CN113779976A

CN113779976A - Judgment rule extraction method, system, device and medium

Info

Publication number: CN113779976A
Application number: CN202111132644.2A
Authority: CN
Inventors: 翁洋; 向迪; 杨帅; 王竹; 李鑫; 刘沛琦; 宋凌波; 其他发明人请求不公开姓名
Original assignee: Chengdu Shuzhilian Technology Co Ltd
Current assignee: Chengdu Shuzhilian Technology Co Ltd
Priority date: 2021-09-27
Filing date: 2021-09-27
Publication date: 2021-12-10
Anticipated expiration: 2041-09-27
Also published as: CN113779976B

Abstract

The invention discloses a judging rule extraction method, a system, a device and a medium, relating to the field of natural language processing information extraction and comprising the following steps: extracting referee causal relationship mentions from a plurality of historical referee documents; extracting and obtaining a referee cause and effect event from the referee cause and effect relationship mention, and constructing a referee cause and effect network based on the referee cause and effect event; carrying out generalized processing on the referee causal network to obtain a general referee rule network; building a prediction model, embedding the general referee rule network into the prediction model to obtain a referee rule extraction model, wherein the input of the referee rule extraction model is a referee document, and the output of the referee rule extraction model is a referee rule; and acquiring a referee document to be processed, inputting the referee document to be processed into the referee rule extraction model, and outputting a referee rule corresponding to the referee document to be processed. The invention realizes the automatic extraction of the judgment rule based on the causal relationship of the event.

Description

Judgment rule extraction method, system, device and medium

Technical Field

The present invention relates to the field of extraction of natural language processing information, and in particular, to a method, a system, an apparatus, and a medium for extracting referee rules.

Background

The judgment rules are formed by analyzing legal problems related to the dispute focus of the case and are established for judgment conclusions, are the core contents and soul of the case, and have inspiring, guiding, standardizing and referencing effects on facts identified and law adaptation of judges in the same case.

The referee rule automatic extraction task is to extract referee rules of court decisions from referee documents. This is a key technology of legal assistance system. On the one hand, a low-cost but high-quality legal advisory service can be provided for the masses who are not familiar with legal terms and complex judicial procedures. On the other hand, the method can be used as a convenient reference for professionals (such as lawyers and judges) to improve the working efficiency of the professionals.

However, the existing automatic extraction method for referee rules only focuses on the cause and effect relationship between specific events, does not find general legal referee rules, and when the referee rules are used for judging and predicting, some problems are encountered: 1) the potential symbolicity of tuple matching or phrase matching greatly limits the flexibility of matching; 2) symbolic forms of causal relationships are difficult to generalize to trial and prediction applications.

Disclosure of Invention

The invention aims to realize automatic extraction of the judge rule based on the event cause and effect relationship, not only can extract the general judge rule in the judge text, but also can embed the general judge rule into a continuous vector space to simplify judge prediction, and the invention is beneficial to legal assistance, judge prediction and the like.

In order to achieve the above object, the present invention provides a referee rule extraction method, comprising:

extracting referee causal relationship mentions from a plurality of historical referee documents;

extracting and obtaining a referee cause and effect event from the referee cause and effect relationship mention, and constructing a referee cause and effect network based on the referee cause and effect event;

carrying out generalized processing on the referee causal network to obtain a general referee rule network;

building a prediction model, embedding the general referee rule network into the prediction model to obtain a referee rule extraction model, wherein the input of the referee rule extraction model is a referee document, and the output of the referee rule extraction model is a referee rule;

and acquiring a referee document to be processed, inputting the referee document to be processed into the referee rule extraction model, and outputting a referee rule corresponding to the referee document to be processed.

The method can obtain a general referee rule network, can embed the general referee rule network into the prediction model to obtain a referee rule extraction model, and extracts the referee rule by using the referee rule extraction model.

Preferably, the extracting and obtaining referee causal relationship from referee documents in the method specifically comprises:

constructing a first extraction rule based on a first factor, a second factor and the priority between the first factor and the second factor, wherein the first factor is a regular expression of a preset causal connecting word in a referee document, and the second factor is sentence statement constraint in the referee document;

and extracting and obtaining referee causal relationship mention from the referee document based on the first extraction rule.

By the method, the judgment causal relationship can be extracted and obtained from the judgment document, and the judgment causal relationship can be conveniently extracted and obtained subsequently.

Preferably, the method for acquiring the first factor includes:

identifying and obtaining a plurality of referee cause-effect pair information from referee documents;

judging whether the causal connection words are accurately used or not based on the judgment causal pair information to obtain a plurality of first causal connection words which are accurately used;

counting the occurrence frequency of each first causal connection word in the referee document, and sequencing the first causal connection words from large to small based on the occurrence frequency to obtain a first sequence;

and extracting the first causal connection words of a plurality of first-ranked bits from the first sequence to obtain the preset causal connection words.

The method comprises the steps of firstly identifying and obtaining a plurality of judgment cause and effect pair information from a judgment document, then obtaining a plurality of first cause and effect connection words which are accurate in use, then counting the occurrence frequency of the first cause and effect connection words in the judgment document, then extracting the first cause and effect connection words with high occurrence frequency, obtaining an accurate regular expression with representative significance through the method, and further enabling the subsequent regular expression to be used to extract and obtain accurate judgment cause and effect relation mentions with the representative significance from the judgment document.

Preferably, in the method, the first causal connectives ranked at the top 4 bits are extracted from the first sequence to obtain the preset causal connectives.

The reason for extracting the top 4 bits of the sequence is to give certain labeled data, and it is found that more than 80% of events in the given data are composed of the four conjuncts, if 3, the problem of coverage rate of the conjuncts of the causal relationship needs to be considered, and 5 are possible, but the extraction difficulty after the extraction needs to be considered, and the subsequent difficulty is increased, so that the extraction of the top 4 bits of the sequence is considered comprehensively and is a better choice.

Preferably, the method for constructing the referee causal network includes:

extracting verbs and nouns from the referee causal relationship mention and obtaining the sequence between the verbs and the nouns;

obtaining referee cause and effect events based on the verbs and the nouns and the sequence between the verbs and the nouns;

constructing the referee cause and effect network based on the referee cause and effect event; wherein the referee cause and effect event corresponds to a node in the referee cause and effect network, and an edge of the referee cause and effect network is associated with the referee cause and effect pair information and points to an effect from a cause.

The judgment cause-and-effect network can be constructed based on the judgment cause-and-effect events in the mode, the judgment cause-and-effect network constructed in the mode can accurately reflect the judgment cause-and-effect events, and the subsequent judgment rules can be accurately extracted conveniently.

Preferably, in the method, the performing a generalized process on the referee cause and effect network to obtain a general referee rule network specifically includes:

generalizing words in the plurality of referee causal events by using a preset dictionary to obtain a plurality of first referee causal events after generalized processing;

counting the occurrence frequency of each first referee causal event in a preset referee document, and sequencing the events from large to small based on the occurrence frequency to obtain a second sequence;

taking the first referee cause-and-effect events of a plurality of first ranked bits in the second sequence as nodes in the general referee rule network; wherein the edges of the general referee cause and effect network are associated with the referee cause and effect pair information, pointing from cause to effect.

In order to make the referee causal network have generality, the referee causal network is subjected to generalization processing, and is convenient to be applied in practice subsequently.

Preferably, the preset dictionary in the method comprises a WordNet dictionary and a VerbNet dictionary, the WordNet dictionary is used for generalizing nouns, and the VerbNet dictionary is used for generalizing verbs.

Preferably, in the method, the general referee rule network is embedded into the prediction model to obtain a referee rule extraction model, which specifically includes:

embedding the general referee rule network into a continuous vector space, and coding by using key attributes of the referee causal event causal relationship, wherein the key attributes comprise: dissymmetry of the referee cause and effect event cause and effect relationship, many-to-many relationship of the referee cause and effect event cause and effect relationship, and transitivity of the referee cause and effect event cause and effect relationship.

The model encodes key attributes of judgment causal relationship, and embeds judgment causal relationship networks into a continuous vector space. Prediction of future events and trial results is expressed as a link prediction task on an embedded referee causality network. By learning and manipulating the potentially continuous representation, the embedding method can greatly improve the flexibility of matching, thereby improving the accuracy of the prediction.

Preferably, the method in the present method further comprises training the referee rule extraction model, wherein the referee rule extraction model passes through an energy function

Learning a cause vector and a result vector embedded in the referee cause-effect pair and a transfer vector { t, tau };

；

c is a cause vector in the judgment cause-effect pair, e is an effect vector in the judgment cause-effect pair, t is a transfer vector from a cause to an effect in the cause-effect relationship, tau is a transfer vector from an effect to a cause in the judgment cause-effect relationship, | c + t-e |₁Representing the loss of the cause mapping to the result, | | e + τ -c | | purple₁Indicating a loss of the mapping of the result to the cause.

Preferably, in the method, the loss function of the referee rule extraction model is as follows:

where { x } is the cause vector and the result vector in the referee cause-effect pair, P⁺For a set of real referee cause-and-effect pairs, P, found in said network of generic referee rules^-Is a damage judge cause-effect pair, γ, constructed by replacing the cause or effect in (c, e)>0 is the boundary separating the real referee cause and effect pair and the damage referee cause and effect pair,

to represent

α is the regularization superparameter.

The invention also provides a referee rule extraction system, which comprises:

the judging causal relationship mention obtaining unit is used for extracting and obtaining judging causal relationship mention from a plurality of historical judging documents;

the judgment cause and effect network construction unit is used for extracting and obtaining judgment cause and effect events from the judgment cause and effect relationship mention and constructing a judgment cause and effect network based on the judgment cause and effect events;

a generalization unit, configured to perform a generalization process on the referee causal network to obtain a general referee rule network;

a referee rule extraction model obtaining unit, configured to construct a prediction model, embed the general referee rule network into the prediction model to obtain a referee rule extraction model, where an input of the referee rule extraction model is a referee document, and an output is a referee rule;

and the judging rule obtaining unit is used for obtaining a judging document to be processed, inputting the judging document to be processed into the judging rule extraction model and outputting a judging rule corresponding to the judging document to be processed.

The invention also provides a referee rule extraction device, which comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, wherein the processor realizes the steps of the referee rule extraction method when executing the computer program.

The present invention also provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the steps of the referee rule extraction method.

One or more technical schemes provided by the invention at least have the following technical effects or advantages:

the invention researches a judge rule network extracted based on the event cause and effect relationship from a judge text and predicts future events and judge results by using the judge rule network. Firstly, a general judgment rule network and a hierarchical judgment cause and effect relationship generation method are provided, and the general judgment rule network is established on a specific judgment cause and effect relationship network. From this network, the trial logic of the court decisions is obtained. In addition, a new dual causal transformation model is designed, which encodes key attributes of the referee causality and embeds the referee causality network into a continuous vector space. Prediction of future events and trial results is expressed as a link prediction task on an embedded referee causality network. By learning and manipulating the potentially continuous representation, the embedding method can greatly improve the flexibility of matching, thereby improving the accuracy of the prediction.

Drawings

The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention;

FIG. 1 is a schematic diagram of a referee rule extraction method;

FIG. 2 is a schematic diagram of the process of building and embedding referee rules;

FIG. 3 is a schematic diagram of the referee rule extraction system.

Detailed Description

In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments of the present invention and features of the embodiments may be combined with each other without conflicting with each other.

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described and thus the scope of the present invention is not limited by the specific embodiments disclosed below.

It should be understood that "system", "device", "unit" and/or "module" as used herein is a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, other words may be substituted by other expressions if they accomplish the same purpose.

As used in this specification and the appended claims, the terms "a," "an," "the," and/or "the" are not intended to be inclusive in the singular, but rather are intended to be inclusive in the plural, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that steps and elements are included which are explicitly identified, that the steps and elements do not form an exclusive list, and that a method or apparatus may include other steps or elements.

The present description uses flowcharts to illustrate operations performed by a system according to embodiments of the present description. It should be understood that the preceding or following operations are not necessarily performed in the exact order in which they are performed. Rather, the various steps may be processed in reverse order or simultaneously. Meanwhile, other operations may be added to the processes, or a certain step or several steps of operations may be removed from the processes.

Example one

Referring to fig. 1, fig. 1 is a schematic diagram of a referee rule extraction method, and a first embodiment provides the referee rule extraction method, which includes:

The following describes the steps of the method in detail:

the present invention aims at automatically extracting referee rules from referee texts. The invention mainly comprises two major parts: general referee rule construction and general referee rule embedding models.

And (3) building a referee rule:

a referee rule based on hierarchical event referee causal relationship is constructed from referee documents, and the construction is completed by the following steps, wherein a specific flow chart is shown in the attached figure 2:

(1) referee cause and effect relationship mentions extraction: this step identifies potential causal pairs from the official document. (cause-effect pairs represent pairs of events with cause-effect relationships consisting of cause effects, such as in referee documents (lie, labor) → (wang, unpaid), such pairs of events with cause-derived effects, called cause-effect pairs.) the frequency with which cause-effect conjunctions are correct (cause-effect conjunctions representing words, phrases, clauses, or conjunctions with cause-effect relationships between sentences, such as because, so, cause, after.) is judged by human annotators, resulting in the four cause-effect conjunctions with the highest frequency. And constructing a rule < a regular expression containing the selected causal connection words, sentence statement constraints, and the priority of the rule when a plurality of rules are matched > to extract referee causal relation mentions.

Wherein the selected causal connectives are the four causal connectives with the highest frequency.

The sentence and sentence constraints are regular expressions containing causal conjunctions, and other grammatical constraints need to be formulated and are artificially formulated according to writing habits. Given a connector "after", we use the regular expressions "after [ sensor 1], [ sensor 2 ]" containing "after" to extract referee causal relations mentions with constraints, where [ sensor 1] cannot begin with numbers, which is a sentence constraint. Because if sentence restriction is not set, the American congress mansion repair project is completed after 2 years, and the sentence restriction is extracted by 'after [ sensor 1], [ sensor 2 ]', but the sentence restriction is not a cause and effect relationship.

When the rules are matched, the rules in the priority of the rules simultaneously satisfy the regular expression and sentence statement constraints, and the more obvious causal relationship is established when the priority selection satisfies the rules. For example, when a sentence contains the causal connection words "bicause" and "after" because "bicause" indicates that the probability of a causal mention is greater than "after", the cause and the result of an event are extracted according to the structure of "bicause" in the subsequent referee causal mention, rather than "after".

The judgment causal relation is referred to and extracted, and the judgment documents are extracted by utilizing causal relation connecting words to contain causal relations. For example, in the 2 nd month and 3 rd of 2019, the original notice and the two defenders are settled together, after settlement, the two defenders still owe the original notice to pay the labor cost of 6700 yuan, and the two defenders issue one piece of 'owing article' to the original notice. After the release of the "debt-article", the two reports are not paid the debt until now, which is prompted by the original notice for many times. After the "debt" is issued, the two reports are urged many times and the debt is not paid so far. "

The reason for selecting the four causal connectives that result in the highest frequency here is: given certain labeled data, it is found that more than 80% of event causal relationships in the given data are composed of the four conjuncts, if the number of the causal relationships is set to 3, the problem of causal relationship conjunct coverage needs to be considered, and 5 are possible, but the extraction difficulty needs to be considered later.

The purpose of constructing the rules to extract referee causal relationship is as follows: after the causal connection words are extracted, the causal connection words are used for extracting sentences containing judgment causal relations, so that regular expressions containing the causal connection words are constructed, sentence constraints are grammatical constraints on the sentences to which the expressions can be applied, and the priorities are the priorities of rules when the rules are matched.

For example, given a connector "after", we use the regular expressions "after [ sensor 1], [ sensor 2 ]" containing "after" to extract causal references with constraints where [ sensor 1] cannot begin with a number (sentence constraints). Therefore, this pattern can match the word "oil price dropped after iraq war", but not the word "repair of american national association building completed after 2 years". It is clear that (oil prices dropped after iraq war) is a cause and effect, but (after 2 years, the american national association building recovery project was completed) is not. After applying the rules, we obtain pairs of causal mentions, one of which is labeled as a cause and the other as an effect.

(2) Extracting a judgment cause-and-effect relationship:

this step extracts referee cause and effect events from the referee cause and effect relationship mentions determined in step (1). And expressing a specific referee cause and effect event by using the original sequence of the verb and the noun (extracting the verb and the noun from the cause and effect mention and the sequence thereof in the cause and effect mention, and obtaining the specific event based on the extracted verb and the noun), and constructing a specific referee cause and effect network after extracting the referee cause and effect event. Each node of the network corresponds to a specific event, an edge is associated with a referee cause-and-effect pair, pointing from cause to effect, wherein an edge refers to the line between two nodes.

In the previous work, the causal event is more represented as a tuple or noun phrase extending from a (subject, predicate, object) triple, and we consider that it is a better choice to simply represent each specific event by the original order of a set of verbs and nouns. The extracting means extracts verbs and nouns by means of a natural language processing part-of-speech tagging and entity recognition NLP tool, and events are established by using the original sequence of the verbs and the nouns. For example, "government takes effective action to maintain school safety," neither the triplet of principal and predicate (government, taken, effective action) nor the noun phrase "school safety" contains complete information about the event. However, if we use verb and noun notation (government, take, action, maintenance, school, security), no important information is lost.

The judgment causal relationship network consists of nodes and edges, wherein the nodes represent judgment causal events and are composed of (verbs, nouns and original sequences of verb nouns), and the edges represent whether causal relationships exist or not; such as (Zhao, Zuo, Wang Yi) → (Zhao Yi, fed in, prison); (qian, kill, grandson) → (qian, sent in, prison) → (qian, prison, run away) → (police, grab, qian); () Representing a node (event), → representing an edge (causal relationship), pointing to the result by the cause.

(3) Generalization of referee cause and effect events: the step is to generalize the referee causal relationship network in the step (2) so as to obtain a general referee rule network. Firstly, words appearing in a specific referee cause and effect event are generalized (generalized corresponding to Generalization) by using WordNet and VerbNet, secondly, frequently co-occurring word pairs are used for representing nodes of a general referee rule network, and then, an edge is created between corresponding rules in the general referee rule network by using the edge of the specific referee cause and effect relationship network.

Where WordNet is for nouns and VerbNet is for verbs, this step eliminates the negative effects of word diversity and can help us find frequent patterns from a large number of specific causal events.

Wherein, the frequency exceeds the preset times, for example, the word of 5 is considered as a frequently co-occurring word; the nodes in the network correspond to events (nouns, verbs and the original sequence of noun nouns) and are word pairs formed by the noun verbs; the frequent co-occurrence is an event which occurs 5 times in a given text, and when the network has a plurality of nodes, the word pairs, namely the events which occur 5 times frequently are put into a general referee rule network.

General referee rule embedding model:

and B, designing a dual causal transformation model, embedding the judgment rule network obtained in the step A into a continuous vector space, and coding by using key attributes of the event causal relationship, namely asymmetry of the event causal relationship, many-to-many relationship of the event causal relationship and transitivity of the event causal relationship. By the energy function:

learning the embedding { x } and the transfer vector { t, τ }, x being c and e in f (c, e). Wherein c is a cause vector (cause), e is a vector representing influence (effect), and can also be understood as an effect, t is a transfer vector from a cause to an effect in a cause-effect relationship, τ is a transfer vector from an effect to a cause in a cause-effect relationship, bidirectional mapping from a cause to an effect in a vector space in a dual cause-effect transfer vector model receives the influence of the cause-effect transfer vector, | | c + t-e | y₁A loss that represents a cause mapping to a result; | e + tau-c | ceiling₁The loss of the result mapped to the cause is represented, both losses are reduced and the energy function is minimized.

Given a true causal pair (c, e), we would like the model to be able to predict the correct events and trial outcomes if cause c or result e were missing. The goal of the training is to learn the energy function f so that it can successfully rank the true pair (c, e) under all other possible pairs, i.e. minimize:

wherein, P⁺Is a set of true causal event pairs, P, found in a referee rule network^-Is a damage pair constructed by replacing the cause or result in (c, e), γ>0 is the boundary separating the true causal pair and the damage pair, and [ x]₊= max (0, x) denotes the positive part of x, α>0 is the regularization superparameter. c' is that after the assumed cause loss, we arbitrarily select other causes to constitute the lossTo (c', e)^’Assuming that the result effect is lost, we arbitrarily choose other results to constitute the lost pair (c, e)^’），f（c’，e^’) Like the above formula definition, | t + τ | | non-woven cells₂Indicating that it is desirable to have the cause-to-result transition vector t differ significantly from the result-to-cause transition vector τ. Intuitively, given a true causal pair (c, e), if cause c or result e is missing, we would like the model to be able to predict the correct event. The training aim is to learn the energy function f, so that the energy function of the true pair (c, e) is smaller than a negative sample; it is a natural realization of the expected criteria for a true causal pair to have lower energy than a damage pair.

The following examples describe the use of the method:

the referee document is:

"the hospital is approved and determined as follows: in 2018, the original report is engaged in the work at the two-quilt department. And on 3 days 2 months in 2019, the original notice and the second defendant are subjected to settlement together, the second defendant owes the original notice for labor fee 6700 yuan after the settlement, and the second defendant issues one debt article to the original notice. After the release of the "debt-article", the two reports are not paid the debt until now, which is prompted by the original notice for many times. "

"the institute considers: the civil activities should obey the principle of honest credit, the primary notice provides labor for the second quilt, and the second quilt should pay the labor fee according to the agreement. The original and the defendant do not agree to pay, the original defendant can require the two defendant to pay at any time, so the court is supported by the litigation request that the original defendant requires the two defendant to pay the arrearage of 6700 Yuan. The second defendant is listened to fulfill the payment responsibility and surely causes the fund occupation loss to the original notice, and the home confirms that the second defendant calculates the fund occupation loss according to the loan market quotation interest rate published by the nationwide bank consummate borrowing center from the day of appeal (5/15/2020). In conclusion, litigation of the original report is requested to be supported by the court. "

According to the eighth, sixty-sixth and one hundred and nine-ninth terms of the contract law of the people's republic of China, the one hundred and forty-fourth terms of the litigation law of the people's republic of China, the following decisions are made: the limitation is reported to Xuxx, Tianxx and pays 6700 yuan of original Notification xx within ten days after the judgment is effective. "

The judgment rule extracted by the method is as follows: (original, labor) → (defended, unpaid) → (defended, owing)

(advertised, paid, labor fee).

Example two

Referring to fig. 3, fig. 3 is a schematic diagram of a referee rule extraction system, in which a second embodiment of the present invention further provides a referee rule extraction system, including:

EXAMPLE III

The third embodiment of the present invention provides a referee rule extraction apparatus, which includes a memory, a processor, and a computer program stored in the memory and operable on the processor, wherein the processor implements the steps of the referee rule extraction method when executing the computer program.

The processor may be a Central Processing Unit (CPU), or other general-purpose processor, a digital signal processor (digital signal processor), an Application Specific Integrated Circuit (Application Specific Integrated Circuit), an off-the-shelf programmable gate array (field programmable gate array) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory can be used for storing the computer program and/or the module, and the processor can realize various functions of the referee rule extraction device in the invention by operating or executing the data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function (such as a sound playing function, an image playing function, etc.), and the like. Further, the memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card, a secure digital card, a flash memory card, at least one magnetic disk storage device, a flash memory device, or other volatile solid state storage device.

Example four

A fourth embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps of the referee rule extraction method.

The referee rule extraction means, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on such understanding, all or part of the flow in the method of implementing the embodiments of the present invention may also be stored in a computer readable storage medium through a computer program, and when the computer program is executed by a processor, the computer program may implement the steps of the above-described method embodiments. Wherein the computer program comprises computer program code, an object code form, an executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying said computer program code, a recording medium, a usb-disk, a removable hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, a point carrier signal, a telecommunications signal, a software distribution medium, etc. It should be noted that the computer readable medium may contain content that is appropriately increased or decreased as required by legislation and patent practice in the jurisdiction.

While the invention has been described with respect to the basic concepts, it will be apparent to those skilled in the art that the foregoing detailed disclosure is only by way of example and not intended to limit the invention. Various modifications, improvements and adaptations to the present description may occur to those skilled in the art, although not explicitly described herein. Such modifications, improvements and adaptations are proposed in the present specification and thus fall within the spirit and scope of the exemplary embodiments of the present specification.

Also, the description uses specific words to describe embodiments of the description. Reference throughout this specification to "one embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic described in connection with at least one embodiment of the specification is included. Therefore, it is emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, some features, structures, or characteristics of one or more embodiments of the specification may be combined as appropriate.

Moreover, those skilled in the art will appreciate that aspects of the present description may be illustrated and described in terms of several patentable species or situations, including any new and useful combination of processes, machines, manufacture, or materials, or any new and useful improvement thereof. Accordingly, aspects of this description may be performed entirely by hardware, entirely by software (including firmware, resident software, micro-code, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data block," module, "" engine, "" unit, "" component, "or" system. Furthermore, aspects of the present description may be represented as a computer product, including computer readable program code, embodied in one or more computer readable media.

The computer storage medium may comprise a propagated data signal with the computer program code embodied therewith, for example, on baseband or as part of a carrier wave. The propagated signal may take any of a variety of forms, including electromagnetic, optical, etc., or any suitable combination. A computer storage medium may be any computer-readable medium that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code located on a computer storage medium may be propagated over any suitable medium, including radio, cable, fiber optic cable, RF, or the like, or any combination of the preceding.

Computer program code required for the operation of various portions of this specification may be written in any one or more programming languages, including an object oriented programming language such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C + +, C #, VB.NET, Python, and the like, a conventional programming language such as C, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, a dynamic programming language such as Python, Ruby, and Groovy, or other programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any network format, such as a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet), or in a cloud computing environment, or as a service, such as a software as a service (SaaS).

Additionally, the order in which the elements and sequences of the process are recited in the specification, the use of alphanumeric characters, or other designations, is not intended to limit the order in which the processes and methods of the specification occur, unless otherwise specified in the claims. While various presently contemplated embodiments of the invention have been discussed in the foregoing disclosure by way of example, it is to be understood that such detail is solely for that purpose and that the appended claims are not limited to the disclosed embodiments, but, on the contrary, are intended to cover all modifications and equivalent arrangements that are within the spirit and scope of the embodiments herein. For example, although the system components described above may be implemented by hardware devices, they may also be implemented by software-only solutions, such as installing the described system on an existing server or mobile device.

Similarly, it should be noted that in the preceding description of embodiments of the present specification, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure aiding in the understanding of one or more of the embodiments. This method of disclosure, however, is not intended to imply that more features than are expressly recited in a claim. Indeed, the embodiments may be characterized as having less than all of the features of a single embodiment disclosed above.

For each patent, patent application publication, and other material, such as articles, books, specifications, publications, documents, etc., cited in this specification, the entire contents of each are hereby incorporated by reference into this specification. Except where the application history document does not conform to or conflict with the contents of the present specification, it is to be understood that the application history document, as used herein in the present specification or appended claims, is intended to define the broadest scope of the present specification (whether presently or later in the specification) rather than the broadest scope of the present specification. It is to be understood that the descriptions, definitions and/or uses of terms in the accompanying materials of this specification shall control if they are inconsistent or contrary to the descriptions and/or uses of terms in this specification.

Finally, it should be understood that the embodiments described herein are merely illustrative of the principles of the embodiments of the present disclosure. Other variations are also possible within the scope of the present description. Thus, by way of example, and not limitation, alternative configurations of the embodiments of the specification can be considered consistent with the teachings of the specification. Accordingly, the embodiments of the present description are not limited to only those embodiments explicitly described and depicted herein.

While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims

1. The judgment rule extraction method is characterized by comprising the following steps:

2. The referee rule extraction method according to claim 1, wherein the extraction of referee causal relationship references from referee documents specifically comprises:

3. The referee rule extraction method according to claim 2, wherein the first factor is obtained in a manner comprising:

4. The method according to claim 3, wherein the first causal connectives in the top 4 bits of the sequence are extracted from the first sequence to obtain the predetermined causal connectives.

5. The referee rule extraction method according to claim 1, wherein the referee causal network is constructed in a manner comprising:

6. The method according to claim 5, wherein the generalizing the referee cause and effect network to obtain a general referee rule network comprises:

7. The referee rule extraction method according to claim 6, wherein the preset dictionary comprises a WordNet dictionary and a VerbNet dictionary, the WordNet dictionary is used for generalizing nouns, and the VerbNet dictionary is used for generalizing verbs.

8. The referee rule extraction method according to claim 1, wherein embedding the general referee rule network into the prediction model to obtain a referee rule extraction model specifically comprises:

9. The referee rule extraction method according to claim 8,characterized in that the method further comprises training the referee rule extraction model, wherein the referee rule extraction model passes through an energy function

；

10. The referee rule extraction method according to claim 9, wherein the loss function of the referee rule extraction model is:

to represent

α is the regularization superparameter.

11. Referee rule extraction system, characterized in that the system comprises:

12. An official rule extraction device comprising a memory, a processor and a computer program stored in said memory and executable on said processor, characterized in that said processor implements the steps of the official rule extraction method as claimed in any one of claims 1 to 10 when executing said computer program.

13. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the official rule extraction method as claimed in any one of claims 1 to 10.