Intelligent retrieval method and system for legal instruments

Abstract

The invention discloses an intelligent retrieval method and system for legal instruments, which is applied to the technical field of legal instrument retrieval, and comprises the following steps: acquiring a to-be-retrieved data text, extracting key information of the legal instruments, including first key information and second key information, and acquiring a legal application scene based on the first key information; based on the second key information, the legal instruments in the legal application scene are retrieved, the similarity between the legal instruments and all the second key information is obtained, and weighted scoring is conducted on the legal instruments by means of the first preset weight of all the second key information, so that display sorting of the legal instruments is conducted; wherein the first preset weight is obtained based on a second preset weight of the second key information; and counting effective retrieval information of the retrieval according to the click condition and the residence time of the legal document, and changing a second preset weight of the second key information for next retrieval. According to the method, the retrieval precision of the legal instruments is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of legal document retrieval, and in particular to an intelligent retrieval method and system for legal documents. Background Art

[0002] An essential step in a lawyer's case handling process is analyzing previous similar cases, including court rulings and case background information contained in legal documents, and then integrating this information into their own case to arrive at a judgment. Therefore, the ability to accurately identify relevant cases based on the case's needs is a key factor in a lawyer's ability to successfully handle a case.

[0003] At present, the commonly used method for retrieving related cases is to perform similarity matching of related legal documents based on the existing content of the case to be processed, and to display and arrange them based on the similarity. Due to the partial similarity of the content of legal documents in different legal application scenarios, and the presence of a large amount of invalid text irrelevant to the case situation in the process of directly performing similarity matching based on existing content, the final similarity matching results are less accurate. There are even huge differences in the legal application scenarios of the matched results, making it difficult for lawyers to obtain effective information from the matching results, which does not meet the lawyers' ultimate expectations for case needs.

[0004] To this end, how to provide an intelligent retrieval method and system for legal documents that can effectively improve the similarity matching accuracy between pending cases and legal documents, and effectively improve the problem of huge differences in legal application scenarios of the matching results of legal documents, and realize effective and accurate positioning of relevant cases according to case requirements is an issue that technical personnel in this field urgently need to solve. Summary of the Invention

[0005] In view of this, the present invention proposes an intelligent retrieval method and system for legal documents.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] An intelligent retrieval method for legal documents, comprising:

[0008] Step 1: Obtain the data text to be retrieved, extract key information of the legal document, including: first key information and second key information, and obtain the legal application scenario of the data text to be retrieved based on the first key information;

[0009] Step 2: Based on the second key information, search for legal documents in the legal application scenario, obtain the similarity between the legal documents and each second key information, and based on the similarity, weight the legal documents using the first preset weight of each second key information to perform a weighted score to sort the display of the legal documents; wherein the first preset weight is obtained based on the second preset weight of the second key information;

[0010] Step 3: According to the click situation and dwell time of the legal document, the effective search information of this search is counted, and the second preset weight of the second key information is changed according to the effective search information for the next search.

[0011] Optionally, in step 1, the key information of the legal document may be extracted by:

[0012] Extracting feature sequences from the data text to be retrieved, including: a character sequence of the data text to be retrieved, a part-of-speech sequence of the data text to be retrieved, and a syntactic sequence of the data text to be retrieved;

[0013] The feature sequence is input into the key information extraction model to obtain the key information of the legal document; wherein the key information extraction model is trained by inputting labeled data text into a neural network model; the labels include: character tags, part-of-speech tags and syntactic tags.

[0014] Optionally, in step 1, the labeled data text is input into the neural network model to train a key information extraction model, specifically:

[0015] The character tags, part-of-speech tags, and syntactic tags of the data text are converted into corresponding vector sequences, and feature fusion is performed to obtain a text feature vector sequence;

[0016] The text feature vector sequence is sequentially input into the deep feature extraction module and each fully connected layer in the neural network model to obtain character prediction results, part of speech prediction results, and syntax prediction results;

[0017] The prediction loss is calculated based on the character prediction results, part-of-speech prediction results, syntax prediction results, as well as character tags, part-of-speech tags, and syntax tags, and the neural network model parameters are adjusted according to the loss until the preset loss threshold is met or the maximum number of iterations is reached to obtain the key information extraction model.

[0018] Optionally, in step 1, the first key information includes but is not limited to at least one of the following key information:

[0019] Case type, cause of case, and court classification.

[0020] Optionally, in step 1, the second key information includes but is not limited to at least one of the following key information:

[0021] Plaintiff, defendant, plaintiff's law firm, defendant's law firm, plaintiff's attorney, defendant's attorney, focus of dispute, and relevant legal provisions cited.

[0022] Optionally, in step 1, the legal application scenario of the data text to be retrieved is obtained based on the first key information, specifically:

[0023] A preset legal document classification database includes: legal documents in different legal application scenarios;

[0024] Using the first key information in different legal application scenarios as input and different legal application scenarios as output, a neural network-based legal application scenario acquisition model is trained;

[0025] The first key information of the data text to be retrieved is input into the legal application scenario acquisition model to obtain the legal application scenario of the data text to be retrieved.

[0026] Optionally, in step 2, based on the similarity, the legal document is weightedly scored using the first preset weight of each second key information, as follows:

[0027]

[0028] Where S is the weighted score of the legal document; n is the number of the second key information of the legal document extracted from the data text to be retrieved; α i is the similarity between the legal document and the second key information extracted from the data to be retrieved; β i The first preset weight of the i-th second key information extracted from the data text to be retrieved.

[0029] Optionally, in step 2, the first preset weight is obtained based on the second preset weight of the second key information as follows:

[0030]

[0031] Among them, β i is the first preset weight of the i-th second key information extracted from the data text to be retrieved; γ i The second preset weight of the i-th second key information extracted from the data text to be retrieved.

[0032] Optionally, in step 3, the second preset weight of the second key information is changed according to the valid search information, specifically:

[0033] The initial second preset weight of the second key information is as follows:

[0034]

[0035] Wherein, δ is the initial second preset weight of each second key information; m is the maximum number of second key information that can be extracted from the data text to be retrieved;

[0036] The second preset weight of the second key information is changed according to the valid search information as follows:

[0037]

[0038] Among them, ε i is the second preset weight of the i-th second key information extracted from the text of the data to be searched after the effective search information is changed; j is the number of clicks on the legal document; The first preset weight of the i-th second key information extracted from the data text to be retrieved in the d-th legal document clicked during this retrieval process; t d is the residence time of the dth legal document; n is the number of the second key information of the legal document extracted from the data text to be retrieved; γ i The second preset weight of the i-th second key information extracted from the data text to be retrieved is changed for the first time ε i When γ i =δ.

[0039] The present invention also provides an intelligent retrieval system for legal documents using an intelligent retrieval method for legal documents, comprising:

[0040] Key information extraction module: used to obtain the data text to be retrieved, extract key information of the legal document, including: first key information and second key information, and obtain the legal application scenario of the data text to be retrieved based on the first key information;

[0041] Legal document scoring module: used to search for legal documents in legal application scenarios based on the second key information, obtain the similarity between the legal documents and each second key information, and based on the similarity, use the first preset weight of each second key information to weight the legal documents to sort the display of the legal documents; wherein the first preset weight is obtained based on the second preset weight of the second key information;

[0042] The second preset weight modification module is used to count the effective search information of this search based on the click situation and dwell time of the legal document, and change the second preset weight of the second key information based on the effective search information for the next search.

[0043] As can be seen from the above technical solution, compared to the prior art, the present invention provides an intelligent retrieval method and system for legal documents. By extracting key information from legal documents based on text characters, parts of speech, and text syntax, this method not only solves the problem of a large number of invalid text searches in existing related case retrieval methods, but also further improves the accuracy of key information extraction based on text characters, parts of speech, and text syntax. By first acquiring the legal application scenario of the data text to be retrieved based on the first key information, and then retrieving the legal documents under the legal application scenario based on the second key information, the retrieved legal documents are weightedly scored and displayed and sorted based on the similarity between the extracted second key information and the retrieved legal documents, in combination with the first preset weight of the extracted second key information (calculated based on the second preset weight), and the second preset weight of the extracted second key information is changed according to the click situation and the dwell time of the legal document after each retrieval. This not only effectively improves the problem of huge differences in the legal application scenarios of the matching results of the legal documents, but also improves the similarity matching accuracy between the cases to be processed and the legal documents, and realizes the effective and accurate positioning of relevant cases according to the case requirements. The method of making corresponding changes to the second preset weight of the extracted key information based on the click situation and the dwell time of the legal document after each retrieval further records the lawyer's personalized needs for the retrieval of related cases, and effectively improves the lawyer's retrieval accuracy for case requirements. In summary, the present invention effectively improves the retrieval accuracy of legal documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0045] Figure 1 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0046] 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 creative efforts are within the scope of protection of the present invention.

[0047] Example 1:

[0048] Embodiment 1 of the present invention discloses an intelligent retrieval method for legal documents, such as Figure 1 Shown, including:

[0049] Step 1: Obtain the data text to be retrieved, extract key information of the legal document, including: first key information and second key information, and obtain the legal application scenario of the data text to be retrieved based on the first key information.

[0050] The method for extracting key information from legal documents is as follows:

[0051] Extracting feature sequences from the data text to be retrieved, including: a character sequence of the data text to be retrieved, a part-of-speech sequence of the data text to be retrieved, and a syntactic sequence of the data text to be retrieved;

[0052] The feature sequence is input into the key information extraction model to obtain the key information of the legal document; wherein the key information extraction model is trained by inputting labeled data text into a neural network model; the labels include: character tags, part-of-speech tags and syntactic tags.

[0053] According to the labeled data text input into the neural network model, the key information extraction model is trained, specifically:

[0054] The character tags, part-of-speech tags, and syntactic tags of the data text are converted into corresponding vector sequences, and feature fusion is performed to obtain a text feature vector sequence;

[0055] The text feature vector sequence is sequentially input into the deep feature extraction module and each fully connected layer in the neural network model to obtain character prediction results, part of speech prediction results, and syntax prediction results;

[0056] The prediction loss is calculated based on the character prediction results, part-of-speech prediction results, syntax prediction results, as well as character tags, part-of-speech tags, and syntax tags, and the neural network model parameters are adjusted according to the loss until the preset loss threshold is met or the maximum number of iterations is reached to obtain the key information extraction model.

[0057] The first key information includes but is not limited to at least one of the following key information:

[0058] Case type, cause of case, and court classification.

[0059] The second key information includes but is not limited to at least one of the following key information:

[0060] Plaintiff, defendant, plaintiff's law firm, defendant's law firm, plaintiff's attorney, defendant's attorney, focus of dispute, and relevant legal provisions cited.

[0061] The legal application scenarios for obtaining the data text to be retrieved based on the first key information are as follows:

[0062] A preset legal document classification database includes: legal documents in different legal application scenarios;

[0063] Using the first key information in different legal application scenarios as input and different legal application scenarios as output, a neural network-based legal application scenario acquisition model is trained;

[0064] The first key information of the data text to be retrieved is input into the legal application scenario acquisition model to obtain the legal application scenario of the data text to be retrieved.

[0065] Step 2: Based on the second key information, search for legal documents in the legal application scenario, obtain the similarity between the legal documents and each second key information, and based on the similarity, use the first preset weight of each second key information to perform weighted scoring on the legal documents to display and sort the legal documents; wherein the first preset weight is obtained based on the second preset weight of the second key information.

[0066] Based on the similarity, the legal documents are weighted and scored using the first preset weights of each second key information as follows:

[0067]

[0068] Where S is the weighted score of the legal document; n is the number of the second key information of the legal document extracted from the data text to be retrieved; α i is the similarity between the legal document and the second key information extracted from the data to be retrieved; β i The first preset weight of the i-th second key information extracted from the data text to be retrieved.

[0069] The first preset weight is obtained based on the second preset weight of the second key information as follows:

[0070]

[0071] Among them, β i is the first preset weight of the i-th second key information extracted from the data text to be retrieved; γ i The second preset weight of the i-th second key information extracted from the data text to be retrieved.

[0072] Step 3: According to the click situation and dwell time of the legal document, the effective search information of this search is counted, and the second preset weight of the second key information is changed according to the effective search information for the next search.

[0073] The second preset weight of the second key information is changed according to the valid search information, specifically:

[0074] The initial second preset weight of the second key information is as follows:

[0075]

[0076] Wherein, δ is the initial second preset weight of each second key information; m is the maximum number of second key information that can be extracted from the data text to be retrieved;

[0077] The second preset weight of the second key information is changed according to the valid search information as follows:

[0078]

[0079] Among them, ε i is the second preset weight of the i-th second key information extracted from the text of the data to be searched after the effective search information is changed; j is the number of clicks on the legal document; The first preset weight of the i-th second key information extracted from the data text to be retrieved in the d-th legal document clicked during this retrieval process; t d is the residence time of the dth legal document; n is the number of the second key information of the legal document extracted from the data text to be retrieved; γ i The second preset weight of the i-th second key information extracted from the data text to be retrieved is changed for the first time ε i When γ i =δ.

[0080] Example 2:

[0081] Embodiment 2 of the present invention discloses an intelligent retrieval system for legal documents using an intelligent retrieval method for legal documents, comprising:

[0082] Key information extraction module: used to obtain the data text to be retrieved, extract key information of the legal document, including: first key information and second key information, and obtain the legal application scenario of the data text to be retrieved based on the first key information;

[0083] Legal document scoring module: used to search for legal documents in legal application scenarios based on the second key information, obtain the similarity between the legal documents and each second key information, and based on the similarity, use the first preset weight of each second key information to weight the legal documents to sort the display of the legal documents; wherein the first preset weight is obtained based on the second preset weight of the second key information;

[0084] The second preset weight modification module is used to count the effective search information of this search based on the click situation and dwell time of the legal document, and change the second preset weight of the second key information based on the effective search information for the next search.

[0085] The present invention discloses an intelligent search method and system for legal documents. By extracting key information from legal documents based on text characters, parts of speech, and syntax, this method not only addresses the problem of large amounts of invalid text retrieval in existing related case search methods, but also further improves the accuracy of key information extraction based on text characters, parts of speech, and syntax. By first acquiring the legal application scenario of the data text to be retrieved based on the first key information, and then retrieving the legal documents under the legal application scenario based on the second key information, the retrieved legal documents are weightedly scored and displayed and sorted based on the similarity between the extracted second key information and the retrieved legal documents, in combination with the first preset weight of the extracted second key information (calculated based on the second preset weight), and the second preset weight of the extracted second key information is changed according to the click situation and the dwell time of the legal document after each retrieval. This not only effectively improves the problem of huge differences in the legal application scenarios of the matching results of the legal documents, but also improves the similarity matching accuracy between the cases to be processed and the legal documents, and realizes the effective and accurate positioning of relevant cases according to the case requirements. The method of making corresponding changes to the second preset weight of the extracted key information based on the click situation and the dwell time of the legal document after each retrieval further records the lawyer's personalized needs for the retrieval of related cases, and effectively improves the lawyer's retrieval accuracy for case requirements. In summary, the present invention effectively improves the retrieval accuracy of legal documents.

[0086] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0087] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent retrieval method for legal documents, characterized in that: include: Step 1: Obtain the data text to be retrieved, extract key information of the legal document, including: first key information and second key information, and obtain the legal application scenario of the data text to be retrieved based on the first key information; Step 2: Based on the second key information, search for legal documents in the legal application scenario, obtain the similarity between the legal document and each of the second key information, and based on the similarity, weight the legal document using the first preset weight of each of the second key information to perform a weighted score on the legal document, so as to perform a display sorting of the legal document; wherein the first preset weight is obtained based on the second preset weight of the second key information; Step 3: According to the clicks and dwell time of the legal document, the effective search information of this search is counted, and the second preset weight of the second key information is changed according to the effective search information for the next search.

2. The intelligent retrieval method for legal documents according to claim 1, characterized in that: In step 1, the key information of the legal document is extracted in the following manner: Extracting a feature sequence from the data text to be retrieved, including: a character sequence of the data text to be retrieved, a part-of-speech sequence of the data text to be retrieved, and a syntactic sequence of the data text to be retrieved; The feature sequence is input into a key information extraction model to obtain key information of the legal document; wherein the key information extraction model is trained by inputting labeled data text into a neural network model; the labels include: character labels, part-of-speech labels and syntactic labels.

3. The intelligent retrieval method for legal documents according to claim 2, characterized in that: In step 1, the labeled data text is input into the neural network model to train the key information extraction model, specifically: The character tags, part-of-speech tags and syntactic tags of the data text are respectively converted into corresponding vector sequences, and feature fusion is performed to obtain a text feature vector sequence; Inputting the text feature vector sequence into the deep feature extraction module and each fully connected layer in the neural network model in sequence to obtain character prediction results, part of speech prediction results and syntax prediction results; The prediction loss is calculated based on the character prediction result, the part-of-speech prediction result, the syntactic prediction result, the character tag, the part-of-speech tag, and the syntactic tag, and the neural network model parameters are adjusted according to the loss until a preset loss threshold is met or a maximum number of iterations is reached, thereby obtaining the key information extraction model.

4. The intelligent retrieval method for legal documents according to claim 1, characterized in that: In step 1, the first key information includes but is not limited to at least one of the following key information: Case type, cause of case, and court characterization.

5. The intelligent retrieval method for legal documents according to claim 1, characterized in that: In step 1, the second key information includes but is not limited to at least one of the following key information: The plaintiff, the defendant, the plaintiff's law firm, the defendant's law firm, the plaintiff's lawyer, the defendant's lawyer, the focus of the dispute and the relevant legal provisions cited.

6. The intelligent retrieval method for legal documents according to claim 1, characterized in that: In step 1, the legal application scenario of the data text to be retrieved is obtained based on the first key information, specifically: A legal document classification database is preset, including: legal documents in different legal application scenarios; Taking the first key information in different legal application scenarios as input and different legal application scenarios as output, a neural network-based legal application scenario acquisition model is trained; The first key information of the data text to be retrieved is input into the legal application scenario acquisition model to acquire the legal application scenario of the data text to be retrieved.

7. The intelligent retrieval method for legal documents according to claim 1, characterized in that: In step 2, based on the similarity, the legal document is weighted and scored using the first preset weight of each of the second key information, as follows: Wherein, S is the weighted score of the legal document; n is the number of the second key information of the legal document extracted from the data text to be retrieved; α i is the similarity between the legal document and the i-th second key information extracted from the data text to be retrieved; β i A first preset weight for the i-th second key information extracted from the data text to be retrieved.

8. The intelligent retrieval method for legal documents according to claim 1, characterized in that: In step 2, the first preset weight is obtained based on the second preset weight of the second key information as follows: Among them, β i is the first preset weight of the i-th second key information extracted from the data text to be retrieved; γ i The second preset weight of the i-th second key information extracted from the data text to be retrieved.

9. The intelligent retrieval method for legal documents according to claim 1, characterized in that: In step 3, the second preset weight of the second key information is changed according to the valid search information, specifically: The initial second preset weight of the second key information is as follows: Wherein, δ is the initial second preset weight of each second key information; m is the maximum number of second key information that can be extracted from the data text to be retrieved; The second preset weight of the second key information is changed according to the effective search information as follows: Among them, ε i is the second preset weight of the i-th second key information extracted from the to-be-searched data text after the effective search information is changed; j is the number of clicks on the legal document; The first preset weight of the i-th second key information extracted from the data text to be searched in the d-th legal document clicked during this search process; t d is the residence time of the dth legal document; n is the number of the second key information of the legal document extracted from the data text to be retrieved; γ i The second preset weight of the i-th second key information extracted from the data text to be retrieved is i When γ i =δ.

10. An intelligent retrieval system for legal documents using the intelligent retrieval method for legal documents according to any one of claims 1 to 9, characterized in that: include: Key information extraction module: used to obtain the data text to be retrieved, extract the key information of the legal document, including: first key information and second key information, and obtain the legal application scenario of the data text to be retrieved based on the first key information; A legal document scoring module: used to search for legal documents in the legal application scenario based on the second key information, obtain the similarity between the legal document and each of the second key information, and based on the similarity, perform weighted scoring on the legal document using the first preset weight of each of the second key information, so as to perform display sorting of the legal documents; wherein the first preset weight is obtained based on the second preset weight of the second key information; The second preset weight modification module is used to count the effective search information of this search according to the click situation and dwell time of the legal document, and change the second preset weight of the second key information according to the effective search information for the next search.

Images