A legal fact analysis method based on judicial court trial and a trial auxiliary system

Abstract

This invention discloses a legal fact analysis method and a trial assistance system based on judicial trials. The legal fact analysis method includes the following steps: determining the fact type of the legal fact to be proven from the dialogue information of the judicial trial; extracting fact elements from the dialogue information according to the fact type of the legal fact to be proven; obtaining evidentiary materials related to the legal fact to be proven based on the fact elements; and identifying the legal fact to be proven based on the fact elements and the evidentiary materials. This invention also discloses a computing device for performing the above method.

Description

Technical Field

[0001] This invention relates to the field of natural language processing technology, and in particular to a legal fact analysis method and trial assistance system based on judicial trials. Background Technology

[0002] With socio-economic development and improved education levels, people's legal awareness is increasing, leading to a rise in judicial cases. Traditional judicial trials involve the judge questioning both parties and third parties (e.g., the plaintiff and their representatives, the defendant and their representatives, and witnesses), who then answer relevant questions. A trial transcript is created based on the dialogue between the judge, the parties, and the third parties. The judge analyzes the legal facts and elements of the case based on this transcript, combining it with the evidence provided by the parties to conduct a judicial trial. This method of trial is time-consuming, requires judges to constantly review documents, and is relatively resource-intensive and inefficient.

[0003] In view of this, how to provide a legal fact analysis scheme based on judicial trials to improve trial efficiency has become an urgent technical problem to be solved. Summary of the Invention

[0004] Therefore, the present invention provides a legal fact analysis method and trial assistance system based on judicial trials, in an attempt to solve or at least alleviate at least one of the above-mentioned problems.

[0005] According to one aspect of the present invention, a method for analyzing legal facts based on judicial trials is provided, comprising the steps of: determining the fact type of the legal fact to be proven from the dialogue information of the judicial trial; extracting fact elements from the dialogue information according to the fact type of the legal fact to be proven; obtaining evidentiary materials related to the legal fact to be proven according to the fact elements; and identifying the legal fact to be proven according to the fact elements and the evidentiary materials.

[0006] Optionally, in the method according to the invention, the step of determining the fact type of the legal fact to be proved from the dialogue information of a judicial trial includes: encoding the dialogue information to obtain a first semantic vector about the dialogue information; and inputting the first semantic vector into a fact type classifier to obtain the fact type of the dialogue information.

[0007] Optionally, in the method according to the present invention, the step of extracting factual elements from dialogue information includes: encoding the dialogue information to obtain a second semantic vector about the dialogue information; and obtaining factual elements in the dialogue information by entity extraction from the second semantic vector.

[0008] Optionally, in the method according to the invention, the step of obtaining evidentiary materials related to the legal fact to be proved based on factual elements includes: extracting at least one type of evidence from dialogue information; identifying the type of evidence related to the legal fact to be proved from the at least one type of evidence, and obtaining evidentiary materials corresponding to the identified type of evidence.

[0009] Optionally, in the method according to the present invention, the step of extracting at least one type of evidence from the dialogue information includes: encoding the dialogue information to obtain a third semantic vector of the dialogue information; and inputting the third semantic vector into an evidence type classifier to extract at least one type of evidence contained in the dialogue information.

[0010] Optionally, in the method according to the invention, the step of identifying the type of evidence related to the legal fact to be proved from at least one type of evidence and obtaining the evidence material corresponding to the identified type of evidence further includes: determining, from the evidence material corresponding to the identified type of evidence, the evidence material that has passed the three-fold determination as evidence material related to the legal fact to be proved.

[0011] Optionally, in the method according to the present invention, the step of determining the legal facts to be proven based on factual elements and evidentiary materials includes: encoding the dialogue information and the determined evidentiary materials to generate a fourth semantic vector about the dialogue information; and inputting the fourth semantic vector into a fact-determining classifier to obtain the fact-determining result of the legal facts to be proven.

[0012] Optionally, the method according to the present invention further includes the steps of: performing speech recognition on the acquired dialogue information to generate dialogue text; sequentially selecting N sentences from the dialogue text as a dialogue information segment, wherein two adjacent dialogue information segments overlap by M sentences, where N and M are both positive integers.

[0013] Optionally, in the method according to the present invention, the step of encoding the dialogue information to obtain a first semantic vector of the dialogue information further includes: encoding the dialogue information in combination with the speaking object pointed to by each sentence in the dialogue information to obtain a first semantic vector of the dialogue information, wherein the speaking object pointed to by the sentence includes the judge in the judicial trial, both parties and third parties.

[0014] Optionally, in the method according to the present invention, after the step of obtaining factual elements in the dialogue information by entity extraction from the second semantic vector, the method further includes the step of optimizing the obtained factual elements by means of a preset rule base.

[0015] According to another aspect of the present invention, a legal fact analysis apparatus based on judicial trials is provided, comprising: a fact type identification unit, adapted to determine the fact type of the legal fact to be proved from the dialogue information of the judicial trial; a fact element extraction unit, adapted to extract fact elements from the dialogue information according to the fact type of the legal fact to be proved; an evidence material extraction unit, adapted to obtain evidence materials related to the legal fact to be proved according to the fact elements; and a fact determination unit, adapted to determine the legal fact to be proved according to the fact elements and the evidence materials.

[0016] According to another aspect of the present invention, a trial assistance system is also provided, comprising: a legal fact analysis device as described above; an interaction module adapted to receive input from a judge and display the results processed by the fact-finding device to the judge; the legal fact analysis device is further adapted to combine the judge's input to determine whether the legal fact to be proven is true.

[0017] According to another aspect of the invention, a computing device is provided, comprising: at least one processor; and a memory storing program instructions, wherein the program instructions are configured to be executed by the at least one processor, the program instructions including instructions for performing any of the methods described above.

[0018] According to another aspect of the invention, a readable storage medium storing program instructions is provided, which, when read and executed by a computing device, cause the computing device to perform any of the methods described above.

[0019] According to the present invention, key legal facts in judicial trials are automatically identified and presented through algorithms to help judges grasp the complete case context in real time. Compared with traditional judicial trials, this saves judges the steps of searching for documents and can effectively improve trial efficiency. Attached Figure Description

[0020] To achieve the foregoing and related objectives, certain illustrative aspects are described herein in conjunction with the following description and accompanying drawings. These aspects indicate various ways in which the principles disclosed herein may be practiced, and all aspects and their equivalents are intended to fall within the scope of the claimed subject matter. The foregoing and other objectives, features, and advantages of this disclosure will become more apparent from the following detailed description, taken in conjunction with the accompanying drawings. Throughout this disclosure, the same reference numerals generally refer to the same parts or elements.

[0021] Figure 1 A schematic diagram of a trial assistance system 100 according to an embodiment of the present invention is shown;

[0022] Figure 2 A schematic diagram of the interactive interface of a trial assistance system 100 according to an embodiment of the present invention is shown;

[0023] Figure 3 A schematic diagram of a computing device 300 according to an embodiment of the present invention is shown;

[0024] Figure 4 A flowchart illustrating a legal fact analysis method 400 based on a judicial trial according to an embodiment of the present invention is shown.

[0025] Figure 5 A schematic diagram of the structure of a fact type identification model 500 according to an embodiment of the present invention is shown; and

[0026] Figure 6 A flowchart illustrating a legal fact analysis method 600 based on judicial trials according to another embodiment of the present invention is shown. Detailed Implementation

[0027] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0028] Figure 1 A schematic diagram of a trial assistance system 100 according to an embodiment of the present invention is shown. The trial assistance system 100 is applied in judicial trials to provide intelligent trial services, and more specifically, to provide auxiliary services for judges' trial work. According to an embodiment of the present invention, the trial assistance system 100 performs legal fact analysis based on real-time dialogue information between the judge, both parties, and third parties during judicial trials to identify the legal facts to be proven in the case and determine the fact-finding results.

[0029] like Figure 1 As shown, the trial assistance system 100 includes a legal fact analysis device 110 and an interaction module 120. Of course, to facilitate the collection of dialogue information during court proceedings, the trial assistance system 100 may also include a voice acquisition device (not shown) for collecting on-site voice data and a voice processing device (not shown) for performing voice recognition on the voice data. In one embodiment, the trial assistance system 100 collects voice data of the judge's questions and the plaintiff's and defendant's answers to the questions, and performs voice recognition processing on the collected voice data. The collection and recognition of voice data are known in the art, and any known or future known voice acquisition and voice recognition technologies can be combined with the technical solutions of this disclosure to implement the trial assistance system 100; therefore, they will not be elaborated upon here.

[0030] According to an embodiment of the present invention, the legal fact analysis device 110 identifies the legal facts involved in the case based on real-time dialogue information during judicial proceedings and extracts the structured elements of the relevant legal facts (hereinafter referred to as fact elements). A legal fact refers to a phenomenon stipulated by law that can cause the creation, modification, or termination of a legal relationship. Taking a private lending case as an example, a legal fact could be the delivery of funds, and fact elements would involve the amount delivered, the name of the payee, and the payee's account number. Simultaneously, the legal fact analysis device 110 can also analyze the relevant evidence submitted by both parties regarding the legal fact. Finally, based on the above analysis (including the legal facts, fact elements, and related evidence materials identified from the dialogue information), it determines whether the legal fact is true, whether it is supported by evidence, and whether the evidence is sufficient. Generally, if the legal fact identified from the dialogue information is true and the supporting evidence is sufficient, then the legal fact is determined to be true.

[0031] Furthermore, in one embodiment of the present invention, the legal fact analysis device 110 includes at least: a fact type identification unit 112, a fact element extraction unit 114, an evidence material extraction unit 116, and a fact determination unit 118, such as... Figure 1 As shown.

[0032] Specifically, the fact type identification unit 112 determines the fact type of the legal fact to be proven from the dialogue information during the judicial trial. The fact element extraction unit 114 extracts the fact elements from the dialogue information based on the fact type of the legal fact to be proven. The evidence material extraction unit 116 obtains the evidence materials related to the legal fact to be proven based on the fact elements. Finally, the fact determination unit 118 determines the legal fact to be proven based on the fact elements and the evidence materials.

[0033] The interaction module 120 provides an interface for interaction with the user (i.e., the judge). On one hand, through the interaction module 120, the judge can browse case-related content, including the information analyzed by the legal fact analysis device 110, as well as case-related complaint texts, answers, and evidence submitted by both parties, etc., and is not limited to these. On the other hand, when the judge believes that the displayed content contains errors or inaccuracies, the interaction module 120 can also receive input from the judge to modify, delete, or add to the displayed content. This allows the legal fact analysis device 110 to optimize the analysis results based on the judge's input, ultimately determining whether the legal facts to be proven are true.

[0034] Figure 2A schematic diagram of the interactive interface of a trial assistance system 100 according to an embodiment of the present invention is shown. According to an embodiment of the present invention, multiple dialogue segments are sequentially acquired during a single judicial trial. By executing the legal fact analysis method according to the present invention, the fact type, factual elements, and evidentiary materials are determined from each dialogue segment of the judicial trial, and the analysis results are sequentially displayed on the interactive interface, such as... Figure 2 As shown.

[0035] Figure 2 This involves a private lending dispute case. Figure 2 The system contains multiple sub-modules 210, each representing the analysis results of a segment of dialogue information, and from top to bottom includes: fact type, fact elements, and evidence materials. For example... Figure 2 As shown, the types of legal facts involved are, in order: loan agreement, payment, repayment, demand for repayment, guarantee, joint repayment, and statute of limitations. Below each type of fact, the factual elements contained in the corresponding dialogue information are listed. Taking "loan agreement" as an example, the factual elements include: time, principal, interest rate during the period, overdue interest rate, penalty, location, and agreed method. Below the factual elements, the corresponding evidentiary materials are listed. Taking "payment" as an example, the evidentiary materials include: bank receipts and Alipay statements. In embodiments of the present invention, judges can modify and supplement the above analysis results displayed on the interactive interface.

[0036] The interactive interface clearly shows the entire court proceedings. Therefore, the trial assistance system according to this invention can help judges understand the complete case details, identify and present key legal facts during the trial, thereby better assisting judges in mastering the trial process. Furthermore, the system can, through real-time interaction with the judge, complete or correct any incomplete or erroneous information identified by the algorithm. Simultaneously, it can output results to downstream systems, thereby better automating downstream tasks such as predicting trial outcomes, calculating amounts involved in litigation, and automatically generating judgments.

[0037] According to an embodiment of the present invention, the trial assistance system 100 and its components can all be implemented by the computing device 300 as described below. Figure 3 A schematic diagram of a computing device 300 according to an embodiment of the present invention is shown.

[0038] like Figure 3 As shown, in the basic configuration 302, the computing device 300 typically includes a system memory 306 and one or more processors 304. A memory bus 308 can be used for communication between the processors 304 and the system memory 306.

[0039] Depending on the desired configuration, processor 304 can be any type of processor, including but not limited to: microprocessor (µP), microcontroller (µC), digital information processor (DSP), or any combination thereof. Processor 304 may include one or more levels of cache such as L1 cache 310 and L2 cache 312, processor core 314, and registers 316. Example processor core 314 may include an arithmetic logic unit (ALU), floating-point unit (FPU), digital signal processing core (DSP core), or any combination thereof. Example memory controller 318 may be used with processor 304, or in some implementations, memory controller 318 may be an internal part of processor 304.

[0040] Depending on the desired configuration, system memory 306 can be any type of memory, including but not limited to volatile memory (such as RAM), non-volatile memory (such as ROM, flash memory, etc.), or any combination thereof. System memory 306 may include operating system 320, one or more applications 322, and program data 324. In some embodiments, applications 322 may be arranged to execute instructions on the operating system using program data 324 by one or more processors 304.

[0041] The computing device 300 may also include an interface bus 340 that facilitates communication from various interface devices (e.g., output devices 342, peripheral interfaces 344, and communication devices 346) to the basic configuration 302 via a bus / interface controller 330. Example output devices 342 include a graphics processing unit 348 and an audio processing unit 350. They may be configured to facilitate communication with various external devices such as displays or speakers via one or more A / V ports 352. Example peripheral interfaces 344 may include a serial interface controller 354 and a parallel interface controller 356, which may be configured to facilitate communication with external devices such as input devices (e.g., keyboards, mice, pens, voice input devices, touch input devices) or other peripherals (e.g., printers, scanners, etc.) via one or more I / O ports 358. Example communication devices 346 may include a network controller 360, which may be arranged to facilitate communication with one or more other computing devices 362 via a network communication link through one or more communication ports 364.

[0042] A network communication link can be an example of a communication medium. A communication medium can typically be embodied in computer-readable instructions, data structures, or program modules within a modulated data signal, such as a carrier wave or other transmission mechanism, and can include any information delivery medium. A “modulated data signal” can be a signal whose data set, or whose modifications, can be encoded with information within the signal. As a non-limiting example, a communication medium can include wired media such as wired networks or leased lines, and various wireless media including sound, radio frequency (RF), microwave, infrared (IR), or other wireless media. The term “computer-readable medium” as used herein can include both storage media and communication media.

[0043] The computing device 300 can be implemented as a server, such as a file server, database server, application server, and web server, or as a personal computer including desktop and laptop computer configurations. Of course, the computing device 300 can also be implemented as part of a small-sized portable (or mobile) electronic device. In an embodiment of the invention, the computing device 300 is configured to execute the legal fact analysis method according to the invention. Application 322 of the computing device 300 includes a legal fact analysis apparatus 110, which contains a plurality of program instructions that can instruct the processor 304 to execute the legal fact analysis method.

[0044] Figure 4 A flowchart illustrating a legal fact analysis method 400 based on a judicial trial according to an embodiment of the present invention is shown. Method 400 is executed in the aforementioned system 100, particularly the legal fact analysis device 110, as follows: Figure 4 The method 400 begins at step S410.

[0045] In step S410, the type of fact to be proven is determined from the dialogue information of the judicial trial.

[0046] According to an embodiment of the present invention, before performing step S410, the step of obtaining dialogue information of the judicial trial is further included.

[0047] In one embodiment, firstly, speech recognition is performed on the real-time collected dialogue information to generate corresponding dialogue text, which serves as the court transcript. Then, the latest segment of dialogue is extracted from the dialogue text and used as the dialogue information for real-time processing. It should be noted that speech recognition technology (ASR) is a common technique in this field, and therefore will not be elaborated upon further here.

[0048] In this embodiment, the latest dialogue information is obtained from the dialogue text using a sliding window. Specifically, N sentences are selected sequentially from the dialogue text as a segment of dialogue information, with M sentences overlapping between adjacent segments. Both N and M are positive integers. In other words, N represents the window size, and M represents the sliding step size. Assuming that sentences in the courtroom dialogue are sequentially identified by numbers, and a segment of dialogue information is generated every 5 sentences, with 2 sentences overlapping between adjacent segments (i.e., window size 5, sliding step size 2), then it can be represented as {1,2,3,4,5} as one segment of dialogue information, and {3,4,5,6,7} as the next segment. It should be noted that considering the dialogue information may contain some irrelevant "nonsense," N and M can be selected according to the actual situation. In actual courtroom settings, different causes of action, judges, and other factors will affect the setting of N and M, and this embodiment of the invention does not impose any limitations on this.

[0049] After obtaining the dialogue information, each segment of dialogue can be processed using a pre-trained fact type recognition model to obtain fact labels representing the fact type. In some embodiments, the fact type recognition model is based on deep learning methods, such as... Figure 5 The diagram shows a structural schematic of a fact type identification model 500 according to an embodiment of the present invention.

[0050] like Figure 5 As shown, the fact type recognition model 500 includes an encoder 510 and a fact type classifier 520 that are coupled to each other. First, the dialogue information is input into the encoder 510, which encodes each sentence in the dialogue information to obtain a first semantic vector about the dialogue information. Then, the first semantic vector is input into the fact type classifier 520, which processes it and outputs a fact label representing the fact type of the dialogue information.

[0051] In one embodiment, the encoder 510 can employ a time-series-based neural network, such as an RNN network, an LSTM network, or a GRU network. In another implementation, an attention mechanism can be introduced into the encoder 510. Specifically, each sentence is segmented into multiple words for each sentence, and each word is then converted into a word embedding, resulting in a sequence of word vectors related to the dialogue information. Next, each word vector in the word vector sequence is processed in the encoder, generating a corresponding hidden vector. By obtaining the attention weights corresponding to each hidden vector and performing a weighted summation of the hidden vector sequence corresponding to the word vector sequence based on these attention weights, a first semantic vector with attention can be obtained.

[0052] The fact type classifier 520 can be implemented using the Softmax classifier or other well-known classifiers, but is not limited to these. The fact type classifier 520 can classify legal facts. Taking the cause of action of private lending as an example, the fact types include lending agreement, guarantee situation, demand for payment, joint repayment, payment delivery, repayment situation, etc., and each category has a fact label (i.e., category code).

[0053] In one embodiment of the present invention, when encoding each sentence, the encoder 510, in addition to performing word embedding to obtain word vectors, also performs role embedding on the speaker to which each sentence refers, obtaining a vector about the conversational role. The two vectors are then concatenated to obtain the vector for the corresponding sentence. Finally, the vectors of each sentence are concatenated to obtain the first semantic vector for that segment of dialogue information. That is, the encoder 510 combines the speaker to which each sentence refers in the dialogue information to encode that segment of dialogue information, obtaining the first semantic vector for that segment of dialogue information. In judicial proceedings, the speaker of each sentence includes the judge and both parties (the plaintiff and their agents, the defendant and their agents). Of course, it can also include third parties, such as fact witnesses, expert witnesses, etc. Figure 5 The encoder shown has three inputs, which receive the words spoken by the judge, the plaintiff (and his / her agent), and the defendant (and his / her agent) in the dialogue information.

[0054] The fact type classifier 520 processes the first semantic vector in the input and calculates a classification result, which serves as the fact label for the content of that segment. This fact label points to the fact type of the dialogue information.

[0055] The following is a brief explanation of the training process for the fact type recognition model 500. First, multiple dialogues are collected and processed, including speech recognition, to obtain corresponding dialogue information, which serves as training data. Simultaneously, each dialogue is labeled with its corresponding fact label. Typically, the dialogues in the training data should contain as many fact types as possible. Next, this labeled training data is input into a neural network model (i.e., the initial fact type recognition model) for learning. After multiple learning iterations, the finally trained fact type recognition model 500 is obtained.

[0056] Table 1 illustrates an example of a dialogue in a judicial trial. The first column of the table represents the speaker's role (i.e., the person being addressed), and the second column represents the content of the dialogue.

[0057] Table 1. Examples of Dialogue Information

[0058]

[0059] After processing by the Fact Type Recognition Model 500, the dialogue information was analyzed to involve the fact of "payment delivery".

[0060] Subsequently, in step S420, factual elements are extracted from the dialogue information according to the fact type of the legal fact to be proven.

[0061] In an embodiment of the present invention, for each dialogue message, a pre-trained fact element extraction model can be used to process it and extract the fact elements. Fact elements typically include elements such as the time, place, and people involved in the legal event. Taking the dialogue message in Table 1 as an example, the extracted fact elements include at least:

[0062] Delivery date: December 17th

[0063] Payment amount: 30,000 yuan

[0064] Delivery location: B's (defendant's name) home

[0065] Payment method: Cash

[0066] Recipient's Name: B (Defendant's Name)

[0067] Recipient's account: xxxx-xxxx-xxx

[0068] In one embodiment, the fact element extraction model is based on a neural network, including an encoder and an entity extraction layer coupled to each other. The encoder part can refer to the structure of the encoder in the fact type recognition model described above, encoding the dialogue information to obtain a second semantic vector about the dialogue information. Of course, other encoders can also be used, and this embodiment of the invention does not limit this. Then, the second semantic vector is input to the entity extraction layer for entity extraction to extract the fact elements of the dialogue information. The entity extraction layer can obtain fact elements by extracting text fragments from sentences. In this embodiment, the entity extraction layer can use a general NER (Named Entity Recognition) model, such as LSTM, LSTM+CRF, BERT, BERT+CRF, etc., and is not limited to these.

[0069] The training of the fact element extraction model can be referenced in the previous description of the training process for the fact type recognition model 500. Similarly, a neural network model is trained using dialogue information with pre-annotated fact elements, serving as the trained fact element extraction model. Given that model training is a standard technique in deep learning, it will not be elaborated upon here due to space limitations.

[0070] According to other embodiments, after extracting factual elements, the extracted factual elements are further optimized using a preset rule base. The preset rule base may contain some general information, such as city name, time, etc., or it may contain information extracted from dialogue information, including the names of the parties involved (such as plaintiff, defendant, witness, etc.), time, delivery method, etc. The embodiments of the present invention do not limit this.

[0071] In one embodiment, the extracted factual elements are compared with a preset rule base using pattern matching to optimize them. For example, if a factual element extracted from the dialogue information is "Hangzhou City, Jiangsu Province," and the preset rule base indicates that this element is incorrect, it can be deleted or corrected by considering the context. Similarly, if an extracted factual element is a person's name, but the name does not appear in the case, it can be confirmed as erroneous and deleted or modified. Furthermore, if the extracted factual element is the date "December 2002," but the preset rule base indicates that the case occurred in 2020 and does not involve 2002, this element can be optimized. In some preferred embodiments, different optimization methods can be set for each type of factual element; these will not be listed here.

[0072] Subsequently, in step S430, evidentiary materials related to the legal facts to be proven are obtained based on the factual elements.

[0073] In some embodiments, step S430 is implemented through the following two steps.

[0074] The first step is to extract at least one type of evidence from the dialogue information and set the extracted evidence type as L1.

[0075] Types of evidence may include, but are not limited to, indictments, answers, business licenses, ID cards, employment contracts, business registration information, loan agreements, pledge contracts, payment vouchers, logistics tracking information, news reports, etc.

[0076] Similarly, a pre-trained evidence type extraction model is used to extract evidence types from dialogue information. In one embodiment, the evidence type extraction model is based on a neural network and includes an encoder and an evidence type classifier that are coupled to each other. First, the encoder encodes the dialogue information to obtain a third semantic vector of the dialogue information. Then, the third semantic vector is input into the evidence type classifier to extract at least one evidence type contained in the dialogue information. In one embodiment, the structure of the evidence type extraction model can refer to the fact type recognition model described above. The difference is that when training the evidence type extraction model, evidence types are used to label each training data, which will not be elaborated here.

[0077] The second step is to identify the type of evidence relevant to the legal facts to be proven from at least one type of evidence, and to obtain the evidentiary materials corresponding to the identified type of evidence.

[0078] According to embodiments of the present invention, different types of evidence are predefined for different types of facts. Table 2 illustrates several types of evidence corresponding to different types of facts.

[0079] Table 2 Examples of Evidence Types Corresponding to Fact Types

[0080]

[0081] Therefore, based on the predefined facts and the fact types determined in the preceding steps, the type of evidence related to that fact type (denoted as L2) can be identified. Next, the obtained evidence types L1 and L2 are cross-validated to obtain their intersection, which is evidence type L3. Finally, from all the evidence submitted by the parties, evidence belonging to category L3 is selected, i.e., evidence related to the legal facts to be proven.

[0082] Furthermore, considering that in judicial proceedings, judges do not accept all evidence submitted by the parties. Evidence accepted by the judge must meet the "three criteria" of authenticity, legality, and relevance. Therefore, in some embodiments, after obtaining evidence corresponding to the identified evidence type (i.e., evidence belonging to category L3) through the above process, it is necessary to further determine which evidence passes the three criteria and is ultimately accepted as evidence related to the legal facts to be proven. According to embodiments of the present invention, whether evidence meets the three criteria can be determined through interaction with the judge. For example, through an interactive module, evidence belonging to category L3 can be displayed to the judge, who can then determine whether it meets the three criteria and whether to accept the evidence.

[0083] According to other embodiments of the present invention, another legal fact analysis method 600 based on judicial trials is provided. In method 600, after analyzing the dialogue information of judicial trials and obtaining the above-mentioned analysis results, it is further determined whether the legal facts to be proven are true based on the above-mentioned analysis results.

[0084] It should be noted that since steps S610 to S630 in method 600 are the same as steps S410 to S430 mentioned above, they will not be repeated here.

[0085] After completing steps S610 to S630, in step S640, the legal facts to be proven are determined based on the factual elements and the identified evidentiary materials.

[0086] According to one embodiment of the present invention, a pre-trained fact-finding model is used to determine whether the legal facts to be proven are true. The fact-finding model is based on deep learning methods and includes an encoder and a fact-finding classifier that are coupled together.

[0087] The encoder encodes the dialogue information and the identified evidence materials to generate a fourth semantic vector about the dialogue information. That is, unlike the encoder in the fact type recognition model mentioned above, here the dialogue information and evidence materials are input into the encoder together for encoding. In embodiments of the present invention, the encoder can be implemented using known or future-known neural networks such as RNN networks, LSTM networks, and GRU networks. An attention mechanism can also be introduced into the encoder, which will not be elaborated further here.

[0088] Next, the fourth semantic vector is input into the fact-finding classifier to obtain the fact-finding result of the legal facts to be proven. The fact-finding classifier can be implemented using a Softmax classifier or other known classifiers, but is not limited to these. The fact-finding classifier performs a binary classification of the legal facts to be proven, outputting "passed" or "failed". For example, an output of "1" indicates that the fact-finding is passed, and an output of "0" indicates that the fact-finding is failed.

[0089] It should be noted that, like the aforementioned models, the fact-finding model needs to be trained beforehand. For details on the training process of the fact-finding model, please refer to the descriptions of the training processes for other models above; due to space limitations, they will not be repeated here.

[0090] It should be understood that for a detailed description of method 600, please refer to the previous descriptions of system 100 and method 400. Due to space limitations, it will not be repeated here.

[0091] According to the legal fact analysis scheme of this invention, key legal facts in judicial trials are automatically identified and presented through algorithms to help judges grasp the complete case context in real time. Compared with traditional judicial trials, this saves judges from steps such as searching for materials, effectively improving trial efficiency. Simultaneously, it can also output results to downstream tasks (such as trial outcome prediction, calculation of amounts involved in litigation, and automatic generation of judgment documents), facilitating the automation of downstream tasks.

[0092] The above embodiments are applied to an online court hearing system designed for courts. Based on similar principles, the legal fact analysis method of this invention can also be applied to other application scenarios, such as police interrogations and procuratorates. In police or procuratorate interrogation scenarios, interrogation records can be automatically generated and key facts extracted.

[0093] The various techniques described herein can be implemented in combination with hardware or software, or a combination thereof. Thus, the methods and apparatus of the present invention, or certain aspects or portions thereof, can take the form of program code (i.e., instructions) embedded in a tangible medium, such as a removable hard disk, USB flash drive, floppy disk, CD-ROM, or any other machine-readable storage medium, wherein when the program is loaded into and executed by a machine such as a computer, the machine becomes an apparatus for practicing the present invention.

[0094] When the program code is executed on a programmable computer, the computing device generally includes a processor, a processor-readable storage medium (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. The memory is configured to store program code; the processor is configured to execute the method of the present invention according to instructions in the program code stored in the memory.

[0095] By way of example, and not limitation, readable media include readable storage media and communication media. Readable storage media stores information such as computer-readable instructions, data structures, program modules, or other data. Communication media generally embodies computer-readable instructions, data structures, program modules, or other data in the form of modulated data signals such as carrier waves or other transmission mechanisms, and includes any information delivery medium. Any combination of the above is also included within the scope of readable media.

[0096] In the specification provided herein, the algorithms and displays are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used with the examples of this invention. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0097] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0098] Similarly, it should be understood that, in order to streamline this disclosure and aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into this detailed description, wherein each claim itself is a separate embodiment of the invention.

[0099] Those skilled in the art will understand that modules, units, or components of the devices disclosed in the examples herein can be arranged in the devices described in this embodiment, or alternatively, can be located in one or more devices different from the devices in this example. The modules in the foregoing examples can be combined into a single module or, in addition, can be divided into multiple sub-modules.

[0100] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0101] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0102] Furthermore, some of the embodiments described herein are methods or combinations of method elements that can be implemented by a processor of a computer system or by other means of performing the functions. Therefore, a processor having the necessary instructions for implementing the methods or method elements forms means for implementing the methods or method elements. Furthermore, the elements described herein in the apparatus embodiments are examples of means for implementing the functions performed by elements for the purposes of carrying out the invention.

[0103] As used herein, unless otherwise specified, the use of ordinal numbers such as “first,” “second,” “third,” etc., to describe ordinary objects merely indicates different instances of similar objects and is not intended to imply that the objects being described must have a given order in time, space, ordering, or any other manner.

[0104] Although the invention has been described with respect to a limited number of embodiments, those skilled in the art will understand from the foregoing description that other embodiments are conceivable within the scope of the invention described herein. Furthermore, it should be noted that the language used in this specification has been chosen primarily for readability and edibility purposes, and not for the purpose of interpreting or limiting the subject matter of the invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the invention is illustrative rather than restrictive, and the scope of the invention is defined by the appended claims.

Claims

1. A method for analyzing legal facts based on judicial trials, comprising the following steps: Identify the type of fact from the dialogue information during judicial proceedings; Based on the type of fact of the legal fact to be proven, extract factual elements from the dialogue information; Based on the aforementioned factual elements, obtain evidentiary materials related to the legal facts to be proven; The legal facts to be proven are determined based on the stated factual elements and the stated evidentiary materials; in, The method of determining the fact type of the legal fact to be proven from the dialogue information of the judicial trial includes: encoding each sentence in the dialogue information to obtain the word vector corresponding to the sentence, and encoding the speaking object to which each sentence in the dialogue information refers to to obtain the vector about the speaking object; concatenating the word vector corresponding to the sentence and the vector about the speaking object to obtain a first semantic vector; and classifying the first semantic vector to obtain the fact type.

2. The method as described in claim 1, wherein, The step of classifying the first semantic vector to obtain the fact type includes: The first semantic vector is input into the fact type classifier to obtain the fact type of the dialogue information.

3. The method as described in claim 2, wherein, The steps for extracting factual elements from the dialogue information include: The dialogue information is encoded to obtain a second semantic vector about the dialogue information; By extracting entities from the second semantic vector, the factual elements in the dialogue information are obtained.

4. The method of claim 3, wherein, The step of obtaining evidentiary materials related to the legal fact to be proven based on the factual elements includes: Extract at least one type of evidence from the dialogue information; Identify the type of evidence related to the legal fact to be proved from the at least one type of evidence, and obtain the evidentiary materials corresponding to the identified type of evidence.

5. The method of claim 4, wherein, The step of extracting at least one type of evidence from the dialogue information includes: The dialogue information is encoded to obtain a third semantic vector of the dialogue information; The third semantic vector is input into the evidence type classifier to extract at least one type of evidence contained in the dialogue information.

6. The method of claim 4, wherein, The step of identifying the type of evidence related to the legal fact to be proven from the at least one type of evidence, and obtaining evidentiary materials corresponding to the identified type of evidence, further includes: From the evidentiary materials corresponding to the identified evidence types, those that have passed the three-fold verification are determined as evidence materials related to the legal facts to be proven.

7. The method of claim 6, wherein, The steps for determining the legal facts to be proven based on the factual elements and the evidentiary materials include: The dialogue information and the identified evidence materials are encoded to generate a fourth semantic vector about the dialogue information; The fourth semantic vector is input into the fact-finding classifier to obtain the fact-finding result of the legal fact to be proven.

8. The method of claim 1, further comprising the step of: The acquired dialogue information is subjected to speech recognition to generate dialogue text; N sentences are selected sequentially from the dialogue text as a dialogue segment, wherein M sentences overlap between adjacent dialogue segments, and N and M are both positive integers.

9. The method as described in claim 2, wherein the speech recipient referred to by the sentence includes the judge in the judicial trial, both parties, and third parties.

10. The method of claim 3, wherein, After the step of obtaining factual elements in the dialogue information by entity extraction from the second semantic vector, the method further includes the step of: The obtained factual elements are optimized by using a pre-set rule base.

11. A legal fact analysis device based on judicial trials, comprising: The fact type identification unit is suitable for determining the fact type of the legal fact to be proved from the dialogue information in judicial trials; The fact element extraction unit is adapted to extract fact elements from the dialogue information based on the fact type of the legal fact to be proven. The evidence extraction unit is adapted to obtain evidence materials related to the legal facts to be proven based on the factual elements. The fact-finding unit is adapted to determine the legal facts to be proven based on the factual elements and the evidentiary materials. The fact type identification unit is used to determine the fact type of the legal fact to be proven from the dialogue information of the judicial trial through the following steps: encoding each sentence in the dialogue information to obtain the word vector corresponding to the sentence, and encoding the speaking object to which each sentence in the dialogue information refers to to obtain the vector about the speaking object; concatenating the word vector corresponding to the sentence and the vector about the speaking object to obtain a first semantic vector; and classifying the first semantic vector to obtain the fact type.

12. A trial support system, comprising: The legal fact analysis device as described in claim 11; An interactive module is adapted to receive input from the judge and display the results processed by the fact-finding device to the judge; The legal fact analysis device is also adapted to combine the judge's input to determine whether the legal fact to be proved is true.

13. A computing device, comprising: At least one processor; and A memory storing program instructions configured to be executed by the at least one processor, the program instructions including instructions for performing the method as claimed in any one of claims 1-10.

14. A readable storage medium storing program instructions that, when read and executed by a computing device, cause the computing device to perform the method as described in any one of claims 1-10.

Images