2→ch3→…→ch d , among which ch1, ch2, ch3, ... and ch d They represent the average values ​​of the correlations between the first, second, third, …, and dth groups of text feature words in the target clause respectively;

[0033] Step S702: Obtain the corpus LA corresponding to the qth clause element in the target clause q , match LA from summary information q The text feature words existing in the corpus are collected and recorded into the qth corpus matching set MT q , obtain the corpus LA corresponding to the q+1th clause element in the target clause q+1 , match LA from summary information q+1 The text feature words existing in the corpus are collected and recorded into the q+1th corpus matching set MT q+1 ;

[0034] Step S703: Calculate MT q With MT q+1 The correlation between any two text feature words in the text constructs the qth evaluation factor δ q , δ q =|ch q -y q |, where ch q Indicates the correlation between the qth group of text feature words in the target clause, y q Indicates MT q With MT q+1The correlation between any set of text feature words, where q≤d;

[0035] Compare the logical relationship of the text feature words in the target text with the logical sequence of the target clauses. When the difference in text relevance is the smallest, it means that the relationship of the text feature words best matches the logical sequence of the target clauses.

[0036] Step S704: Constructing a correlation evaluation function Select text feature words from the corpus matching set and calculate them separately When φ reaches the minimum value, the word sequence composed of text feature words is obtained, and the zth text feature word V in the word sequence is z and the zth clause element fac in the target clause z , forming the zth feature relationship pair (V z ,fac z ),

[0037] In order to better implement the above method, a legal information interpretation optimization system based on intelligent interaction is also proposed, which includes:

[0038] A matching model management module, a corpus management module, a logical sequence management module, a related word management module, a prompt word management module, a summary acquisition module, a feature relationship pair management module and an annotation management module, wherein the matching model management module is used to manage the matching model between the legal text and the legal clause, the corpus management module is used to manage the corpus corresponding to the clause elements, the logical sequence management module is used to manage the logical sequence corresponding to the legal clause, the related word management module is used to obtain the related words of the target text, the prompt word management module is used to generate adjustment prompt words according to the difference between the related words before and after the adjustment, the summary acquisition module is used to obtain the summary information of the target text generated by the large language model, the feature relationship pair management module is used to manage the feature relationship pairs composed of the clause elements of the target clause and the text feature words in the summary information, and the annotation management module is used to annotate the clause elements in the target clause through the text feature words in the feature relationship pairs;

[0039] Furthermore, the corpus management module includes: a matching record management unit, an annotation management unit and a corpus collection unit, wherein the matching record management unit is used to manage matching records between legal texts and legal clauses, the annotation management unit is used to manage annotation information of clause elements and text feature words in matching records, and the corpus collection unit is used to match clause elements with text feature words in the same legal clause;

[0040] Furthermore, the logic sequence management module includes: a text feature word management unit, a correlation calculation unit and a logic sequence management unit, wherein the text feature word management unit is used to manage text feature words in the legal text, the correlation calculation unit is used to calculate the correlation between text feature words, and the logic sequence management unit is used to concatenate the average values ​​of the text feature word correlations to obtain the logic sequence of the legal clauses;

[0041] Furthermore, the prompt word management module includes: an adjustment capture unit, an adjustment management unit and a prompt word association unit, wherein the adjustment capture unit is used to capture the user's adjustment record of the text title, the adjustment management unit is used to associate the adjustment prompt word, and the prompt word association unit is used to manage the prompt word;

[0042] Furthermore, the feature relationship pair management module includes: a target clause management unit, a corpus matching unit, an evaluation factor calculation unit, an evaluation function association unit and a feature relationship pair management unit, wherein the target clause management unit is used to obtain the logical sequence of the target clause, the corpus matching unit is used to match the text feature words in the target text, the evaluation factor calculation unit is used to calculate the evaluation factor, the evaluation function association unit is used to obtain the word sequence composed of the text feature words when the evaluation function is minimum, and the feature relationship pair management unit is used to manage the feature relationship pairs.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows: by obtaining the logical relationship between text feature words in the legal text, a reference of the correspondence between clause elements and instances in the legal clause is obtained. By obtaining the difference between the user's demand for the legal text and the analysis result of the large language model, a prompt word is generated to input the large language model, so that the text summary output by the large language model meets the user's needs. Instances are obtained from the text summary, and the clause elements in the legal clause are annotated, so that the user can better understand the legal clause by bringing it into the context of the legal text. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A schematic diagram of the structure of a legal information interpretation optimization system based on intelligent interaction according to the present invention;

[0045] Figure 2 This is a flow chart of the first part of a legal information interpretation optimization method based on intelligent interaction according to the present invention;

[0046] Figure 3 This is a flow chart of the second part of a legal information interpretation optimization method based on intelligent interaction according to the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0048] Example: Figure 1-Figure 3 As shown, the present invention provides a technical solution, a legal information interpretation optimization method based on intelligent interaction;

[0049] Step S100: Organize the legal clause matching records of all historical legal texts, and build a matching model between legal texts and legal clauses through a deep learning algorithm.

[0050] Step S200: marking the clause elements in the legal clause, obtaining matching records of the same legal clause and several legal texts, marking the text feature words corresponding to the clause elements in the legal clause in the legal texts, and compiling a corpus of the clause elements in the same legal clause;

[0051] In the embodiment, clause elements are words in legal clauses that limit scenarios or objects, such as parties, defendants, obligations, claims and debts. These clause elements have corresponding examples in actual legal texts.

[0052] Wherein, step S200 includes:

[0053] Step S201: Obtain a matching record (text, cla) of a certain legal text and legal clause, where text represents the legal text in the matching record, and cla represents the legal clause in the matching record;

[0054] Step S202: Mark the clause elements in cla, annotate the instances corresponding to the clause elements in text, and record the words corresponding to the instances as text feature words;

[0055] Step S203: Gather several legal texts matching the legal clause cla to form a text matching set corresponding to the legal clause cla, and separately gather text feature words corresponding to each clause element in the legal clause cla to obtain a corpus corresponding to each clause element.

[0056] Step S300: according to the arrangement order of the clause elements in the legal clause, the text feature words in the legal text matched by the legal clause are grouped into a text feature word sequence, and the associations of the text feature words in the text feature word sequence are collected to obtain a logical sequence of the clause elements in the legal clause;

[0057] Wherein, step S300 includes:

[0058] Step S301: Obtain the i-th matching legal text text from the text matching set corresponding to the legal clause cla i , get the jth clause element in the legal clause cla in the legal text text i The corresponding text feature word v j，i , the j+1th clause element in the legal text i The corresponding text feature word v j+1，i ;

[0059] Step S302: Obtain text feature word v j , i and text feature word v j+1，i The correlation between them is denoted as r ij ;

[0060] Step S303: Calculate the average value c of the correlation between the text feature words of the jth clause element and the j+1th clause element in the legal clause cla j , Where m represents the number of legal texts in the text matching set corresponding to the legal clause cla;

[0061] Step S304: Arrange the average values ​​of the correlations between the text feature words corresponding to the clause elements according to the order of the clause elements in the legal clause cla, and obtain the logical sequence L of the clause elements in the legal clause cla, L: c1→c2→c3→…→c k , where k represents the total number of clause elements in the legal clause cla.

[0062] Step S400: a legal text input by the user into the large language model is set as the target text, the text title of the target text is obtained through the large language model, words that are semantically related to the title feature words in the text title are obtained from the target text, and the words are recorded as related words.

[0063] Step S500: the user adjusts the text title, obtains the related words in the adjusted text title and the target text, and generates adjustment prompt words according to the difference between the related words before and after the adjustment;

[0064] Wherein, step S500 includes:

[0065] Step S501: Obtain the text title output by the large language model. When the user modifies a word in the text title, it is recorded as a target modification record, wherein the word is recorded as a first target word, and after the user modifies the word, the word that replaces the word is recorded as a second target word;

[0066] Step S502: from the target text, collect the related words of the first target word to form a first word set, denoted as E1, obtain the semantic association between each related word in E1 and the first target word, collect the related words of the second target word to form a second word set, denoted as E2, and obtain the semantic association between each related word in E2 and the second target word;

[0067] Step S503: Obtain a word reference set E0, E0 = E1 ∪ E2, obtain any word w in E0, and record the semantic association between w and the first target word as α w , the semantic correlation between w and the second target word is recorded as β w , set the strengthening word set H1 and the weakening word set H2;

[0068] When β w >α w When w is recorded in the intensifier set H1, when α w >β w When , record w into the weakened word set H2;

[0069] Step S504: generating strengthening prompt words according to the words in the strengthening word set H1, generating weakening prompt words according to the words in the weakening word set H2, and the adjusted prompt words include strengthening prompt words and weakening prompt words;

[0070] The prompt words reinforce the context, situation, requirement or angle that needs to be analyzed by the large language model;

[0071] Examples of generated prompts include: "Please summarize the main points of the following text", "Please analyze the following text from the perspective of the topic", "Please analyze this information based on the context of the following text", and "The content in a certain aspect is insufficient, and more details can be added for analysis."

[0072] Step S600: Input the target text and adjustment prompt words into the large language model to obtain summary information of the target text, and set the legal clauses that match the target text as target clauses through the matching model of the legal text and the legal clauses.

[0073] Step S700: Obtain the clause elements in the target clause, match the text feature words corresponding to each clause element in the summary information according to the corpus of the clause elements, obtain the logical sequence of the target clause, calculate the correlation between the text feature words corresponding to the clause elements according to the logical sequence order, and when the correlation is the smallest with the correlation between the clause elements corresponding to the logical sequence of the target clause, form a feature relationship pair with the clause elements and the text feature words;

[0074] Wherein, step S700 includes:

[0075] Step S701: Obtain the logical sequence Lh of the target clause, Lh: ch1→ch2→ch3→…→ch d , among which ch1, ch2, ch3, ... and ch d They represent the average values ​​of the correlations between the first, second, third, …, and dth groups of text feature words in the target clause respectively;

[0076] Step S702: Obtain the corpus LA corresponding to the qth clause element in the target clause q , match LA from summary information q The text feature words existing in the corpus are collected and recorded into the qth corpus matching set MT q , obtain the corpus LA corresponding to the q+1th clause element in the target clause q+1 , match LA from summary information q+1 The text feature words existing in the corpus are collected and recorded into the q+1th corpus matching set MT q+1 ;

[0077] Step S703: Calculate MT q With MT q+1 The correlation between any two text feature words in the text constructs the qth evaluation factor δ q , δ q =|ch q -y q |, where ch q Indicates the correlation between the qth group of text feature words in the target clause, y q Indicates MT q With MT q+1 The correlation between any set of text feature words, where q≤d;

[0078] Step S704: construct a correlation evaluation function φ, Select text feature words from the corpus matching set, calculate the value of φ respectively, and when φ reaches the minimum value, obtain the word sequence composed of text feature words, and replace the zth text feature word V in the word sequence with the zth text feature word V. z and the zth clause element fac in the target clause z , forming the zth feature relationship pair (V z ,fac z ).

[0079] Step S800: Collect all feature relationship pairs, and annotate the clause elements in the target clause through the text feature words in the feature relationship pairs;

[0080] For example: According to a legal provision, a party shall perform its obligations as agreed upon. If it fails to perform or performs its obligations incompletely, it shall bear the liability for breach of contract.

[0081] Note: The party (person A in the text) shall perform its obligations (examples of obligations in the text) in accordance with the agreement (examples of events in the text). If it fails to perform or fails to fully perform its obligations, it shall bear the liability for breach of contract (examples of liability for breach of contract in the text).

[0082] The system includes:

[0083] Matching model management module, corpus management module, logical sequence management module, related word management module, prompt word management module, summary acquisition module, feature relationship pair management module and annotation management module;

[0084] Among them, the matching model management module is used to manage the matching model between legal texts and legal clauses;

[0085] The corpus management module is used to manage the corpus corresponding to the clause elements, wherein the corpus management module includes: a matching record management unit, an annotation management unit and a corpus collection unit, wherein the matching record management unit is used to manage the matching records between the legal text and the legal clauses, the annotation management unit is used to manage the annotation information of the clause elements and the text feature words in the matching records, and the corpus collection unit is used to match the clause elements with the text feature words in the same legal clause;

[0086] The logic sequence management module is used to manage the logic sequence corresponding to the legal clauses, wherein the logic sequence management module includes: a text feature word management unit, a correlation calculation unit and a logic sequence management unit, wherein the text feature word management unit is used to manage the text feature words in the legal text, the correlation calculation unit is used to calculate the correlation between the text feature words, and the logic sequence management unit is used to concatenate the average values ​​of the text feature word correlations to obtain the logic sequence of the legal clauses;

[0087] Among them, the related word management module is used to obtain the related words of the target text;

[0088] The prompt word management module is used to generate adjustment prompt words according to the difference between related words before and after adjustment, and the logical sequence management module includes: a text feature word management unit, a correlation calculation unit and a logical sequence management unit, wherein the text feature word management unit is used to manage text feature words in the legal text, the correlation calculation unit is used to calculate the correlation between text feature words, and the logical sequence management unit is used to splice the average value of the text feature word correlation to obtain the logical sequence of the legal clauses;

[0089] The summary acquisition module is used to obtain summary information of the target text generated by the large language model, and the feature relationship pair management module includes: a target clause management unit, a corpus matching unit, an evaluation factor calculation unit, an evaluation function association unit and a feature relationship pair management unit, wherein the target clause management unit is used to obtain the logical sequence of the target clause, the corpus matching unit is used to match the text feature words in the target text, the evaluation factor calculation unit is used to calculate the evaluation factor, the evaluation function association unit is used to obtain the word sequence composed of the text feature words when the evaluation function is the minimum value, and the feature relationship pair management unit is used to manage the feature relationship pairs;

[0090] The feature relationship pair management module is used to manage the feature relationship pairs composed of the clause elements of the target clause and the text feature words in the summary information. The feature relationship pair management module includes: a target clause management unit, a corpus matching unit, an evaluation factor calculation unit, an evaluation function association unit and a feature relationship pair management unit. The target clause management unit is used to obtain the logical sequence of the target clause, the corpus matching unit is used to match the text feature words in the target text, the evaluation factor calculation unit is used to calculate the evaluation factor, the evaluation function association unit is used to obtain the word sequence composed of the text feature words when the evaluation function is at the minimum value, and the feature relationship pair management unit is used to manage the feature relationship pairs.

[0091] Among them, the annotation management module is used to annotate the clause elements in the target clause through the text feature words in the feature relationship pair.

[0092] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A legal information interpretation optimization method based on intelligent interaction, characterized by: The method comprises the following steps: Step S100: sort out the legal clause matching records of all historical legal texts, and build a matching model between legal texts and legal clauses through a deep learning algorithm; Step S200: marking the clause elements in the legal clause, obtaining matching records of the same legal clause and several legal texts, marking the text feature words corresponding to the clause elements in the legal clause in the legal texts, and compiling a corpus of the clause elements in the same legal clause; Step S300: according to the arrangement order of the clause elements in the legal clause, the text feature words in the legal text matched by the legal clause are grouped into a text feature word sequence, and the associations of the text feature words in the text feature word sequence are collected to obtain a logical sequence of the clause elements in the legal clause; Step S400: a legal text input by the user into the large language model is set as the target text, a text title of the target text is obtained through the large language model, words that are semantically related to the title feature words in the text title are obtained from the target text, and the words are recorded as related words; Step S500: the user adjusts the text title, obtains related words in the adjusted text title and the target text, and generates adjustment prompt words according to the difference between the related words before and after the adjustment; Step S500 includes: Step S501: obtaining a text title output by a large language model, and recording a target modification record after a user modifies a certain word in the text title, wherein the certain word is recorded as a first target word, and after the user modifies the certain word, the word that replaces the certain word is recorded as a second target word; Step S502: from the target text, collect the related words of the first target word to form a first word set, denoted as E1, obtain the semantic association between each related word in E1 and the first target word, collect the related words of the second target word to form a second word set, denoted as E2, and obtain the semantic association between each related word in E2 and the second target word; Step S503: Obtain a word reference set E0, E0 = E1 ∪ E2, obtain any word w in E0, and record the semantic association between w and the first target word as α w , the semantic relevance between w and the second target word is denoted as β w , set the strengthening word set H1 and the weakening word set H2; When β w >α w When w is recorded in the intensifier set H1, when α w >β w When , record w into the weakened word set H2; Step S504: generating strengthening prompt words according to the words in the strengthening word set H1, and generating weakening prompt words according to the words in the weakening word set H2, wherein the adjustment prompt words include strengthening prompt words and weakening prompt words; Step S600: input the target text and the adjustment prompt words into the large language model to obtain summary information of the target text, and set the legal clauses that match the target text as target clauses through the matching model of the legal text and the legal clauses; Step S700: Obtain the clause elements in the target clause, match the text feature words corresponding to each clause element in the summary information according to the corpus of the clause elements, obtain the logical sequence of the target clause, calculate the correlation between the text feature words corresponding to the clause elements according to the logical sequence order, and when the difference between the correlation and the correlation of the clause elements corresponding to the logical sequence of the target clause is the smallest, form a feature relationship pair between the clause elements and the text feature words; Step S800: Collect all feature relationship pairs, and annotate the clause elements in the target clause through the text feature words in the feature relationship pairs.

2. According to claim 1, a legal information interpretation optimization method based on intelligent interaction is characterized by: Step S200 includes: Step S201: obtaining a matching record (text, cla) of a certain legal text and a legal clause, wherein text represents the legal text in the matching record, and cla represents the legal clause in the matching record; Step S202: Mark the clause elements in cla, annotate the instances corresponding to the clause elements in text, and record the words corresponding to the instances as text feature words; Step S203: Gather several legal texts matching the legal clause cla to form a text matching set corresponding to the legal clause cla, and separately gather text feature words corresponding to each clause element in the legal clause cla to obtain a corpus corresponding to each clause element.

3. The method for optimizing legal information interpretation based on intelligent interaction according to claim 2 is characterized in that: Step S300 includes: Step S301: Obtain the i-th matching legal text text from the text matching set corresponding to the legal clause cla i , get the jth clause element in the legal clause cla in the legal text text i The corresponding text feature word v j，i , the j+1th clause element in the legal text i The corresponding text feature word v j+1，i ; Step S302: Obtain text feature word v j , i and text feature word v j+1，i The correlation between them is denoted as r ij ; Step S303: Calculate the average value c of the correlation between the text feature words of the jth clause element and the j+1th clause element in the legal clause cla j , Where m represents the number of legal texts in the text matching set corresponding to the legal clause cla; Step S304: Arrange the average values ​​of the correlations between the text feature words corresponding to the clause elements according to the order of the clause elements in the legal clause cla, and obtain the logical sequence L of the clause elements in the legal clause cla, L: c1→c2→c3→…→c k , where k represents the total number of clause elements in the legal clause cla.

4. The method for optimizing legal information interpretation based on intelligent interaction according to claim 3 is characterized by: Step S700 includes: Step S701: Obtain the logical sequence Lh of the target clause, Lh: ch1→ch2→ch3→…→ch d , among which ch1, ch2, ch3, ... and ch d They represent the average values ​​of the correlation between the first, second, third, … and dth groups of text feature words in the target clause respectively; Step S702: Obtain the corpus LA corresponding to the qth clause element in the target clause q , match LA from the summary information q The text feature words existing in the qth corpus matching set MT are collected and recorded into the qth corpus matching set MT q , obtain the corpus LA corresponding to the q+1th clause element in the target clause q+1 , match LA from the summary information q+1 The text feature words existing in the corpus are collected and recorded into the q+1th corpus matching set MT q+1 ; Step S703: Calculate MT q With MT q+1 The correlation between any two text feature words in the text constructs the qth evaluation factor δ q , δ q =|ch q -y q |, where ch q Indicates the correlation between the qth group of text feature words in the target clause, y q Indicates MT q With MT q+1 The correlation between any set of text feature words, where q≤d; Step S704: Constructing a correlation evaluation function Select text feature words from the corpus matching set and calculate them separately The value of When the minimum value is obtained, the word sequence composed of text feature words is obtained, and the zth text feature word V in the word sequence is z and the zth clause element fac in the target clause z , forming the zth feature relationship pair (V z ,fac z ).

5. A legal information interpretation optimization system based on intelligent interaction, used to execute a legal information interpretation optimization method based on intelligent interaction as described in any one of claims 1 to 4, characterized in that: The system includes: A matching model management module, a corpus management module, a logical sequence management module, a related word management module, a prompt word management module, a summary acquisition module, a feature relationship pair management module and an annotation management module, wherein the matching model management module is used to manage the matching model between the legal text and the legal clause, the corpus management module is used to manage the corpus corresponding to the clause elements, the logical sequence management module is used to manage the logical sequence corresponding to the legal clause, the related word management module is used to obtain the related words of the target text, the prompt word management module is used to generate adjustment prompt words according to the difference between the related words before and after the adjustment, the summary acquisition module is used to obtain the summary information of the target text generated by the large language model, the feature relationship pair management module is used to manage the feature relationship pairs composed of the clause elements of the target clause and the text feature words in the summary information, and the annotation management module is used to annotate the clause elements in the target clause through the text feature words in the feature relationship pairs.

6. The legal information interpretation optimization system based on intelligent interaction according to claim 5 is characterized by: The corpus management module includes: a matching record management unit, an annotation management unit and a corpus collection unit, wherein the matching record management unit is used to manage the matching records of legal texts and legal clauses, the annotation management unit is used to manage the annotation information of clause elements and text feature words in the matching records, and the corpus collection unit is used to match the clause elements with the text feature words in the same legal clause.

7. The legal information interpretation optimization system based on intelligent interaction according to claim 5 is characterized by: The logical sequence management module includes: a text feature word management unit, a correlation calculation unit and a logical sequence management unit, wherein the text feature word management unit is used to manage text feature words in the legal text, the correlation calculation unit is used to calculate the correlation between text feature words, and the logical sequence management unit is used to concatenate the average values ​​of the text feature word correlations to obtain the logical sequence of the legal clauses.

8. The legal information interpretation optimization system based on intelligent interaction according to claim 5 is characterized by: The prompt word management module includes: an adjustment capture unit, an adjustment management unit and a prompt word association unit, wherein the adjustment capture unit is used to capture the user's adjustment record of the text title, the adjustment management unit is used to associate the adjustment prompt words, and the prompt word association unit is used to manage the prompt words.

9. The legal information interpretation optimization system based on intelligent interaction according to claim 5 is characterized by: Feature Relationship The pair management module includes: a target clause management unit, a corpus matching unit, an evaluation factor calculation unit, an evaluation function association unit and a feature relationship pair management unit, wherein the target clause management unit is used to obtain the logical sequence of the target clause, the corpus matching unit is used to match the text feature words in the target text, the evaluation factor calculation unit is used to calculate the evaluation factor, the evaluation function association unit is used to obtain the word sequence composed of the text feature words when the evaluation function is at the minimum value, and the feature relationship pair management unit is used to manage the feature relationship pairs.

Images