A legal label system-based document labeling method
By preprocessing and converting the legal tag database and combining it with a tree diagram for legal document annotation, the problems of low annotation efficiency and misjudgment are solved, and efficient and accurate legal tag annotation and difficulty level generation are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING ZHIYUN XINGHE INFORMATION TECH CO LTD
- Filing Date
- 2023-02-10
- Publication Date
- 2026-06-05
AI Technical Summary
Existing technologies for labeling legal documents are slow and prone to misinterpretation.
By preprocessing the legal tag database, merging and analyzing to identify common tags, performing numerical conversion, generating a numerical conversion table and a comparison tree diagram, processing legal documents using the numerical conversion table and color annotation, adding annotations based on the comparison tree diagram, and determining the difficulty of the annotated content.
It improves the efficiency and accuracy of legal labeling, enabling rapid labeling processing and the generation of difficulty levels for easier subsequent interpretation.
Smart Images
Figure CN116227445B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of document annotation technology, specifically a document annotation method based on a legal labeling system. Background Technology
[0002] Legally labeled documents are documents used by judicial and administrative organs, parties involved, lawyers, etc., in resolving litigation and non-litigation cases. They also include non-standard documents from judicial organs, including both standard and non-standard documents.
[0003] Patent publication number CN108334500B discloses a method and apparatus for annotating judicial documents based on machine learning algorithms. The method includes: collecting a text set of judicial documents to be annotated; performing structural segmentation on the text in the text set; establishing a semantic tag library; manually annotating the judicial documents to be annotated based on the semantic tag library; selecting a portion of the manually annotated judicial documents as a standard data set for machine learning to train and optimize a preliminary annotation model; selecting the remaining manually annotated judicial document samples as a verification data set to improve the semantic tag library and iterate and optimize the preliminary annotation model to obtain a mature judicial document annotation model; and inputting the target judicial document to be annotated, after structural segmentation, into the mature judicial document annotation model to obtain the annotation result. This invention solves the problems of incomplete extraction of legal elements and low accuracy of case information extraction in related technologies.
[0004] When legal documents are labeled with legal tags, they are generally analyzed based on the legal tags existing in the database. The legal documents are then identified and labeled one by one according to the existing legal tags. However, this identification and labeling method is relatively slow in practice. At the same time, due to the complexity of the text, misidentification can easily occur during the identification process. Summary of the Invention
[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a document annotation method based on a legal labeling system to solve the technical problems of slow efficiency and easy mislabeling of legal labels in the overall document.
[0006] To achieve the above objectives, according to an embodiment of the first aspect of the present invention, a document annotation method based on a legal labeling system is proposed, comprising the following steps:
[0007] S1. Preprocess the legal tag database, merge and analyze several groups of legal tags, identify the common tags in several groups of legal tags, perform numerical conversion on such common tags, and obtain the numerical conversion table of the corresponding common tags. At the same time, according to the layer-by-layer screening method, obtain the comparison tree diagram of this legal tag database.
[0008] S2. Extract the legal documents that need to be annotated, extract the processed numerical conversion table, and convert the common tags inside the legal documents using the numerical conversion table. After the conversion is completed, the document to be processed is obtained.
[0009] S3. The document to be processed is annotated. Based on the comparison tree diagram obtained after processing, the whole sentences with color annotation in the document to be processed are extracted, and then the extracted sentences with color annotation are uniformly annotated.
[0010] S4. Analyze the content marked in the legal label text, determine the difficulty of different legal label texts by the number of times the marked area exists and the corresponding area area, and give the difficulty level.
[0011] Preferably, in step S1, the specific method for merging and analyzing several groups of legal tags is as follows:
[0012] S11. Extract all legal tags from the legal tag database and extract common tags from all legal tags. Then, use conversion values to convert the common tags. The conversion values are represented by i, where i = 1, 2, ..., n. When i is 11, the value of i is 1-1. After processing the common tags of the first category, bind the replaced conversion value i with the corresponding common tag to obtain the value conversion table.
[0013] S12. After the common tags of the first category are extracted and processed, the common tags of the processed legal tags are extracted again. By using a layer-by-layer filtering method, a comparison tree diagram belonging to this legal tag database is obtained. The common tags of the second category and subsequent categories do not need to be converted by conversion values.
[0014] S13. After processing all the legal tags in steps S11 and S12, store the corresponding numerical conversion table and the comparison tree diagram obtained from the analysis.
[0015] Preferably, in step S2, the specific method for converting the common tags within the legal document using a numerical conversion table is as follows:
[0016] S21. Extract the corresponding characters existing in the legal document through the common tags inside the numerical conversion table. After extraction, fill the extraction position with the conversion values belonging to the common tags from the numerical conversion table.
[0017] S22. After the conversion values are filled in, the filled conversion values are color-coded. The colors used for color coding are uniform, and the specific colors of the uniform color coding are determined by the operator. The legal documents with color coding are then set as documents to be processed.
[0018] Preferably, in step S3, the specific steps for annotating the document to be processed are as follows:
[0019] S31. Lock the area marked with color in the document to be processed, extend forward and backward according to the locked area until the corresponding punctuation mark is found, and extract the whole sentence with color mark.
[0020] S32. Organize several sentences, pre-determine the branches of the tree diagram to be compared based on the conversion values, and then compare the locked color-coded area with the legal labels of the branches in the tree diagram to be compared. During the comparison process, the color-coded area can be extended forward and backward to determine the legal labels existing in the corresponding sentences, and color-code the determined legal labels. At this time, the determined color labels are marked with another color scheme. Among them, the common labels of the first category use a unified color scheme, the common labels of the second category use a unified color scheme, and the legal labels of each group of different categories are marked with different unified color schemes. The different color schemes are all determined by the operator.
[0021] S33. Fill the original unprocessed document with the corrected sentences. After the unprocessed document is annotated, you will get the legal label annotation text.
[0022] Preferably, in step S4, the specific method for determining the difficulty of different legal label texts is as follows:
[0023] S41. Extract the number of color matchings and the area of the color matching area of the first category of common labels from the legal label annotation text, mark the corresponding number of color matchings as CS1, and mark the corresponding total area as MJ1. Use ND1 = CS1 × C1 + MJ1 × C2 to obtain the difficulty assessment value ND1 of the first category of common labels, where C1 and C2 are preset fixed coefficient factors.
[0024] S42. Then, process the color area of the shared label in the second category in the same way as in step S41 to obtain the difficulty assessment value ND2 of the shared label in the second category. Then, process the difficulty assessment values ND2 of subsequent labels belonging to the same category in turn. k The data is obtained, and the difficulty assessment values of k groups are summed to obtain the parameter DHD to be verified.
[0025] S43. Compare the parameter DHD to be checked with the preset parameters Y1 and Y2. The preset parameters Y1 and Y2 are preset values. Y1 < Y2. When DHD < Y1, mark the corresponding legal label text as simple text. When Y1 ≤ DHD < Y2, mark the corresponding legal label text as regular text. When Y2 ≤ DHD, mark the corresponding legal label text as difficult text.
[0026] S44. Transmit the legal label text that has been determined in terms of difficulty to the external display terminal, and restore the common labels of the first category according to the numerical conversion table for external personnel to review.
[0027] Compared with the prior art, the beneficial effects of the present invention are as follows: by merging and analyzing several groups of legal tags, identifying the common tags existing in several groups of legal tags, performing numerical conversion on such common tags, and obtaining the corresponding numerical conversion table of common tags, the common tags inside the legal document are converted by using the numerical conversion table. After the conversion is completed, the document to be processed is obtained. The document to be processed is annotated. Through the comparison tree diagram obtained by the processing, several sentences with color annotations are uniformly annotated. The common tags of the first category are converted by numerical conversion in advance, and the common tags of the original text are converted and color-coded at the same time. In the subsequent operation, the corresponding common tags can be quickly found. Subsequently, the whole sentence with common tags is directly extracted, the legal tags existing in the sentence are analyzed and annotated, which can ensure the accuracy of legal tag annotation, as well as the efficiency of legal tag annotation, quickly complete the annotation process, and improve timeliness.
[0028] The difficulty level of the annotated text will then be determined by the existing color-coded indicators, and a difficulty level will be generated and linked to the corresponding annotated text. This will allow the relevant personnel to understand the difficulty level of the legal text in advance and make appropriate interpretation plans. Attached Figure Description
[0029] Figure 1 This is a schematic diagram of the method flow of the present invention;
[0030] Figure 2 This is a tree-like classification diagram of common identifiers in this invention. Detailed Implementation
[0031] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0032] Please see Figure 1 This application provides a document annotation method based on a legal labeling system, comprising the following steps:
[0033] S1. Preprocess the legal tag database by merging and analyzing several groups of legal tags to identify common tags among them. Perform numerical conversion on these common tags and obtain a corresponding numerical conversion table. Simultaneously, using a hierarchical common tag filtering method, obtain a comparison tree diagram of the legal tag database. The specific method for merging and analyzing is as follows:
[0034] S11. Extract all legal tags from the legal tag database and extract common tags from all legal tags. Then, convert the common tags using conversion values. The conversion values are represented by i, where i = 1, 2, ..., n. When i is 11, to avoid subsequent identification errors, the value of i is represented as 1-1. 1-1 is used to distinguish it from common tags with i value of 1. If it is 11 (without a hyphen), it is easy to be judged as two common tags with i value of 1. After processing the common tags of the first category, bind the replaced conversion value i with the corresponding common tag to obtain a value conversion table (this value conversion table will be used in the subsequent document annotation process).
[0035] S12. After extracting and processing the common tags of the first category, perform another common tag extraction process on the processed legal tags. Using a layer-by-layer filtering method, obtain a comparison tree diagram belonging to this legal tag database. The common tags of the second category and subsequent categories do not need to be converted using conversion values (e.g., if the three sets of legal tags are: "Advertisement Article", "Example of Legal Article", and "Final Chapter of Legal Article", then the common tag of the first category is "Article (i)", and the common tag of the second category is "Legal Article"). Parameter Figure 2 There may be shared labels for a third category or a fourth category later, which will not be elaborated here. Only the shared labels for the first category have been numerically converted, and the conversion value i is used to convert them.
[0036] S13. After processing all the legal tags in steps S11 and S12, store the corresponding numerical conversion table and the comparison tree diagram obtained from the analysis.
[0037] S2. Extract the legal documents that need to be annotated, extract the processed numerical conversion table, and convert the common tags within the legal documents using the numerical conversion table. After conversion, the document to be processed is obtained. The specific conversion method is as follows:
[0038] S21. Extract the corresponding characters existing in the legal document through the common tags inside the numerical conversion table. After extraction, fill the extraction position with the conversion value belonging to the common tag from the numerical conversion table.
[0039] S22. After the conversion values are filled in, the converted values are marked with colors (without color marking, it is easy to cause confusion between the converted values and the original values existing in the legal documents). The colors used for color marking are uniform, and the specific colors of the uniform color scheme are determined by the operator. The legal documents marked with colors are set as documents to be processed.
[0040] S3. The document to be processed is annotated. Using the resulting comparison tree diagram, sentences with color annotations are extracted from the document (a document consists of several paragraphs, and each paragraph consists of several sentences). The extracted sentences with color annotations are then uniformly annotated. The specific annotation process is as follows:
[0041] S31. Lock the area marked with color in the document to be processed, extend forward and backward according to the locked area until the corresponding punctuation mark is found, and extract the whole sentence with color mark.
[0042] S32. Organize several sentences, pre-determine the branches of the tree diagram to be compared based on the conversion values, and then compare the locked color-coded area with the legal labels of the branches in the tree diagram to be compared. During the comparison process, the color-coded area can be extended forward and backward to determine the legal labels existing in the corresponding sentences, and color-code the determined legal labels. At this time, the determined color labels are marked with another color scheme. Among them, the common labels of the first category use a unified color scheme, the common labels of the second category use a unified color scheme, and the legal labels of each group of different categories are marked with different unified color schemes. The different color schemes are all determined by the operator.
[0043] S33. Fill in the original positions of the documents to be processed with the sorted sentences. After the documents to be processed are marked, the legal label text is obtained.
[0044] S4. Analyze the content marked within the legal label text. Assess the difficulty of different legal label texts by the frequency of the marked areas and their corresponding area sizes, and assign a difficulty level. The specific method for difficulty assessment is as follows:
[0045] S41. Extract the number of color matchings and the area of the color matching area of the first category of common labels from the legal label annotation text, mark the corresponding number of color matchings as CS1, and mark the corresponding total area as MJ1. Use ND1 = CS1 × C1 + MJ1 × C2 to obtain the difficulty assessment value ND1 of the first category of common labels, where C1 and C2 are preset fixed coefficient factors, and the specific values of C1 and C2 are determined by the operator based on experience.
[0046] S42. Then, process the color area of the shared label in the second category in the same way as in step S41 to obtain the difficulty assessment value ND2 of the shared label in the second category. Then, process the difficulty assessment values ND2 of subsequent labels belonging to the same category in turn. k The data is obtained, and the difficulty assessment values of k groups are summed to obtain the parameter DHD to be verified.
[0047] S43. Compare the parameter DHD to be checked with the preset parameters Y1 and Y2. The preset parameters Y1 and Y2 are preset values, and the specific values of Y1 and Y2 are determined by the operator. When Y1 < Y2, the corresponding legal label text is marked as simple text when DHD < Y1, the corresponding legal label text is marked as regular text when Y1 ≤ DHD < Y2, and the corresponding legal label text is marked as difficult text when Y2 ≤ DHD.
[0048] S44. Transmit the legal label text that has been determined in terms of difficulty to the external display terminal, and restore the common labels of the first category according to the numerical conversion table for external personnel to review.
[0049] The data in the above formula are all calculated by removing the dimensions and taking the numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.
[0050] The working principle of this invention is as follows: Several sets of legal tags are merged and analyzed to identify common tags among them. These common tags are then numerically converted, and a corresponding numerical conversion table is obtained. Simultaneously, a comparison tree diagram of the legal tag database is obtained through a layered common tag filtering process. Legal documents requiring annotation are extracted, and the processed numerical conversion table is also extracted. The common tags within these legal documents are then converted using the numerical conversion table. After conversion, the documents to be processed are obtained. These documents are then annotated, and the comparison tree diagram is used to identify the relevant legal documents. The process involves extracting entire sentences with color annotations from documents, then uniformly annotating several of these sentences. First, the shared tags in the first category are converted numerically, and simultaneously, the shared tags in the original text are converted and color-coded. In subsequent operations, the corresponding shared tags can be quickly located. Then, the entire sentences with shared tags are directly extracted, and the legal tags within those sentences are analyzed and annotated. This ensures both the accuracy and efficiency of legal tag annotation, enabling rapid completion of the annotation process and improving timeliness.
[0051] The difficulty level of the annotated text will then be determined by the existing color-coded indicators, and a difficulty level will be generated and linked to the corresponding annotated text. This will allow the relevant personnel to understand the difficulty level of the legal text in advance and make appropriate interpretation plans.
[0052] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. A document annotation method based on a legal labeling system, characterized in that, Includes the following steps: S1. Preprocess the legal tag database by merging and analyzing several groups of legal tags to identify common tags among them. Perform numerical conversion on these common tags and obtain a corresponding numerical conversion table. Simultaneously, using a layer-by-layer filtering method, obtain a comparison tree diagram of the legal tag database. Specifically: S11. Extract all legal tags from the legal tag database and extract common tags from all legal tags. Then, use conversion values to convert the common tags. The conversion values are represented by i, where i = 1, 2, ..., n. When i is 11, the value of i is 1-1. After processing the common tags of the first category, bind the replaced conversion value i with the corresponding common tag to obtain the value conversion table. S12. After the common tags of the first category are extracted and processed, the common tags of the processed legal tags are extracted again. By using a layer-by-layer filtering method, a comparison tree diagram belonging to this legal tag database is obtained. The common tags of the second category and subsequent categories do not need to be converted by conversion values. S13. After processing all the legal tags in steps S11 and S12, store the corresponding numerical conversion table and the comparison tree diagram obtained from the analysis. S2. Extract the legal documents that need to be annotated, extract the processed numerical conversion table, and convert the common tags within the legal documents using the numerical conversion table. After conversion, the document to be processed is obtained. The specific method is as follows: S21. Extract the corresponding characters existing in the legal document through the common tags inside the numerical conversion table. After extraction, fill the extraction position with the conversion values belonging to the common tags from the numerical conversion table. S22. After the conversion values are filled in, the filled conversion values are color-coded. The colors used for color coding are uniform, and the specific colors of the uniform color coding are determined by the operator. The legal documents with color coding are set as documents to be processed. S3. The document to be processed is annotated. Based on the comparison tree diagram obtained after processing, the whole sentences with color annotation in the document to be processed are extracted, and then the extracted sentences with color annotation are uniformly annotated. S4. Analyze the content marked within the legal label text. Assess the difficulty of different legal label texts by the frequency of the marked areas and their corresponding area sizes, and assign a difficulty level. The specific method is as follows: S41. Extract the number of color combinations and the area of the color-matching regions in the first category of common labels from the legal label annotation text, and label the corresponding number of color combinations as CS1 and the corresponding total area as MJ1. The difficulty assessment value ND1 for the common label of the first category is obtained, where C1 and C2 are both preset fixed coefficient factors; S42. Then, process the color area of the shared label in the second category in the same way as in step S41 to obtain the difficulty assessment value ND2 of the shared label in the second category. Then, process the difficulty assessment values ND2 of subsequent labels belonging to the same category in turn. k The data is obtained, and the difficulty assessment values of k groups are summed to obtain the parameter DHD to be verified. S43. Compare the parameter DHD to be checked with the preset parameters Y1 and Y2. The preset parameters Y1 and Y2 are preset values. Y1 < Y2. When DHD < Y1, mark the corresponding legal label text as simple text. When Y1 ≤ DHD < Y2, mark the corresponding legal label text as regular text. When Y2 ≤ DHD, mark the corresponding legal label text as difficult text. S44. Transmit the legal label text that has been determined in terms of difficulty to the external display terminal, and restore the common labels of the first category according to the numerical conversion table for external personnel to review.
2. The document annotation method based on a legal labeling system according to claim 1, characterized in that, In step S3, the specific steps for annotating the document to be processed are as follows: S31. Lock the area marked with color in the document to be processed, extend forward and backward according to the locked area until the corresponding punctuation mark is found, and extract the whole sentence with color mark. S32. Organize several sentences, pre-determine the branches of the tree diagram to be compared based on the conversion values, and then compare the locked color-coded area with the legal labels of the branches in the tree diagram to be compared. During the comparison process, the color-coded area can be extended forward and backward to determine the legal labels existing in the corresponding sentences, and color-code the determined legal labels. At this time, the determined color labels are marked with another color scheme. Among them, the common labels of the first category use a unified color scheme, the common labels of the second category use a unified color scheme, and the legal labels of each group of different categories are marked with different unified color schemes. The different color schemes are all determined by the operator. S33. Fill the original unprocessed document with the corrected sentences. After the unprocessed document is annotated, you will get the legal label annotation text.