WO2021057202A1 - 一种判决结果处理方法及装置 - Google Patents

一种判决结果处理方法及装置 Download PDF

Info

Publication number
WO2021057202A1
WO2021057202A1 PCT/CN2020/102026 CN2020102026W WO2021057202A1 WO 2021057202 A1 WO2021057202 A1 WO 2021057202A1 CN 2020102026 W CN2020102026 W CN 2020102026W WO 2021057202 A1 WO2021057202 A1 WO 2021057202A1
Authority
WO
WIPO (PCT)
Prior art keywords
legal
document
judgment result
processed
text content
Prior art date
Application number
PCT/CN2020/102026
Other languages
English (en)
French (fr)
Inventor
曾祥辉
冯鸳鹤
Original Assignee
北京国双科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 北京国双科技有限公司 filed Critical 北京国双科技有限公司
Publication of WO2021057202A1 publication Critical patent/WO2021057202A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/18—Legal services

Definitions

  • the present invention relates to the field of machine learning technology, and more specifically, to a method and device for processing judgment results.
  • the judgments of judicial cases are mainly led by case reviewers.
  • the case reviewers obtain the judgment results of the judicial cases they hear based on the expert rules obtained in advance.
  • the process can be: obtain judicial experts and summarize from the provisions of laws and regulations. And associate the legal elements with the corresponding laws and regulations to obtain the expert rules of the corresponding relationship between the legal elements and the laws and regulations.
  • obtain the case reviewer for the pending documents obtain the laws and regulations corresponding to the legal elements from the expert rules, and then obtain the judgment results of the documents to be processed based on the laws and regulations corresponding to the legal elements.
  • the legal element is a summary of a law and regulation. Taking the crime of causing traffic accidents as an example, the corresponding laws and regulations are: violation of traffic and transportation management regulations, resulting in a major accident, causing serious injury or death, or causing major losses to public and private property , If sentenced to fixed-term imprisonment or criminal detention of less than three years, then the legal elements of the laws and regulations can be: at least one of causing serious injury, causing death, and causing major losses. These legal elements are used to summarize the laws and regulations.
  • the present invention provides a method and device for processing judgment results that overcome the above-mentioned problems or at least partially solve the above-mentioned problems, which are used to compare the respective judgment results corresponding to the documents to be processed with the text content corresponding to the respective legal elements.
  • the technical solutions are as follows:
  • the present invention provides a method for processing judgment results, the method including:
  • the obtaining the text content corresponding to each of the legal elements in the document to be processed includes:
  • the correlation degree that satisfies the first preset condition is selected from the correlation between each word and the legal element, and the correlation degree is based on the correlation degree that satisfies the first preset condition.
  • the word to which the degree of relevance belongs is obtained from the document to be processed, and the text content corresponding to the legal element is obtained.
  • the obtaining the text content corresponding to each of the legal elements in the document to be processed includes:
  • each of the legal elements from the preset related word set, and the words in the preset related word set are based on the known judgment results in each first legal document corresponding to each first legal document
  • the correlation degree of each legal element and/or the number of co-occurrences of each word in each first legal document is obtained, each legal element corresponding to each first legal document, and each word in each first legal document and each first legal document
  • the correlation degree of each legal element corresponding to the document is obtained based on the element recognition model, which takes each second legal document with known judgment results as input, and will label each legal element and each second legal document as input.
  • the correlation between each word in each second legal document and each labeled legal element is obtained as output training;
  • the method further includes:
  • the determination of the legal element corresponding to each judgment result from the at least one legal element includes:
  • the correlation degree that satisfies the second preset condition For any one of the judgment results, select the correlation degree that satisfies the second preset condition from the correlation degree of each legal element with the judgment result, and calculate the correlation degree that satisfies the second preset condition.
  • the element of the legal article to which it belongs is determined as the element of the legal article corresponding to the judgment result.
  • the obtaining at least one judgment result corresponding to the document to be judged based on at least one legal element corresponding to the document to be judged includes:
  • At least one legal element corresponding to the document to be judged is used as the input of the judgment result prediction model, and at least one judgment result corresponding to the document to be judged output by the judgment result prediction model is obtained, and the judgment result prediction model also outputs The degree of relevance of each legal element to each judgment result;
  • the judgment result prediction model takes each legal element corresponding to each legal document for which the judgment result is known as input, and uses the judgment result corresponding to each legal document and the correlation between each legal element and each judgment result as output training to obtain .
  • the method further includes: marking the text content associated with the legal element in the document to be processed.
  • the text content associated with each legal element corresponds to a text relevance
  • the special marking of the text content associated with the legal element in the document to be processed includes: obtaining a special marking form corresponding to the text content based on the text relevance of the text content associated with the legal element, and The special mark form corresponding to the text content makes a special mark on the text content.
  • the present invention also provides a judgment result processing device, which includes:
  • the first obtaining unit is configured to obtain at least one legal element corresponding to the document to be processed, and to obtain the text content corresponding to each legal element in the document to be processed, and the document to be processed is the document of the case to be judged ;
  • the second obtaining unit is configured to obtain at least one judgment result corresponding to the document to be processed based on at least one legal element corresponding to the document to be processed, and to determine the corresponding judgment result from the at least one legal element Legal elements
  • the association unit is used for associating each judgment result with the text content corresponding to the corresponding legal element, so that each judgment result corresponds to the text content.
  • the present invention also provides an electronic device that includes at least one processor, and at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus
  • the processor is used to call the program instructions in the memory to execute the above judgment result processing method.
  • the present invention also provides a computer readable medium on which a computer program is stored, and when the program is executed by a processor, the above judgment result processing method is realized.
  • the judgment result processing method and device can obtain at least one legal element corresponding to the document to be processed and the text content corresponding to each legal element in the document to be processed for a judgment document, based on At least one legal element corresponding to the document to be processed is obtained, at least one judgment result corresponding to the document to be processed is obtained, and the legal element corresponding to each judgment result is determined, and then the text content corresponding to each judgment result and the corresponding legal element is performed Correlation, thereby establishing the correspondence between each judgment result corresponding to the document to be processed and the text content in the document to be processed, so that the judgment result corresponds to the text content, then for any judgment result, it is possible to know which part of the document to be processed is due to
  • the text content leads to the judgment result, so that the text content corresponding to the judgment result is used to interpret the judgment result to know the reason for the judgment result, so that not only the judgment result corresponding to the document to be processed can be obtained, but also the judgment result can be obtained.
  • the obtaining of the judgment result and the association between the judgment result and the text content in this embodiment are based on the intermediate reference of the legal element, which improves the accuracy compared with obtaining the judgment result from the entire content of the document to be processed.
  • Fig. 1 shows a flowchart of a method for processing a judgment result provided by an exemplary embodiment of the present disclosure
  • FIG. 2 shows a schematic diagram of a model architecture of an element recognition model provided by an exemplary embodiment of the present disclosure
  • Fig. 3 shows a flowchart of another method for processing a judgment result provided by an exemplary embodiment of the present disclosure
  • FIG. 4 shows a flowchart of still another method for processing a judgment result provided by an exemplary embodiment of the present disclosure
  • FIG. 5 shows a schematic structural diagram of a judgment result processing apparatus provided by an exemplary embodiment of the present disclosure
  • FIG. 6 shows a schematic structural diagram of another judgment result processing apparatus provided by an exemplary embodiment of the present disclosure
  • Fig. 7 shows a schematic structural diagram of an electronic device provided by an exemplary embodiment of the present disclosure.
  • FIG. 1 shows a judgment result processing method provided by an exemplary embodiment of the present disclosure to be able to explain the judgment result corresponding to the document to be processed, which may include the following steps:
  • the document to be processed is the document of the case to be judged (that is, the document of the case without a judgment result).
  • the corresponding legal element is a summary of the laws and regulations applicable to the document to be processed.
  • the legal element corresponding to the document to be processed is based on the document to be processed.
  • Obtain the text content in, for example, obtain the text content based on the facts describing the case in the document to be processed. If the content of the text describing the facts of the case is "the accident caused the victim Zhang's leg bone fracture, fifth-degree disability", then the legal element corresponding to this part of the text content is: causing a person's fifth-degree disability, which is the above " “Causing serious injury”.
  • the method of obtaining at least one legal element corresponding to the document to be processed includes but is not limited to the following methods:
  • this embodiment can display a preset element knowledge graph to assist manual labeling.
  • the preset element knowledge graph is used to indicate the hierarchical relationship between the law elements with the law element as the node.
  • the so-called hierarchical relationship indicates the law element to which a law element in the preset element knowledge graph belongs and a law element
  • the elements of the law are divided to indicate the upper and lower levels of the elements of the law.
  • the names of the legal elements marked by different persons can be unified, and the judgment result prediction error caused by the names of the legal elements can be reduced.
  • Another way take the text content of the document to be processed, especially the text content describing the facts of the case, as the input of the element recognition model, and obtain the legal elements corresponding to the document to be processed output by the element recognition model.
  • the element recognition model is obtained based on the training samples and test samples in the set of documents with known judgment results. If the test sample is recorded as the first legal document and the training sample is recorded as the second legal document, the process of obtaining the element recognition model includes the following steps :
  • each second legal document such as the content describing the facts of the case, the content of the judge's determination of the facts, and the laws and regulations of the second legal document. These specific contents are recorded in the second legal document.
  • the fact-finding section and the court’s opinion section can be extracted from the fact-finding section and the court’s opinion section of the second legal document.
  • the initial element recognition model can also output the correlation between each word in the second legal document and its corresponding legal elements.
  • the pre-built basic model can be a model based on Bi-LSTM+Attention, and its model architecture as shown in picture 2.
  • each layer of the model architecture shown in Figure 2 from bottom to top are: the first layer is an input layer, x T represents the T-th word in the document input to the model, and the second layer is the word vector layer, e T represents the word vector of the T-th word, the third layer is a bidirectional LSTM (Bi-LSTM) layer, which is used to enable the fourth layer to obtain the word vector of each word, and the fourth layer is the attention layer, It is used to calculate the weight of each word.
  • the fifth layer is the output layer, which is used to output the legal element and the weight of each word. The weight of each word is the correlation between each word and the current output legal element .
  • the selection of a sentence is based on the word to which the selected relevance belongs, and other methods can also be used. Taking any first legal document as an example, one method is to obtain the information of each word in the first legal document. The number of co-occurrences. The so-called number of co-occurrences is the total number of occurrences of the word in the first legal document. A certain number of words are selected based on the co-occurrence of each word in the first legal document (such as the highest value of the co-occurrence N words in the first legal document), select the sentence containing the word selected based on the number of co-occurrences from the first legal document. Another way is to select the word based on the relevance and the number of co-occurrences, and then select the first legal document that contains the word Word sentence.
  • the selected words are marked specifically, for example, the selected words are marked with specific colors and special effects (such as bold and/or oblique) to mark the selected words. Words are highlighted, so that people can understand the sentence at a glance based on specific tags, which is convenient for manual summarization. Further, specific tags used for words with different degrees of relevance can also be different. Taking specific colors as an example, words with higher degrees of relevance The more prominent the specific color is used, the more relevant words (indicating that the words can better reflect the facts of the case) will be more prominently displayed.
  • the distance between the same and adjacent words can also be marked, and by marking the distance, it can be known that the higher the number of co-occurrences, the shorter the distance between words. Then based on the distance, it can be determined which words have the higher the number of co-occurrences, indicating that the words with the higher the number of co-occurrences can better reflect the facts of the first legal document. This can be artificially based on the facts of the first legal document.
  • the word of fact sums up the elements of the law.
  • step 7 Based on the legal elements summarized in step 6 and the legal elements referred to in step 2 (that is, the legal elements summarized from the legal provisions of laws and regulations), relabel the legal elements of each second legal document , And then take the specific content in each second legal document as input, and take the relabeled legal elements of each second legal document as output, and retrain the initial element recognition model to obtain the element recognition model.
  • Each first legal document is used to test the element recognition model.
  • the test process please refer to the test process of the basic model, which is not described in this embodiment.
  • One way is to obtain the correlation degree between each word in the document to be processed and each legal element. For any legal element in each legal element, select from the correlation between each word and the legal element. The relevance of the first preset condition is obtained, and the text content corresponding to the legal element is obtained from the document to be processed based on the word that meets the relevance of the first preset condition.
  • the degree of relevance is used to indicate the degree to which the words in the document to be processed can interpret or apply to the element of the law.
  • the higher the degree of correlation the better the interpretation of the element of the law or the more applicable to the judgment result of the element of the law. It is predicted that the correlation between any word and a legal element can be obtained based on the above-mentioned element recognition model.
  • the first preset condition may be determined according to the actual application. For example, the first preset condition may be to select the N relevance degrees with the highest relevance values, so that any legal element corresponding to the document to be processed can be obtained.
  • the N relevance belonging words corresponding to the legal element, and then the sentence where these words are located and/or the laws and regulations corresponding to these words can be used as the text content corresponding to the legal element.
  • Another way obtain the words corresponding to each legal element from the preset related word set, obtain the sentence to which the word corresponding to each legal element belongs and the laws and regulations of the word corresponding to each legal element in the document to be processed, Take the obtained sentences and laws and regulations as the text content corresponding to the elements of each law.
  • the words in the set of presupposed related words are the correlation between each word in each first legal document and each legal element corresponding to each first legal document based on the known judgment result and/or each word in each first legal document
  • the number of co-occurrences is obtained, and the correlation between each legal element corresponding to each first legal document and the correlation between each word in each first legal document and each legal element corresponding to each first legal document is obtained based on the element identification model, and the element identification model It takes each second legal document with a known judgment result as input, and will use each legal element marked in each second legal document and the correlation between each word in each second legal document and each marked legal element as output training Obtained
  • the specific description can refer to the description of the element recognition model provided in this embodiment.
  • each word in each first legal document with a known judgment result After obtaining the correlation between each word in each first legal document with a known judgment result and each legal element corresponding to each first legal document and/or the number of co-occurrences of each word in each first legal document, based on each first legal document 1.
  • the correlation between each word in a legal document and each legal element corresponding to each first legal document and/or the number of co-occurrences of each word in each first legal document select words from each first legal document as the presupposition correlation For the words in the word set, for example, the N words with the highest relevance and/or the N co-occurrences with the highest number of co-occurrences are selected.
  • the crime types corresponding to the judgment results of each first legal document may be different.
  • a preset related word set can be set for each crime type, and then when the preset related word set is obtained , Can be based on the correlation between each word of the first legal document under the same charge type and each legal element corresponding to each first legal document and/or the number of co-occurrences of each word in the first legal document under the charge type, Select words from the first legal document under the crime type.
  • the judgment result corresponding to the document to be processed is based on the elements of the law, including the trial results such as crime and sentence.
  • each judgment result can be applied to at least one element of the law.
  • the applicable legal elements correspond to the judgment results in order to establish the corresponding relationship between the legal elements and the judgment results.
  • one way to obtain at least one judgment result corresponding to the document to be processed is to use at least one legal element corresponding to the document to be processed as the input of the judgment result prediction model to obtain the document to be processed output by the judgment result prediction model Corresponding to at least one judgment result, the judgment result prediction model also outputs the correlation between each legal element and each judgment result.
  • the judgment result prediction model takes each legal element corresponding to each legal document for which the judgment result is known as input, and trains the judgment result corresponding to each legal document and the correlation between each legal element and each judgment result as output training.
  • the judgment result prediction model is similar to the acquisition process of the above-mentioned element recognition model. Both are obtained by training the basic model with various legal documents with known judgment results as samples. The difference lies in the training process of the judgment result prediction model.
  • the input is the legal element corresponding to each legal document
  • the output of the model is the judgment result corresponding to each legal document, such as the crime and/or sentence corresponding to each legal document, so as to obtain a judgment result prediction model from the legal element to the judgment result Therefore, in this embodiment, the training and testing process of the judgment result prediction model will not be described in detail.
  • Another way to obtain at least one judgment result corresponding to the document to be processed in this embodiment is to obtain the judgment result corresponding to the document to be processed based on the expert rules that record the correspondence between the legal elements and the judgment result.
  • the process is as follows: Obtain from the expert rules the same legal element as the legal element of the document to be processed, and the judgment result corresponding to the same legal element is regarded as the judgment result corresponding to the document to be processed. For the establishment process of the expert rule, this embodiment Not to elaborate.
  • One way for the judgment result to correspond to the legal elements is to obtain the correlation between each legal element and each judgment result, and for any judgment result in each judgment result, from each legal element to the judgment From the relevance of the result, the relevance that satisfies the first preset condition is selected, and the legal element of the relevance that satisfies the first preset condition is determined as the legal element corresponding to the judgment result.
  • the correlation between each legal element and each judgment result can be output by the judgment result prediction model.
  • the higher the correlation with the judgment result the more the legal element can adapt to the judgment result, and the corresponding laws and regulations of the legal element are more suitable. It can accurately explain the judgment result. Therefore, the correlation degree that meets the first preset condition can be, but not limited to: the N correlation degrees with the highest correlation degree, so that the N correlation degrees with the highest correlation degree can be correlated
  • the legal element to which the degree belongs is determined as the legal element corresponding to the judgment result.
  • the first preset condition may also be changed according to actual needs, which is not described in this embodiment.
  • each judgment result with the text content corresponding to each corresponding legal element to establish a correspondence between the text content in the document to be processed and each judgment result, so that the judgment result corresponds to the text content, based on the corresponding relationship
  • the content of the text can know which factors of the document to be processed generate the judgment result, so as to know the reason for the judgment result.
  • the sentence to which the word corresponding to the legal element in the document to be processed belongs and the laws and regulations of the word corresponding to each legal element are regarded as the text content corresponding to the judgment result, and the word corresponding to the legal element in the document to be processed
  • the laws and regulations that belong to the sentence and the words corresponding to the elements of each law interpret the judgment result.
  • the judgment result processing method can obtain at least one legal element corresponding to the document to be processed and obtain the text content corresponding to each legal element in the document to be processed for a judgment document. At least one legal element corresponding to the document, obtain at least one judgment result corresponding to the document to be processed, and determine the legal element corresponding to each judgment result, and then associate each judgment result with the text content corresponding to the corresponding legal element, Therefore, the corresponding relationship between each judgment result corresponding to the document to be processed and the text content in the document to be processed can be established. Then, for any judgment result, it can be known which part of the text content in the document to be processed caused the judgment result to be obtained.
  • the text content that has a corresponding relationship with the judgment result is used to interpret the judgment result in order to know the reason for the judgment result, so that not only the judgment result corresponding to the document to be processed can be obtained, but the reason for the judgment result can be explained, so that the judgment result has Interpretability and traceability.
  • the obtaining of the judgment result and the association between the judgment result and the text content in this embodiment are based on the intermediate reference of the legal element, which improves the accuracy compared with obtaining the judgment result from the entire content of the document to be processed.
  • FIG. 3 shows the flow of another judgment result processing method provided by an exemplary embodiment of the present disclosure, which may include the following steps:
  • step 301 and step 302 are the same as the above step 101 and step 102, which will not be described in this embodiment.
  • 303 Obtain the same legal element in the knowledge map of predetermined elements that is the same as each legal element corresponding to the document to be processed, and the predetermined element knowledge map is used to indicate the hierarchical relationship between the various legal elements using the legal element as a node.
  • each legal element in the knowledge map of the preset elements is: a legal element corresponding to laws and regulations obtained by artificially summarizing laws and regulations, or a legal document based on known judgment results
  • the process of obtaining the element identification model artificially summarizes the legal elements from various legal documents, and when all legal elements are obtained, the hierarchical relationship of these legal elements is artificially given, thereby forming an element system.
  • each legal element is associated with the case fragments in the corresponding legal document that can explain the judgment result and the corresponding laws and regulations to form a preset element knowledge map, which is not only Indicate the hierarchical relationship between the various legal elements, and can also show the text content associated with each legal element, such as the case fragments and corresponding laws and regulations in the corresponding legal documents that can explain the judgment result.
  • the text content corresponding to the judgment result is not only the text content in the document to be processed, but also the law related to the law element in the knowledge map of preset elements that is the same as the law element corresponding to the document to be processed Laws and case fragments in legal documents with known judgment results, use case fragments in laws and regulations and legal documents with known judgment results to assist the interpretation of the judgment results in the pending documents, increase the interpretation content of the judgment results, and improve the interpretability And traceability.
  • FIG. 4 shows the flow of another judgment result processing method provided by an exemplary embodiment of the present disclosure. Based on FIG. 1, the following steps may be further included:
  • the text content associated with each legal element corresponds to a text relevance
  • the corresponding way of marking the text content can be: based on the text relevance of the text content of the associated legal element, the text content is obtained
  • the text content is specially marked in the special mark form corresponding to the text content.
  • the text content with different text relevance is marked by different special mark forms.
  • the text content with high text relevance uses the special mark form and the text content with low text relevance uses the special mark form to be more prominent.
  • text content with high text relevance take a specific color as an example.
  • the text content with high text relevance uses a darker specific color than the text content with low text relevance.
  • the method of obtaining the text relevance of the text content is: based on the relevance of each word in the text content to the legal element.
  • the text content is related to the legal element.
  • the relevance of the word with the highest value of the degree is taken as the relevance of the text content.
  • Another way is to perform a weighted summation of the relevance of each word in the text content with the legal elements to obtain the relevance of the text content.
  • One way is to average the relevance of each word in the text content to the legal elements to obtain the relevance of the text content.
  • the exemplary embodiment of the present disclosure also provides a judgment result processing device, the structure of which is shown in FIG. 5, which may include: a first obtaining unit 10, a second obtaining unit 20, and an associating unit 30 .
  • the first obtaining unit 10 is configured to obtain at least one legal element corresponding to the document to be processed, and obtain the text content corresponding to each legal element in the document to be processed, and the document to be processed is the document of the case to be judged.
  • the legal element corresponding to the document to be processed is a summary of the laws and regulations applicable to the document to be processed, and the legal element corresponding to the document to be processed is obtained based on the text content of the document to be processed, for example, based on the facts describing the case in the document to be processed The text content is obtained.
  • the method of obtaining at least one legal element corresponding to the document to be processed includes but is not limited to the following methods:
  • Another way take the text content of the document to be processed, especially the text content describing the facts of the case, as the input of the element recognition model, and obtain the legal elements corresponding to the document to be processed output by the element recognition model.
  • the element recognition model is obtained based on the training samples and test samples in the document collection with known judgment results.
  • the feasible ways for the first obtaining unit 10 to obtain the text content corresponding to each legal element in the document to be processed include but are not limited to the following ways:
  • One way is to obtain the correlation degree between each word in the document to be processed and each legal element. For any legal element in each legal element, select from the correlation between each word and the legal element. The relevance of the first preset condition is obtained, and the text content corresponding to the legal element is obtained from the document to be processed based on the word that meets the relevance of the first preset condition.
  • Another way obtain the words corresponding to each legal element from the preset related word set, obtain the sentence to which the word corresponding to each legal element belongs and the laws and regulations of the word corresponding to each legal element in the document to be processed, Use the obtained sentences and laws and regulations as the text content corresponding to the elements of each law.
  • the second obtaining unit 20 is configured to obtain at least one judgment result corresponding to the document to be processed based on at least one legal element corresponding to the document to be processed, and to determine the legal element corresponding to each judgment result from the at least one legal element.
  • one way to obtain at least one judgment result corresponding to the document to be processed is to use at least one legal element corresponding to the document to be processed as the input of the judgment result prediction model to obtain the document to be processed output by the judgment result prediction model Corresponding to at least one judgment result, the judgment result prediction model also outputs the correlation between each legal element and each judgment result.
  • the judgment result prediction model takes each legal element corresponding to each legal document for which the judgment result is known as input, and uses the judgment result corresponding to each legal document and the correlation between each legal element and each judgment result as output training. Please refer to the relevant description in the method embodiment.
  • Another way to obtain at least one judgment result corresponding to the document to be processed in this embodiment is to obtain the judgment result corresponding to the document to be processed based on the expert rules that record the correspondence between the legal elements and the judgment result.
  • the process is as follows: Obtain from the expert rules the same legal element as the legal element of the document to be processed, and the judgment result corresponding to the same legal element is regarded as the judgment result corresponding to the document to be processed. For the establishment process of the expert rule, this embodiment Not to elaborate.
  • One way for the judgment result to correspond to the legal elements is to obtain the correlation between each legal element and each judgment result, and for any judgment result in each judgment result, from each legal element to the judgment From the relevance of the result, the relevance that satisfies the first preset condition is selected, and the law element of the relevance that satisfies the first preset condition is determined as the law element corresponding to the judgment result.
  • the correlation between each legal element and each judgment result can be output by the judgment result prediction model.
  • the higher the correlation with the judgment result the more the legal element can adapt to the judgment result, and the corresponding laws and regulations of the legal element are more suitable. It can accurately explain the judgment result. Therefore, the correlation degree that meets the first preset condition can be, but not limited to: the N correlation degrees with the highest correlation degree, so that the N correlation degrees with the highest correlation degree can be correlated
  • the legal element to which the degree belongs is determined as the legal element corresponding to the judgment result.
  • the first preset condition may also be changed according to actual needs, which is not described in this embodiment.
  • the associating unit 30 is used for associating each judgment result with the text content corresponding to each corresponding legal element, so as to establish a correspondence between the text content in the document to be processed and each judgment result, so that the judgment result corresponds to the text content, thereby Based on the corresponding text content, it can be known which factors of the document to be processed generated the judgment result, so as to know the reason for the judgment result. For example, the sentence to which the word corresponding to the legal element in the document to be processed belongs and the laws and regulations of the word corresponding to each legal element are regarded as the text content corresponding to the judgment result, and the word corresponding to the legal element in the document to be processed The laws and regulations that belong to the sentence and the words corresponding to the elements of each law interpret the judgment result.
  • the judgment result processing method can obtain at least one legal element corresponding to the document to be processed and obtain the text content corresponding to each legal element in the document to be processed for a judgment document. At least one legal element corresponding to the document, obtain at least one judgment result corresponding to the document to be processed, and determine the legal element corresponding to each judgment result, and then associate each judgment result with the text content corresponding to the corresponding legal element, Therefore, the corresponding relationship between each judgment result corresponding to the document to be processed and the text content in the document to be processed can be established. Then, for any judgment result, it can be known which part of the text content in the document to be processed caused the judgment result to be obtained.
  • the text content that has a corresponding relationship with the judgment result is used to interpret the judgment result in order to know the reason for the judgment result, so that not only the judgment result corresponding to the document to be processed can be obtained, but the reason for the judgment result can be explained, so that the judgment result has Interpretability and traceability.
  • the obtaining of the judgment result and the association between the judgment result and the text content in this embodiment are based on the intermediate reference of the legal element, which improves the accuracy compared with obtaining the judgment result from the entire content of the document to be processed.
  • FIG. 6 shows another judgment result processing device provided by an exemplary embodiment of the present disclosure. Based on the above-mentioned FIG. 5, it may further include: a third obtaining unit 40 and a text content obtaining unit 50.
  • the third obtaining unit 40 is used to obtain the same legal elements in the knowledge map of predetermined elements that correspond to the various legal elements of the document to be processed, and the predetermined element knowledge map is used to indicate the legal elements with the legal elements as nodes. Hierarchical relationship between.
  • each legal element in the knowledge map of the preset elements is: a legal element corresponding to laws and regulations obtained by artificially summarizing laws and regulations, or a legal document based on known judgment results
  • the process of obtaining the element identification model artificially summarizes the legal elements from various legal documents, and when all legal elements are obtained, the hierarchical relationship of these legal elements is artificially given, thereby forming an element system.
  • each legal element is associated with the case fragments in the corresponding legal document that can explain the judgment result and the corresponding laws and regulations to form a preset element knowledge map, which is not only Indicate the hierarchical relationship between the various legal elements, and can also show the text content associated with each legal element, such as the case fragments and corresponding laws and regulations in the corresponding legal documents that can explain the judgment result.
  • the text content obtaining unit 50 is configured to use the law and regulation associated with the same legal element and the case fragments in the legal document with known judgment results as the text content corresponding to the legal element corresponding to the document to be processed.
  • the text content corresponding to the judgment result is not only the text content in the document to be processed, but also the law related to the law element in the knowledge map of preset elements that is the same as the law element corresponding to the document to be processed Laws and case fragments in legal documents with known judgment results, use case fragments in laws and regulations and legal documents with known judgment results to assist the interpretation of the judgment results in the pending documents, increase the interpretation content of the judgment results, and improve the interpretability And traceability.
  • the associating unit 30 provided in this embodiment can also perform a special mark on the text content of the associated legal element in the document to be processed, so as to highlight the text content of the associated legal element in the form of a special mark.
  • the text content associated with each legal element corresponds to a text relevance
  • the corresponding way of marking the text content can be: based on the text relevance of the text content of the associated legal element, the text content is obtained
  • the text content is specially marked in the special mark form corresponding to the text content.
  • the judgment result processing device includes a processor and a memory.
  • the first acquiring unit 10, the second acquiring unit 20, and the associating unit 30 are all stored in the memory as program units, and the program units stored in the memory are executed by the processor. To realize the corresponding function.
  • the processor contains the kernel, and the kernel calls the corresponding program unit from the memory.
  • the kernel can be set to one or more, by adjusting the kernel parameters to associate each judgment result corresponding to the document to be processed with the text content corresponding to each corresponding legal element, so as to interpret the judgment through the text content corresponding to the judgment result
  • the reason for the judgment result not only the judgment result corresponding to the document to be processed can be obtained, but also the reason for the judgment result can be explained, so that the judgment result is interpretable and traceable.
  • the embodiment of the present invention provides a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, the judgment result processing method is implemented.
  • the embodiment of the present invention provides a processor, the processor is used to run a program, wherein the judgment result processing method is executed when the program is running.
  • the embodiment of the present invention provides an electronic device whose structure is shown in FIG. 7.
  • the electronic device includes at least one processor 701, and at least one memory 702 and a bus 703 connected to the processor 701; wherein the processor and the memory pass through The bus completes the communication between each other; the processor is used to call the program instructions in the memory to execute the above-mentioned judgment result processing method.
  • the electronic devices in this article can be servers, PCs, PADs, mobile phones, etc.
  • This application also provides a computer program product, which when executed on a data processing device, is suitable for executing a program that initializes the following method steps:
  • the obtaining the text content corresponding to each of the legal elements in the document to be processed includes:
  • the correlation degree that satisfies the first preset condition is selected from the correlation between each word and the legal element, and the correlation degree is based on the correlation degree that satisfies the first preset condition.
  • the word to which the degree of relevance belongs is obtained from the document to be processed, and the text content corresponding to the legal element is obtained.
  • the obtaining the text content corresponding to each of the legal elements in the document to be processed includes:
  • each of the legal elements from the preset related word set, and the words in the preset related word set are based on the known judgment results in each first legal document corresponding to each first legal document
  • the correlation degree of each legal element and/or the number of co-occurrences of each word in each first legal document is obtained, each legal element corresponding to each first legal document, and each word in each first legal document and each first legal document
  • the correlation degree of each legal element corresponding to the document is obtained based on the element recognition model, which takes each second legal document with known judgment results as input, and will label each legal element and each second legal document as input.
  • the correlation between each word in each second legal document and each labeled legal element is obtained as output training;
  • the determination of the legal element corresponding to each judgment result from the at least one legal element includes:
  • the correlation degree that satisfies the second preset condition For any one of the judgment results, select the correlation degree that satisfies the second preset condition from the correlation degree of each legal element with the judgment result, and calculate the correlation degree that satisfies the second preset condition.
  • the element of the legal article to which it belongs is determined as the element of the legal article corresponding to the judgment result.
  • the obtaining at least one judgment result corresponding to the document to be judged based on at least one legal element corresponding to the document to be judged includes:
  • At least one legal element corresponding to the document to be judged is used as the input of the judgment result prediction model, and at least one judgment result corresponding to the document to be judged output by the judgment result prediction model is obtained, and the judgment result prediction model also outputs The degree of relevance of each legal element to each judgment result;
  • the judgment result prediction model takes each legal element corresponding to each legal document for which the judgment result is known as input, and uses the judgment result corresponding to each legal document and the correlation between each legal element and each judgment result as output training to obtain .
  • a program when executed on a data processing device, it is also suitable for executing a program that initializes the following method steps: in the document to be processed, the text content associated with the legal element is specially marked.
  • the text content associated with each legal element corresponds to a text relevance
  • the special marking of the text content associated with the legal element in the document to be processed includes: obtaining a special marking form corresponding to the text content based on the text relevance of the text content associated with the legal element, and The special mark form corresponding to the text content makes a special mark on the text content.
  • the device includes one or more processors (CPUs), memory, and buses.
  • the device may also include input/output interfaces, network interfaces, and so on.
  • the memory may include non-permanent memory in computer-readable media, random access memory (RAM) and/or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one Memory chip.
  • RAM random access memory
  • ROM read-only memory
  • flash RAM flash memory
  • the memory is an example of a computer-readable medium.
  • Computer-readable media include permanent and non-permanent, removable and non-removable media, and information storage can be realized by any method or technology.
  • the information can be computer-readable instructions, data structures, program modules, or other data.
  • Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, Magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media can be used to store information that can be accessed by computing devices. According to the definition in this article, computer-readable media does not include transitory media, such as modulated data signals and carrier waves.
  • this application can be provided as a method, a system, or a computer program product. Therefore, this application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
  • a computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Biomedical Technology (AREA)
  • Primary Health Care (AREA)
  • Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Technology Law (AREA)
  • Strategic Management (AREA)
  • Marketing (AREA)
  • Human Resources & Organizations (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Animal Behavior & Ethology (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种判决结果处理方法及装置，获得待处理文书对应的至少一个法条要素以及获得待处理文书中与各个法条要素对应的文本内容，基于待处理文书对应的至少一个法条要素，获得待处理文书对应的至少一个判决结果，并确定各个判决结果对应的法条要素，然后将各个判决结果与各自对应的法条要素对应的文本内容进行关联，由此建立待处理文书对应的各个判决结果与待处理文书中的文本内容的对应关系，那么对于任一判决结果可以获知是由于待处理文书中的哪部分文本内容导致得到该判决结果，从而通过与判决结果具有对应关系的文本内容来解读该判决结果，以获知得到该判决结果的原因，使得判决结果具有可解释性和可溯源性。

Description

一种判决结果处理方法及装置
本申请要求于2019年09月25日提交中国专利局、申请号为201910912134.3、发明名称为“一种判决结果处理方法及装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及机器学习技术领域，更具体地说，尤其涉及一种判决结果处理方法及装置。
背景技术
目前司法案件的判决主要由案件审核人员主导，例如案件审核人员基于预先得到的专家规则得出其审理的司法案件的判决结果，其过程可以是：获得司法领域专家从法律法规的规定中总结出的法条要素，并将法条要素与对应的法律法规进行关联，以得到法条要素和法律法规的对应关系的专家规则，在获得任一待处理文书之后获得案件审核人员对该待处理文书标记的法条要素，从专家规则中获得该法条要素对应的法律法规，然后基于该法条要素对应的法律法规，获得待处理文书的判决结果。
其中法条要素是对一条法律法规的摘要，以交通肇事罪为例，其对应的法律法规是：违反交通运输管理法规，因而发生重大事故，致人重伤、死亡或者使公私财产遭受重大损失的，处三年以下有期徒刑或者拘役，那么该法律法规的法条要素可以是：致人重伤、致人死亡和造成重大损失中的至少一个，以这些法条要素对法律法规进行概括说明。
发明内容
鉴于上述问题，本发明提供一种克服上述问题或者至少部分地解决上述问题的判决结果处理方法及装置，用于将待处理文书对应的各个判决结果与各自对应的法条要素对应的文本内容进行关联。技术方案如下：
本发明提供一种判决结果处理方法，所述方法包括：
获得待处理文书对应的至少一个法条要素，并获得所述待处理文书中与各个所述法条要素对应的文本内容，所述待处理文书为待判决案件的文书；
基于所述待处理文书对应的至少一个法条要素，获得所述待处理文书对应的至少一个判决结果，并从所述至少一个法条要素中确定各个判决结果对应的法条要素；
将各个判决结果与各自对应的法条要素对应的文本内容进行关联，以使各个判决结果与文本内容对应。
优选的，所述获得所述待处理文书中与各个所述法条要素对应的文本内容包括：
获得所述待处理文书中各个词分别与各个所述法条要素的相关度；
对所述各个法条要素中的任一法条要素，从各个词分别与该法条要素的相关度中选取满足第一预设条件的相关度，并基于所述满足第一预设条件的相关度所属词，从所述待处理文书中获得该法条要素对应的文本内容。
优选的，所述获得所述待处理文书中与各个所述法条要素对应的文本内容包括：
从预设相关词集合中获得与各个所述法条要素对应的词，所预设相关词集合中的词是基于已知判决结果的各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度和/或各个第一法律文书中各个词的共现次数获得，各个第一法律文书对应的各个法条要素以及各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度基于要素识别模型获得，所述要素识别模型是将已知判决结果的各个第二法律文书作为输入，将为各个第二法律文书标注的各个法条要素以及各个第二法律文书中各个词与标注的各个法条要素的相关度作为输出训练得到；
获得所述待处理文书中与各个所述法条要素对应的词所属句子和与各个所述法条要素对应的词所属法律法规，将所获得的句子和法律法规作为与各个法条要素对应的文本内容。
优选的，所述方法还包括：
获得预设要素知识图谱中与待处理文书对应的各个法条要素相同的法条要素，所述预设要素知识图谱用于以法条要素为节点指示各个法条要素之间的层级关系；
将所述相同的法条要素关联的法律法规和已知判决结果的法律文书中的 案例片段作为所述待处理文书对应的该法条要素对应的文本内容。
优选的，所述从所述至少一个法条要素中确定各个判决结果对应的法条要素包括：
获得所述各个法条要素分别与各个判决结果的相关度；
对所述各个判决结果中的任一判决结果，从各个法条要素分别与该判决结果的相关度中选取满足第二预设条件的相关度，将所述满足第二预设条件的相关度所属法条要素确定为该判决结果对应的法条要素。
优选的，所述基于所述待判决文书对应的至少一个法条要素，获得所述待判决文书对应的至少一个判决结果包括：
将所述待判决文书对应的至少一个法条要素作为判决结果预测模型的输入，获得所述判决结果预测模型输出的所述待判决文书对应的至少一个判决结果，所述判决结果预测模型还输出各个法条要素分别与各个判决结果的相关度；
其中所述判决结果预测模型是将已知判决结果的各个法律文书对应的各个法条要素作为输入，将各个法律文书对应的判决结果以及各个法条要素与各个判决结果的相关度作为输出训练得到。
优选的，所述方法还包括：在所述待处理文书中对关联所述法条要素的文本内容进行特殊标记。
优选的，关联各个法条要素的文本内容对应有文本相关度；
所述在所述待处理文书中对关联所述法条要素的文本内容进行特殊标记包括：基于关联所述法条要素的文本内容的文本相关度，获得该文本内容对应的特殊标记形式，以该文本内容对应的特殊标记形式对该文本内容进行特殊标记。
本发明还提供一种判决结果处理装置，所述装置包括：
第一获取单元，用于获得待处理文书对应的至少一个法条要素，并获得所述待处理文书中与各个所述法条要素对应的文本内容，所述待处理文书为待判决案件的文书；
第二获取单元，用于基于所述待处理文书对应的至少一个法条要素，获得所述待处理文书对应的至少一个判决结果，并从所述至少一个法条要素中确定 各个判决结果对应的法条要素；
关联单元，用于将各个判决结果与各自对应的法条要素对应的文本内容进行关联，以使各个判决结果与文本内容对应。
本发明还提供一种电子设备，所述电子设备包括至少一个处理器、以及与处理器连接的至少一个存储器、总线；其中，所述处理器、所述存储器通过所述总线完成相互间的通信；所述处理器用于调用所述存储器中的程序指令，以执行上述判决结果处理方法。
本发明还提供一种计算机可读介质，其上存储有计算机程序，该程序被处理器执行时实现上述判决结果处理方法。
借由上述技术方案，本发明提供的判决结果处理方法及装置，对于一个判决文书能够获得待处理文书对应的至少一个法条要素以及获得待处理文书中与各个法条要素对应的文本内容，基于待处理文书对应的至少一个法条要素，获得待处理文书对应的至少一个判决结果，并确定各个判决结果对应的法条要素，然后将各个判决结果与各自对应的法条要素对应的文本内容进行关联，由此建立待处理文书对应的各个判决结果与待处理文书中的文本内容的对应关系，使得判决结果与文本内容对应，那么对于任一判决结果可以获知是由于待处理文书中的哪部分文本内容导致得到该判决结果，从而通过与判决结果具有对应关系的文本内容来解读该判决结果，以获知得到该判决结果的原因，使得不单单是得到待处理文书对应的判决结果还能够对判决结果的原因进行说明，使得判决结果具有可解释性和可溯源性。
并且本实施例中判决结果的获得以及判决结果与文本内容的关联是基于法条要素这一中间参考，相对于从待处理文书的全部内容得到判决结果来说，提高准确度。
上述说明仅是本发明技术方案的概述，为了能够更清楚了解本发明的技术手段，而可依照说明书的内容予以实施，并且为了让本发明的上述和其它目的、特征和优点能够更明显易懂，以下特举本发明的具体实施方式。
附图说明
通过阅读下文优选实施方式的详细描述，各种其他的优点和益处对于本领域普通技术人员将变得清楚明了。附图仅用于示出优选实施方式的目的，而并 不认为是对本发明的限制。而且在整个附图中，用相同的参考符号表示相同的部件。在附图中：
图1示出了本公开的示例性实施例提供的一种判决结果处理方法的流程图；
图2示出了本公开的示例性实施例提供的要素识别模型的模型架构示意图；
图3示出了本公开的示例性实施例提供的另一种判决结果处理方法的流程图；
图4示出了本公开的示例性实施例提供的再一种判决结果处理方法的流程图；
图5示出了本公开的示例性实施例提供的一种判决结果处理装置的结构示意图；
图6示出了本公开的示例性实施例提供的另一种判决结果处理装置的结构示意图；
图7示出了本公开的示例性实施例提供的一种电子设备的结构示意图。
具体实施方式
下面将参照附图更详细地描述本公开的示例性实施例。虽然附图中显示了本公开的示例性实施例，然而应当理解，可以以各种形式实现本公开而不应被这里阐述的实施例所限制。相反，提供这些实施例是为了能够更透彻地理解本公开，并且能够将本公开的范围完整的传达给本领域的技术人员。
请参阅图1，其示出了本公开的示例性实施例提供的一种判决结果处理方法，用以能够对待处理文书对应的判决结果进行解释说明，可以包括以下步骤：
101：获得待处理文书对应的至少一个法条要素，并获得待处理文书中与各个法条要素对应的文本内容。
其中待处理文书为待判决案件的文书(即没有判决结果的案件的文书)对应的法条要素是对适用于待处理文书的法律法规的摘要，待处理文书对应的法条要素基于待处理文书中的文本内容获得，例如基于待处理文书中描述案情事实的文本内容获得。如描述案情事实的文本内容是“该事故造成了受害人张某的腿骨骨折，五级伤残”，那么这部分文本内容对应的法条要素是：致人五级 伤残，是上述“致人重伤”的一种情况。在本实施例中，获得待处理文书对应的至少一个法条要素的方式包括但不限于如下方式：
一种方式：获得人工为待处理文书标注的法条要素，例如获得案件审核人员标注出的法条要素。在人工标注法条要素过程中，本实施例可以展示预设要素知识图谱以辅助人工标注。其中预设要素知识图谱用于以法条要素为节点指示各个法条要素之间的层级关系，所谓层级关系表明预设要素知识图谱中的一个法条要素所属的法条要素以及一个法条要素下又划分的法条要素，以指示出法条要素之间的上下层级关系。以上述“致人重伤”和“致人五级伤残”为例，“致人五级伤残”所属的法条要素是“致人重伤”，说明“致人重伤”是“致人五级伤残”的上一层级，“致人重伤”下又划分的法条要素有“致人五级伤残”，说明“致人五级伤残”是“致人重伤”的下一层级。
通过预设要素知识图谱的辅助能够使得不同人员标注的法条要素的名称统一，降低因法条要素的名称导致判决结果预测出错。
另一种方式：将待处理文书中的文本内容，尤其是描述案情事实的文本内容作为要素识别模型的输入，获得要素识别模型输出的待处理文书对应的法条要素。
其中要素识别模型是基于已知判决结果的文书集合中的训练样本和测试样本得到，假如测试样本记为第一法律文书，训练样本记为第二法律文书，要素识别模型的获得过程包括以下步骤：
1、提取各个第二法律文书中的特定内容，如描述案情事实的内容、法官对事实认定的内容以及第二法律文书使用的法律法规中的法条，这些特定内容记录在第二法律文书的事实查明段和本院认为段，由此可以提取第二法律文书的事实查明段和本院认为段。
2、通过已有文书解析算法从各个第二法律文书中的特定内容中识别出法条，对任一法条，基于该法条所属第二法律文书的判决结果中的罪名类型，获得该法条对应的法条要素。
目前法律法规中对不同罪名类型规定有多个法条，对于同一种罪名类型下的各个法条人工总结出每个法条对应的法条要素，那么在识别出法条之后可以基于该法条所述第二法律文书的判决结果中的罪名类型，从已总结出的法条要 素中获得该法条对应的法条要素。
3、将各个第二法律文书中的特定内容作为输入，将各个第二法律文书对应的法条要素作为输出，对预先构建的基础模型进行训练，得到初始要素识别模型，对于任一第二法律文书来说，该初始要素识别模型还能够输出该第二法律文书中各个词与其对应的法条要素的相关度，例如预先构建的基础模型可以是基于Bi-LSTM+Attention的模型，其模型架构如图2所示。
图2所示模型架构从下往上各层功能分别是：第一层是一个输入层，x T表示输入到该模型中的文书中的第T个词，第二层是词向量层，e T表示第T个词的词向量，第三层是双向LSTM(Bi-LSTM)层，用于使第四层能够获得每个词的词向量，第四层是注意力层(Attention Layer)，用于计算每个词的权重，第五层是输出层，用于输出法条要素和每个词的权重，每个词的权重视为是每个词与当前输出的法条要素的相关度。
4、提取各个第一法律文书中的特定内容，将各个第一法律文书中的特定内容输入到初始要素识别模型中，得到该要素识别模型输出的各个第一法律文书对应的法条要素；如果任一第一法律文书，初始要素识别模型输出的该第一法律文书对应的法条要素符合预设条件，如输出的该第一法律文书对应的法条要素是该第一法律文书的罪名类型下预先总结出的一个法条要素，则停止对基础模型的训练，否则对基础模型的模型参数和/或各个第二法律文书的输入和输出等进行修正，以基于修正后的内容对基础模型进行再次训练。
5、获得初始要素识别模型输出的各个第一法律文书中各个词与第一法律文书对应的法条要素的相关度，对任一第一法律文书，从该第一法律文书中各个词与该第一法律文书对应的法条要素的相关度中选取第一数量的相关度(如选取相关度的取值最高的N个相关度)。
6、获得第一法律文书中包含所选取的相关度所属词的句子，并获得人工参照所获得的句子总结出的法条要素。
在本实施例中句子的选取除基于所选取的相关度所属词之外，还可以采用其他方式，以任一第一法律文书为例，一种方式是获得该第一法律文书中各个词的共现次数，所谓共现次数是该词在该第一法律文书中出现的总次数，基于该第一法律文书中各个词的共现次数选取一定数量的词(如共现次数的取值最 高的N个词)，从该第一法律文书中选取包含基于共现次数所选取的词的句子，另一种方式是基于相关度和共现次数选取词，然后选取第一法律文书中包含该词的句子。
为了便于人工参照所获得的句子总结出法条要素，对所选取词进行特定标记，例如对所选取词采用特定颜色、特效(如加粗和/或倾斜)进行标记，以将所选取词突出显示，这样人工基于特定标记能够一目了然其所在句子，便于人工总结，进一步的对于相关度不同的词来说采用的特定标记也可以不同，以特定颜色为例，相关度越高的词采用的特定颜色越突出，这样能够将相关度越高(表明词更能体现案情事实)的词更加的突出显示。
此外在获得第一法律文书中词的共现次数的情况下，还可以标记相同且相邻的词之间的距离，并且通过标记距离可知共现次数越高的词之间的距离越短，那么基于距离可以确定出哪些词的共现次数越高，说明这些共现次数越高的词更能体现所属第一法律文书的案情事实，由此人工可以基于该体现所属第一法律文书的案情事实的词总结出法条要素。
7、基于步骤6总结出的法条要素和步骤2参照的法条要素(即从法律法规规定的法条中总结出的法条要素)，对各个第二法律文书进行法条要素的重新标注，然后将各个第二法律文书中的特定内容作为输入，将各个第二法律文书重新标注的法条要素作为输出，对初始要素识别模型进行重新训练，得到要素识别模型。并利用各个第一法律文书对要素识别模型进行测试，测试过程请参阅对基础模型的测试过程，对此本实施例不在阐述。
而获得待处理文书中与各个法条要素对应的文本内容的可行方式包括但不限于下述方式：
一种方式是：获得待处理文书中各个词分别与各个法条要素的相关度，对各个法条要素中的任一法条要素，从各个词分别与该法条要素的相关度中选取满足第一预设条件的相关度，并基于满足第一预设条件的相关度所属词，从待处理文书获得该法条要素对应的文本内容。
其中相关度用于表明待处理文书中的词能够解释法条要素或者适用于该法条要素的程度，相关度越高说明更能解释法条要素或更适用于该法条要素进行判决结果的预测，任一词与一个法条要素的相关度可以基于上述要素识别模 型获得。
第一预设条件可以根据实际应用而定，例如第一预设条件可以是选取相关度取值最高的N个相关度，由此对于待处理文书对应的任一法条要素，能够获得与该法条要素对应的N个相关度所属词，进而可以将这些词所在句子和/或这些词对应的法律法规作为该法条要素对应的文本内容。
另一种方式：从预设相关词集合中获得与各个法条要素对应的词，获得待处理文书中与各个法条要素对应的词所属句子和与各个法条要素对应的词所属法律法规，将获得的句子和法律法规作为与各个法条要素对应的文本内容。
其中预设相关词集合中的词是基于已知判决结果的各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度和/或各个第一法律文书中各个词的共现次数获得，各个第一法律文书对应的各个法条要素以及各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度基于要素识别模型获得，要素识别模型是将已知判决结果的各个第二法律文书作为输入，将为各个第二法律文书标注的各个法条要素以及各个第二法律文书中各个词与标注的各个法条要素的相关度作为输出训练得到，具体说明可参见本实施例提供的要素识别模型的说明。
在获得已知判决结果的各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度和/或各个第一法律文书中各个词的共现次数之后，基于各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度和/或各个第一法律文书中各个词的共现次数，从各个第一法律文书中选取词作为预设相关词集合中的词，例如选取相关度的取值最高的N个相关度和/或共现次数最高的N个共现次数对应的词。
在这里需要说明的一点是：各个第一法律文书的判决结果对应的罪名类型可能不同，在本实施例可以为每个罪名类型设置一个预设相关词集合，那么在获得预设相关词集合时，可以基于同一个罪名类型下的第一法律文书各个词与各个第一法律文书对应的各个法条要素的相关度和/或该罪名类型下的第一法律文书中各个词的共现次数，从该罪名类型下的第一法律文书中选取词。
102：基于待处理文书对应的至少一个法条要素，获得待处理文书对应的至少一个判决结果，并从至少一个法条要素中确定各个判决结果对应的法条要 素。
也就是说，待处理文书对应的判决结果是以法条要素为参照基础得到的包括诸如罪名和刑期等的审判结果，在判决中每个判决结果能够适用于至少一个法条要素，由此将其适用的法条要素与判决结果对应，以建立法条要素与判决结果的对应关系。
在本实施例中获得待处理文书对应的至少一个判决结果的一种方式是：将待处理文书对应的至少一个法条要素作为判决结果预测模型的输入，获得判决结果预测模型输出的待处理文书对应的至少一个判决结果，判决结果预测模型还输出各个法条要素分别与各个判决结果的相关度。
其中判决结果预测模型是将已知判决结果的各个法律文书对应的各个法条要素作为输入，将各个法律文书对应的判决结果以及各个法条要素与各个判决结果的相关度作为输出训练得到。判决结果预测模型与上述要素识别模型的获得过程相类似，都是以已知判决结果的各个法律文书为样本对基础模型进行训练得到，不同之处在于判决结果预测模型的训练过程中，模型的输入是各个法律文书对应的法条要素，模型的输出是各个法律文书对应的判决结果，如各个法律文书对应的罪名和/或刑期，从而得到一个由法条要素到判决结果的判决结果预测模型，对此本实施例不再对判决结果预测模型的训练和测试过程进行详细说明。
在本实施例中获得待处理文书对应的至少一个判决结果的另一种方式是：基于记录有法条要素与判决结果的对应关系的专家规则获得待处理文书对应的判决结果，其过程是：从专家规则中获得与待处理文书的法条要素相同的法条要素，将该相同的法条要素对应的判决结果视为是待处理文书对应的判决结果，对于专家规则的建立过程本实施例不在阐述。
而判决结果与法条要素的对应的一种方式是：获得各个法条要素分别与各个判决结果的相关度，并对各个判决结果中的任一判决结果，从各个法条要素分别与该判决结果的相关度中选取满足第一预设条件的相关度，将满足第一预设条件的相关度所属法条要素确定为该判决结果对应的法条要素。
其中各个法条要素分别与各个判决结果的相关度可由判决结果预测模型输出，与判决结果的相关度越高说明该法条要素更能适配该判决结果，该法条 要素对应的法律法规更能准确说明该判决结果，为此满足第一预设条件的相关度可以是但不限于是：相关度的取值最高的N个相关度，从而可以将相关度的取值最高的N个相关度所属法条要素确定为该判决结果对应的法条要素，在实际应用中第一预设条件还可以根据实际需求而变化，对此本实施例不在阐述。
103：将各个判决结果与各自对应的法条要素对应的文本内容进行关联，以将待处理文书中的文本内容与各个判决结果建立对应关系，使得判决结果与文本内容对应，从而基于有对应关系的文本内容可知判决结果是由该待处理文书的哪些因素生成，以获知得到该判决结果的原因。例如将待处理文书中与法条要素对应的词所属句子和与各个法条要素对应的词所属法律法规作为与判决结果又对应关系的文本内容，通过待处理文书中与法条要素对应的词所属句子和与各个法条要素对应的词所属法律法规解释判决结果。
借由上述技术方案，本发明提供的判决结果处理方法，对于一个判决文书能够获得待处理文书对应的至少一个法条要素以及获得待处理文书中与各个法条要素对应的文本内容，基于待处理文书对应的至少一个法条要素，获得待处理文书对应的至少一个判决结果，并确定各个判决结果对应的法条要素，然后将各个判决结果与各自对应的法条要素对应的文本内容进行关联，由此建立待处理文书对应的各个判决结果与待处理文书中的文本内容的对应关系，那么对于任一判决结果可以获知是由于待处理文书中的哪部分文本内容导致得到该判决结果，从而通过与判决结果具有对应关系的文本内容来解读该判决结果，以获知得到该判决结果的原因，使得不单单是得到待处理文书对应的判决结果还能够对判决结果的原因进行说明，使得判决结果具有可解释性和可溯源性。
并且本实施例中判决结果的获得以及判决结果与文本内容的关联是基于法条要素这一中间参考，相对于从待处理文书的全部内容得到判决结果来说，提高准确度。
请参阅图3，其示出了本公开的示例性实施例提供的另一种判决结果处理方法的流程，可以包括以下步骤：
301：获得待处理文书对应的至少一个法条要素，并获得待处理文书中与各个法条要素对应的文本内容。
302：基于待处理文书对应的至少一个法条要素，获得待处理文书对应的至少一个判决结果，并从至少一个法条要素中确定各个判决结果对应的法条要素。
在本实施例中，步骤301和步骤302：与上述步骤101和步骤102相同，对此本实施例不在阐述。
303：获得预设要素知识图谱中与待处理文书对应的各个法条要素相同的法条要素，预设要素知识图谱用于以法条要素为节点指示各个法条要素之间的层级关系。
在本实施例中预设要素知识图谱中的每个法条要素是：通过对法律法规进行人工总结得出的与法律法规对应的法条要素，或者是对已知判决结果的各个法律文书基于要素识别模型的获得过程人工从各个法律文书中总结出的法条要素，在获得所有法条要素的情况下由人工给出这些法条要素的层级关系，从而形成要素体系。
在要素体系的基础上将每个法条要素与对应的法律文书中能够解释该判决结果的案例片段和对应的法律法规进行关联，形成预设要素知识图谱，由此该预设要素知识图谱不仅指示各个法条要素之间的层级关系，还能够示出各个法条要素关联的文本内容，例如对应的法律文书中能够解释该判决结果的案例片段和对应的法律法规。
在这里需要说明的一点是：对任一罪名类型，现有法律法规是对该罪名类型下的多种情况进行总结得到的一个涵盖较大范围的多个法条，而实际发生的案件可能是这些法条所涵盖范围的一种情况，如上述“致人重伤”和“致人五级伤残”为例，因此如果单独从法律法规中总结出法条要素，会使得这些法条要素不能涵盖实际发生的案件的范围，为此需要基于法律法规和已知判决结果的各个法律文书，获得具有不同粒度的法条要素，即属于不同层级涵盖不同范围的法条要素。并且每个罪名类型下的法条要素有所不同，为此本实施例能够为每个罪名类型对应一个预设要素知识图谱。
304：将相同的法条要素关联的法律法规和已知判决结果的法律文书中的 案例片段作为待处理文书对应的该法条要素对应的文本内容。
305：将各个判决结果与各自对应的法条要素对应的文本内容进行关联。
从上述技术方案可知，与判决结果对应的文本内容不单单是待处理文书中的文本内容，还可以是预设要素知识图谱中与待处理文书对应的法条要素相同的法条要素关联的法律法规和已知判决结果的法律文书中的案例片段，以通过法律法规和已知判决结果的法律文书中的案例片段辅助待处理文书对判决结果进行解释，增加判决结果的解释内容，提高可解释性和可溯源性。
请参阅图4，其示出了本公开的示例性实施例提供的再一种判决结果处理方法的流程，在图1基础上还可以包括以下步骤：
104：在待处理文书中对关联法条要素的文本内容进行特殊标记，以通过特殊标记形式对关联法条要素的文本内容进行突出显示。
例如对关联法条要素的文本内容采用特定颜色、特效(如加粗和/或倾斜)进行标记，以将关联法条要素的文本内容突出显示，这样人工基于特定标记能够一目了然对应的文本内容，提高获得判决结果的原因的效率。
在本实施例中关联各个法条要素的文本内容对应有文本相关度，相对应的对文本内容进行特殊标记的方式可以是：基于关联法条要素的文本内容的文本相关度，获得该文本内容对应的特殊标记形式，以该文本内容对应的特殊标记形式对该文本内容进行特殊标记。
也就是说具有不同文本相关度的文本内容采用不同的特殊标记形式进行标记，例如文本相关度高的文本内容其采用的特殊标记形式与文本相关度低的文本内容采用的特殊标记形式更能凸显文本相关度高的文本内容，以特定颜色为例，文本相关度高的文本内容采用的特定颜色比文本相关度低的文本内容采用的特定颜色要深。
在本实施例中，文本内容的文本相关度的获得方式是：基于该文本内容中各个词与法条要素的相关度而定，一种方式是：将该文本内容中与法条要素的相关度的取值最高的词的相关度作为该文本内容的相关度，另一种方式是：对该文本内容中各个词与法条要素的相关度进行加权求和得到文本内容的相关度，再一种方式是：对该文本内容中各个词与法条要素的相关度进行平均得到 文本内容的相关度。
与上述方法实施例相对应，本公开的示例性实施例还提供一种判决结果处理装置，其结构如图5所示，可以包括：第一获取单元10、第二获取单元20和关联单元30。
第一获取单元10，用于获得待处理文书对应的至少一个法条要素，并获得待处理文书中与各个法条要素对应的文本内容，待处理文书为待判决案件的文书。
其中待处理文书对应的法条要素是对适用于待处理文书的法律法规的摘要，待处理文书对应的法条要素基于待处理文书中的文本内容获得，例如基于待处理文书中描述案情事实的文本内容获得。在本实施例中，获得待处理文书对应的至少一个法条要素的方式包括但不限于如下方式：
一种方式：获得人工为待处理文书标注的法条要素；
另一种方式：将待处理文书中的文本内容，尤其是描述案情事实的文本内容作为要素识别模型的输入，获得要素识别模型输出的待处理文书对应的法条要素。其中要素识别模型是基于已知判决结果的文书集合中的训练样本和测试样本得到。
对于上述两种获得法条要素的方式可以参阅上述方法实施例，对此本实施例不再阐述。
在本实施例中，第一获得单元10获得待处理文书中与各个法条要素对应的文本内容的可行方式包括但不限于下述方式：
一种方式是：获得待处理文书中各个词分别与各个法条要素的相关度，对各个法条要素中的任一法条要素，从各个词分别与该法条要素的相关度中选取满足第一预设条件的相关度，并基于满足第一预设条件的相关度所属词，从待处理文书获得该法条要素对应的文本内容。
另一种方式：从预设相关词集合中获得与各个法条要素对应的词，获得待处理文书中与各个法条要素对应的词所属句子和与各个法条要素对应的词所属法律法规，将获得的句子和法律法规作为与各个法条要素对应的文本内容。
对于上述两种获得法条要素对应的文本内容的方式可以参阅上述方法实施例，对此本实施例不再阐述。
第二获取单元20，用于基于待处理文书对应的至少一个法条要素，获得待处理文书对应的至少一个判决结果，并从至少一个法条要素中确定各个判决结果对应的法条要素。
在本实施例中获得待处理文书对应的至少一个判决结果的一种方式是：将待处理文书对应的至少一个法条要素作为判决结果预测模型的输入，获得判决结果预测模型输出的待处理文书对应的至少一个判决结果，判决结果预测模型还输出各个法条要素分别与各个判决结果的相关度。
其中判决结果预测模型是将已知判决结果的各个法律文书对应的各个法条要素作为输入，将各个法律文书对应的判决结果以及各个法条要素与各个判决结果的相关度作为输出训练得到，具体请参阅方法实施例中的相关说明。
在本实施例中获得待处理文书对应的至少一个判决结果的另一种方式是：基于记录有法条要素与判决结果的对应关系的专家规则获得待处理文书对应的判决结果，其过程是：从专家规则中获得与待处理文书的法条要素相同的法条要素，将该相同的法条要素对应的判决结果视为是待处理文书对应的判决结果，对于专家规则的建立过程本实施例不在阐述。
而判决结果与法条要素的对应的一种方式是：获得各个法条要素分别与各个判决结果的相关度，并对各个判决结果中的任一判决结果，从各个法条要素分别与该判决结果的相关度中选取满足第一预设条件的相关度，将满足第一预设条件的相关度所属法条要素确定为该判决结果对应的法条要素。
其中各个法条要素分别与各个判决结果的相关度可由判决结果预测模型输出，与判决结果的相关度越高说明该法条要素更能适配该判决结果，该法条要素对应的法律法规更能准确说明该判决结果，为此满足第一预设条件的相关度可以是但不限于是：相关度的取值最高的N个相关度，从而可以将相关度的取值最高的N个相关度所属法条要素确定为该判决结果对应的法条要素，在实际应用中第一预设条件还可以根据实际需求而变化，对此本实施例不在阐述。
关联单元30，用于将各个判决结果与各自对应的法条要素对应的文本内容进行关联，以将待处理文书中的文本内容与各个判决结果建立对应关系，使得判决结果与文本内容对应，从而基于有对应关系的文本内容可知判决结果是 由该待处理文书的哪些因素生成，以获知得到该判决结果的原因。例如将待处理文书中与法条要素对应的词所属句子和与各个法条要素对应的词所属法律法规作为与判决结果又对应关系的文本内容，通过待处理文书中与法条要素对应的词所属句子和与各个法条要素对应的词所属法律法规解释判决结果。
借由上述技术方案，本发明提供的判决结果处理方法，对于一个判决文书能够获得待处理文书对应的至少一个法条要素以及获得待处理文书中与各个法条要素对应的文本内容，基于待处理文书对应的至少一个法条要素，获得待处理文书对应的至少一个判决结果，并确定各个判决结果对应的法条要素，然后将各个判决结果与各自对应的法条要素对应的文本内容进行关联，由此建立待处理文书对应的各个判决结果与待处理文书中的文本内容的对应关系，那么对于任一判决结果可以获知是由于待处理文书中的哪部分文本内容导致得到该判决结果，从而通过与判决结果具有对应关系的文本内容来解读该判决结果，以获知得到该判决结果的原因，使得不单单是得到待处理文书对应的判决结果还能够对判决结果的原因进行说明，使得判决结果具有可解释性和可溯源性。
并且本实施例中判决结果的获得以及判决结果与文本内容的关联是基于法条要素这一中间参考，相对于从待处理文书的全部内容得到判决结果来说，提高准确度。
请参阅图6，其示出了本公开的示例性实施例提供的另一种判决结果处理装置，在上述图5基础上还可以包括：第三获取单元40和文本内容获取单元50。
第三获取单元40，用于获得预设要素知识图谱中与待处理文书对应的各个法条要素相同的法条要素，预设要素知识图谱用于以法条要素为节点指示各个法条要素之间的层级关系。
在本实施例中预设要素知识图谱中的每个法条要素是：通过对法律法规进行人工总结得出的与法律法规对应的法条要素，或者是对已知判决结果的各个法律文书基于要素识别模型的获得过程人工从各个法律文书中总结出的法条要素，在获得所有法条要素的情况下由人工给出这些法条要素的层级关系，从而形成要素体系。
在要素体系的基础上将每个法条要素与对应的法律文书中能够解释该判决结果的案例片段和对应的法律法规进行关联，形成预设要素知识图谱，由此该预设要素知识图谱不仅指示各个法条要素之间的层级关系，还能够示出各个法条要素关联的文本内容，例如对应的法律文书中能够解释该判决结果的案例片段和对应的法律法规。
文本内容获取单元50，用于将相同的法条要素关联的法律法规和已知判决结果的法律文书中的案例片段作为待处理文书对应的该法条要素对应的文本内容。
从上述技术方案可知，与判决结果对应的文本内容不单单是待处理文书中的文本内容，还可以是预设要素知识图谱中与待处理文书对应的法条要素相同的法条要素关联的法律法规和已知判决结果的法律文书中的案例片段，以通过法律法规和已知判决结果的法律文书中的案例片段辅助待处理文书对判决结果进行解释，增加判决结果的解释内容，提高可解释性和可溯源性。
此外，本实施例提供的关联单元30还可以在待处理文书中对关联法条要素的文本内容进行特殊标记，以通过特殊标记形式对关联法条要素的文本内容进行突出显示。
在本实施例中关联各个法条要素的文本内容对应有文本相关度，相对应的对文本内容进行特殊标记的方式可以是：基于关联法条要素的文本内容的文本相关度，获得该文本内容对应的特殊标记形式，以该文本内容对应的特殊标记形式对该文本内容进行特殊标记。
所述判决结果处理装置包括处理器和存储器，上述第一获取单元10、第二获取单元20和关联单元30等均作为程序单元存储在存储器中，由处理器执行存储在存储器中的上述程序单元来实现相应的功能。
处理器中包含内核，由内核去存储器中调取相应的程序单元。内核可以设置一个或以上，通过调整内核参数来将待处理文书对应的各个判决结果与各自对应的法条要素对应的文本内容进行关联，从而通过与判决结果具有对应关系的文本内容来解读该判决结果，以获知得到该判决结果的原因，使得不单单是得到待处理文书对应的判决结果还能够对判决结果的原因进行说明，使得判决结果具有可解释性和可溯源性。
本发明实施例提供了一种计算机可读介质，其上存储有计算机程序，该程序被处理器执行时实现所述判决结果处理方法。
本发明实施例提供了一种处理器，所述处理器用于运行程序，其中，所述程序运行时执行所述判决结果处理方法。
本发明实施例提供了一种电子设备，其结构如图7所示，电子设备包括至少一个处理器701、以及与处理器701连接的至少一个存储器702、总线703；其中，处理器、存储器通过总线完成相互间的通信；处理器用于调用存储器中的程序指令，以执行上述的判决结果处理方法。本文中的电子设备可以是服务器、PC、PAD、手机等。
本申请还提供了一种计算机程序产品，当在数据处理设备上执行时，适于执行初始化有如下方法步骤的程序：
获得待处理文书对应的至少一个法条要素，并获得所述待处理文书中与各个所述法条要素对应的文本内容，所述待处理文书为待判决案件的文书；
基于所述待处理文书对应的至少一个法条要素，获得所述待处理文书对应的至少一个判决结果，并从所述至少一个法条要素中确定各个判决结果对应的法条要素；
将各个判决结果与各自对应的法条要素对应的文本内容进行关联，以使各个判决结果与文本内容对应。
优选的，所述获得所述待处理文书中与各个所述法条要素对应的文本内容包括：
获得所述待处理文书中各个词分别与各个所述法条要素的相关度；
对所述各个法条要素中的任一法条要素，从各个词分别与该法条要素的相关度中选取满足第一预设条件的相关度，并基于所述满足第一预设条件的相关度所属词，从所述待处理文书中获得该法条要素对应的文本内容。
优选的，所述获得所述待处理文书中与各个所述法条要素对应的文本内容包括：
从预设相关词集合中获得与各个所述法条要素对应的词，所预设相关词集合中的词是基于已知判决结果的各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度和/或各个第一法律文书中各个词的共现次 数获得，各个第一法律文书对应的各个法条要素以及各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度基于要素识别模型获得，所述要素识别模型是将已知判决结果的各个第二法律文书作为输入，将为各个第二法律文书标注的各个法条要素以及各个第二法律文书中各个词与标注的各个法条要素的相关度作为输出训练得到；
获得所述待处理文书中与各个所述法条要素对应的词所属句子和与各个所述法条要素对应的词所属法律法规，将所获得的句子和法律法规作为与各个法条要素对应的文本内容。
优选的，当在数据处理设备上执行时，还适于执行初始化有如下方法步骤的程序：
获得预设要素知识图谱中与待处理文书对应的各个法条要素相同的法条要素，所述预设要素知识图谱用于以法条要素为节点指示各个法条要素之间的层级关系；
将所述相同的法条要素关联的法律法规和已知判决结果的法律文书中的案例片段作为所述待处理文书对应的该法条要素对应的文本内容。
优选的，所述从所述至少一个法条要素中确定各个判决结果对应的法条要素包括：
获得所述各个法条要素分别与各个判决结果的相关度；
对所述各个判决结果中的任一判决结果，从各个法条要素分别与该判决结果的相关度中选取满足第二预设条件的相关度，将所述满足第二预设条件的相关度所属法条要素确定为该判决结果对应的法条要素。
优选的，所述基于所述待判决文书对应的至少一个法条要素，获得所述待判决文书对应的至少一个判决结果包括：
将所述待判决文书对应的至少一个法条要素作为判决结果预测模型的输入，获得所述判决结果预测模型输出的所述待判决文书对应的至少一个判决结果，所述判决结果预测模型还输出各个法条要素分别与各个判决结果的相关度；
其中所述判决结果预测模型是将已知判决结果的各个法律文书对应的各个法条要素作为输入，将各个法律文书对应的判决结果以及各个法条要素与各 个判决结果的相关度作为输出训练得到。
优选的，当在数据处理设备上执行时，还适于执行初始化有如下方法步骤的程序：在所述待处理文书中对关联所述法条要素的文本内容进行特殊标记。
优选的，关联各个法条要素的文本内容对应有文本相关度；
所述在所述待处理文书中对关联所述法条要素的文本内容进行特殊标记包括：基于关联所述法条要素的文本内容的文本相关度，获得该文本内容对应的特殊标记形式，以该文本内容对应的特殊标记形式对该文本内容进行特殊标记。
本申请是参照根据本申请实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
在一个典型的配置中，设备包括一个或多个处理器(CPU)、存储器和总线。设备还可以包括输入/输出接口、网络接口等。
存储器可能包括计算机可读介质中的非永久性存储器，随机存取存储器(RAM)和/或非易失性内存等形式，如只读存储器(ROM)或闪存(flash RAM)，存储器包括至少一个存储芯片。存储器是计算机可读介质的示例。
计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁带磁磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算机可读介质不包括暂存电脑可读媒 体(transitory media)，如调制的数据信号和载波。
还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括要素的过程、方法、商品或者设备中还存在另外的相同要素。
本领域技术人员应明白，本申请的实施例可提供为方法、系统或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
以上仅为本申请的实施例而已，并不用于限制本申请。对于本领域技术人员来说，本申请可以有各种更改和变化。凡在本申请的精神和原理之内所作的任何修改、等同替换、改进等，均应包含在本申请的权利要求范围之内。

Claims (10)

  1. 一种判决结果处理方法，其特征在于，所述方法包括：
    获得待处理文书对应的至少一个法条要素，并获得所述待处理文书中与各个所述法条要素对应的文本内容，所述待处理文书为待判决案件的文书；
    基于所述待处理文书对应的至少一个法条要素，获得所述待处理文书对应的至少一个判决结果，并从所述至少一个法条要素中确定各个判决结果对应的法条要素；
    将各个判决结果与各自对应的法条要素对应的文本内容进行关联，以使各个判决结果与文本内容对应。
  2. 根据权利要求1所述的方法，其特征在于，所述获得所述待处理文书中与各个所述法条要素对应的文本内容包括：
    获得所述待处理文书中各个词分别与各个所述法条要素的相关度；
    对所述各个法条要素中的任一法条要素，从各个词分别与该法条要素的相关度中选取满足第一预设条件的相关度，并基于所述满足第一预设条件的相关度所属词，从所述待处理文书中获得该法条要素对应的文本内容。
  3. 根据权利要求1所述的方法，其特征在于，所述获得所述待处理文书中与各个所述法条要素对应的文本内容包括：
    从预设相关词集合中获得与各个所述法条要素对应的词，所预设相关词集合中的词是基于已知判决结果的各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度和/或各个第一法律文书中各个词的共现次数获得，各个第一法律文书对应的各个法条要素以及各个第一法律文书中各个词与各个第一法律文书对应的各个法条要素的相关度基于要素识别模型获得，所述要素识别模型是将已知判决结果的各个第二法律文书作为输入，将为各个第二法律文书标注的各个法条要素以及各个第二法律文书中各个词与标注的各个法条要素的相关度作为输出训练得到；
    获得所述待处理文书中与各个所述法条要素对应的词所属句子和与各个所述法条要素对应的词所属法律法规，将所获得的句子和法律法规作为与各个法条要素对应的文本内容。
  4. 根据权利要求2或3所述的方法，其特征在于，所述方法还包括：
    获得预设要素知识图谱中与待处理文书对应的各个法条要素相同的法条要素，所述预设要素知识图谱用于以法条要素为节点指示各个法条要素之间的层级关系；
    将所述相同的法条要素关联的法律法规和已知判决结果的法律文书中的案例片段作为所述待处理文书对应的该法条要素对应的文本内容。
  5. 根据权利要求1所述的方法，其特征在于，所述从所述至少一个法条要素中确定各个判决结果对应的法条要素包括：
    获得所述各个法条要素分别与各个判决结果的相关度；
    对所述各个判决结果中的任一判决结果，从各个法条要素分别与该判决结果的相关度中选取满足第二预设条件的相关度，将所述满足第二预设条件的相关度所属法条要素确定为该判决结果对应的法条要素。
  6. 根据权利要求1所述的方法，其特征在于，所述方法还包括：在所述待处理文书中对关联所述法条要素的文本内容进行特殊标记。
  7. 根据权利要求6所述的方法，其特征在于，关联各个法条要素的文本内容对应有文本相关度；
    所述在所述待处理文书中对关联所述法条要素的文本内容进行特殊标记包括：基于关联所述法条要素的文本内容的文本相关度，获得该文本内容对应的特殊标记形式，以该文本内容对应的特殊标记形式对该文本内容进行特殊标记。
  8. 一种判决结果处理装置，其特征在于，所述装置包括：
    第一获取单元，用于获得待处理文书对应的至少一个法条要素，并获得所述待处理文书中与各个所述法条要素对应的文本内容，所述待处理文书为待判决案件的文书；
    第二获取单元，用于基于所述待处理文书对应的至少一个法条要素，获得所述待处理文书对应的至少一个判决结果，并从所述至少一个法条要素中确定各个判决结果对应的法条要素；
    关联单元，用于将各个判决结果与各自对应的法条要素对应的文本内容进行关联，以使各个判决结果与文本内容对应。
  9. 一种电子设备，其特征在于，所述电子设备包括至少一个处理器、以 及与处理器连接的至少一个存储器、总线；其中，所述处理器、所述存储器通过所述总线完成相互间的通信；所述处理器用于调用所述存储器中的程序指令，以执行如权利要求1至7中任一项所述的判决结果处理方法。
  10. 一种计算机可读介质，其上存储有计算机程序，其特征在于，该程序被处理器执行时实现如权利要求1至7中任一所述的判决结果处理方法。
PCT/CN2020/102026 2019-09-25 2020-07-15 一种判决结果处理方法及装置 WO2021057202A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910912134.3 2019-09-25
CN201910912134.3A CN112559754A (zh) 2019-09-25 2019-09-25 一种判决结果处理方法及装置

Publications (1)

Publication Number Publication Date
WO2021057202A1 true WO2021057202A1 (zh) 2021-04-01

Family

ID=75029338

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/102026 WO2021057202A1 (zh) 2019-09-25 2020-07-15 一种判决结果处理方法及装置

Country Status (2)

Country Link
CN (1) CN112559754A (zh)
WO (1) WO2021057202A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222251A (zh) * 2021-05-13 2021-08-06 太极计算机股份有限公司 一种基于案件争议焦点的辅助裁判结果预测方法及系统
CN115062918A (zh) * 2022-05-23 2022-09-16 冶金自动化研究设计院有限公司 一种基于规则引擎和事件上报的板坯质量溯源方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108009299A (zh) * 2017-12-28 2018-05-08 北京市律典通科技有限公司 法律审判业务处理方法和装置
CN108304386A (zh) * 2018-03-05 2018-07-20 上海思贤信息技术股份有限公司 一种基于逻辑规则推断法律文书判决结果的方法及装置
CN109117434A (zh) * 2017-06-23 2019-01-01 北京国双科技有限公司 裁判文书检索方法、装置、存储介质及处理器
CN109241528A (zh) * 2018-08-24 2019-01-18 讯飞智元信息科技有限公司 一种量刑结果预测方法、装置、设备及存储介质
CN109359175A (zh) * 2018-09-07 2019-02-19 平安科技（深圳）有限公司 电子装置、诉讼数据处理的方法及存储介质

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110245337B (zh) * 2018-03-09 2022-11-22 北京国双科技有限公司 一种生成裁判文书中经审理查明段的方法及装置
CN109308355B (zh) * 2018-09-17 2020-03-13 清华大学 法律判决结果预测方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109117434A (zh) * 2017-06-23 2019-01-01 北京国双科技有限公司 裁判文书检索方法、装置、存储介质及处理器
CN108009299A (zh) * 2017-12-28 2018-05-08 北京市律典通科技有限公司 法律审判业务处理方法和装置
CN108304386A (zh) * 2018-03-05 2018-07-20 上海思贤信息技术股份有限公司 一种基于逻辑规则推断法律文书判决结果的方法及装置
CN109241528A (zh) * 2018-08-24 2019-01-18 讯飞智元信息科技有限公司 一种量刑结果预测方法、装置、设备及存储介质
CN109359175A (zh) * 2018-09-07 2019-02-19 平安科技（深圳）有限公司 电子装置、诉讼数据处理的方法及存储介质

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113222251A (zh) * 2021-05-13 2021-08-06 太极计算机股份有限公司 一种基于案件争议焦点的辅助裁判结果预测方法及系统
CN115062918A (zh) * 2022-05-23 2022-09-16 冶金自动化研究设计院有限公司 一种基于规则引擎和事件上报的板坯质量溯源方法

Also Published As

Publication number Publication date
CN112559754A (zh) 2021-03-26

Similar Documents

Publication Publication Date Title
CN107798136B (zh) 基于深度学习的实体关系抽取方法、装置及服务器
CN112860841B (zh) 一种文本情感分析方法、装置、设备及存储介质
WO2021057202A1 (zh) 一种判决结果处理方法及装置
CN109271489B (zh) 一种文本检测方法及装置
CN107193796B (zh) 一种舆情事件检测方法及装置
CN115828112B (zh) 一种故障事件的响应方法、装置、电子设备及存储介质
CN110188357B (zh) 对象的行业识别方法及装置
CN113326380B (zh) 基于深度神经网络的设备量测数据处理方法、系统及终端
US11755766B2 (en) Systems and methods for detecting personally identifiable information
CN113849597B (zh) 基于命名实体识别的违法广告词检测方法
CN112966526A (zh) 一种基于情感词向量的汽车在线评论情感分析方法
CN112257444B (zh) 金融信息负面实体发现方法、装置、电子设备及存储介质
Qamar Bhatti et al. Explicit content detection system: An approach towards a safe and ethical environment
CN112052424A (zh) 一种内容审核方法及装置
CN111274786A (zh) 一种自动量刑的方法和系统
WO2021051957A1 (zh) 司法文本识别方法、文本识别模型获得方法及相关设备
CN117349437A (zh) 基于智能ai的政府信息管理系统及其方法
CN112183060A (zh) 多轮对话系统的指代消解方法
CN116881395A (zh) 一种舆情信息检测方法和装置
CN116521871A (zh) 文件的检测方法、装置、处理器以及电子设备
US20110296345A1 (en) Technique For Determining And Indicating Strength Of An Item In A Weighted List Based On Tagging
CN115455934A (zh) 一种企业多种经营范围识别方法与系统
Aarthy et al. Social media analysis for flood nuggets extraction using relevant post filtration
US20210216721A1 (en) System and method to quantify subject-specific sentiment
CN115017894A (zh) 一种舆情风险识别方法及装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20869016

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20869016

Country of ref document: EP

Kind code of ref document: A1