Legal reference information providing method, system, device and equipment and storage medium

Abstract

The invention provides a legal reference information providing method, system, device and equipment and a storage medium, and relates to the technical field of data processing, in particular to the technical fields of artificial intelligence, deep learning, intelligent search, computer vision and the like. According to the specific implementation scheme, in response to natural language input in any form provided by a target object, a target legal field needing to be consulted by the target object is determined; determining at least one agent for the target object based on the target legal domain; interacting with a target object on the basis of target knowledge associated with the target legal field on the basis of at least one agent so as to mine details of the case to obtain case description information; and providing legal reference information for the target object based on the case description information.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to technical fields such as artificial intelligence, deep learning, intelligent search, and computer vision. Background Art

[0002] As legal awareness grows, users are increasingly demanding professional legal advice. Traditional methods of consulting with real lawyers typically include offline and online consultations. Whether this is done offline or online, choosing a consulting firm and lawyer presents challenges for users. Furthermore, offline consultations are costly, making online consultations more appealing to users.

[0003] However, due to the lack of fluency in online consultation, the uneven abilities of real lawyers, and communication barriers between users and real lawyers, online consultation also has certain limitations and the consultation effect is poor.

[0004] With the maturity and development of AI (Artificial Intelligence) technology, AI technology has gradually emerged in various industries. It is worth looking forward to how to use AI technology to improve the effectiveness of legal consulting. Summary of the Invention

[0005] The present disclosure provides a method, system, apparatus, device, and storage medium for providing legal reference information.

[0006] According to one aspect of the present disclosure, a legal consulting method is provided, comprising:

[0007] In response to any form of natural language input provided by the target object, determining the target legal field that the target object needs to consult;

[0008] determining at least one agent for the target object based on the target legal domain;

[0009] Based on the target knowledge associated with the target legal field, at least one intelligent agent interacts with the target object to mine case details and obtain case description information;

[0010] Based on the case description information, provide legal reference information to the target object.

[0011] According to another aspect of the present disclosure, there is provided a legal reference information providing system, comprising:

[0012] The middle platform is configured to respond to any form of natural language input provided by a target subject and determine the target legal field that the target subject needs to consult; determine at least one intelligent agent for the target subject based on the target legal field; based on the target knowledge associated with the target legal field, the at least one intelligent agent interacts with the target subject to mine case details to obtain case description information; and provide legal reference information to the target subject based on the case description information;

[0013] At least one intelligent agent, used to generate elements that need to interact with the target object.

[0014] According to another aspect of the present disclosure, there is provided a device for providing legal reference information, comprising:

[0015] A first determination module is configured to determine a target legal field that the target object needs to consult in response to any form of natural language input provided by the target object;

[0016] a second determination module, configured to determine at least one agent for a target object based on the target legal field;

[0017] An interaction module, configured to interact with a target object based on target knowledge associated with a target legal field by at least one intelligent agent, so as to mine case details and obtain case description information;

[0018] The processing module is used to provide legal reference information to the target object based on the case description information.

[0019] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0020] at least one processor; and

[0021] a memory communicatively connected to the at least one processor; wherein,

[0022] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any method in the embodiments of the present disclosure.

[0023] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any method according to the embodiments of the present disclosure.

[0024] According to another aspect of the present disclosure, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the computer program implements any one of the methods according to the embodiments of the present disclosure.

[0025] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0027] Figure 1 This is a schematic diagram of a possible legal intelligent consulting product;

[0028] Figure 2 is an application scenario architecture diagram of a method for providing legal reference information according to an embodiment of the present disclosure;

[0029] Figure 3 is a flowchart of a method for providing legal reference information according to an embodiment of the present disclosure;

[0030] Figure 4 is a schematic diagram of a knowledge tree according to an embodiment of the present disclosure;

[0031] Figure 5 is a background image of a moot court according to an embodiment of the present disclosure;

[0032] Figure 6 is another flowchart of a method for providing legal reference information according to an embodiment of the present disclosure;

[0033] Figure 7 is a schematic diagram of a visual interface for the think tank matching stage according to an embodiment of the present disclosure;

[0034] Figure 8 is a schematic diagram of a visual interface for the pre-court preparation stage according to an embodiment of the present disclosure;

[0035] Figure 9 is a schematic diagram of a visual interface at the trial scene stage according to an embodiment of the present disclosure;

[0036] Figure 10 is a schematic diagram of a method for providing legal reference information according to an embodiment of the present disclosure;

[0037] Figure 11 is a schematic diagram of a platform for applying the method for providing legal reference information according to an embodiment of the present disclosure;

[0038] Figure 12 is a schematic diagram of a legal reference information providing system according to an embodiment of the present disclosure;

[0039] Figure 13 is a schematic diagram of the architecture of a method for providing legal reference information according to an embodiment of the present disclosure;

[0040] Figure 14 is a structural diagram of a legal reference information providing device according to an embodiment of the present disclosure;

[0041] Figure 15 It is a block diagram of an electronic device used to implement the method for providing legal reference information according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0042] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0043] In related technologies, the main path for users to use legal intelligent consulting products is divided into five steps, such as Figure 1 As shown, it includes: user expression, intent understanding, retrieval, reasoning response, and service recommendation.

[0044] The user expression refers to the question that the user wants to ask in the input box by voice or text.

[0045] Intent understanding involves identifying the intent of the user's inquiry and understanding the specific areas of the inquiry. Examples include, but are not limited to, first-level, second-level, and third-level areas. First-level areas are larger, second-level areas are smaller, and third-level areas are the smallest. Examples of first-level areas include legal consultation, second-level areas include marriage and family affairs, and third-level areas include custody disputes.

[0046] Retrieval is to retrieve relevant corpus information from pre-built high-quality legal corpus and searched resource content based on the questions asked by the user, so as to screen out a preset number of corpus contents that best meet the requirements.

[0047] The inference response involves processing and optimizing the previously filtered corpus content to generate prompts that are then fed into the large model. Examples of these prompts include the role (e.g., a lawyer's professional identity), the corpus content, and the user's question. Based on these prompts, the large model summarizes the answer to the user's question, typically a qualitative conclusion, which is output in a typed format.

[0048] Service recommendation means recommending relevant legal services based on the questions asked by users, including professional documents and recommendations of real professional lawyers in the field, to achieve ultimate service integration.

[0049] However, due to the relatively simple expressions of users, for ordinary non-professional users, their expressions are usually vague and even lack a lot of key content, which may lead to inaccurate final reasoning responses and the final consulting effect needs to be improved.

[0050] In view of this, a method for providing legal reference information is proposed in the embodiments of the present disclosure. To facilitate understanding, some terms involved in the embodiments of the present disclosure are explained, including:

[0051] Common Key Elements: Common Key Elements are essential foundational elements of a case; they provide a clear and comprehensive understanding of the basic circumstances of the case. They can be used to formulate a more qualified indictment, improving case handling efficiency. These elements provide a basic understanding of the case and can therefore be considered essential for the issuance of a judgment (or virtual judgment).

[0052] The first professional element is a personalized element inferred by the first intelligent agent from the perspective of the target object based on experience. As a supplement to the general key elements, it can further clarify the case for the target object and help restore the truth of the case. The first professional element mainly focuses on safeguarding the legitimate rights and interests of the target object, thereby improving the accuracy of the legal reference information provided.

[0053] The second professional element, derived from the second agent's experience, takes the perspective of the target's opponent. This personalized element further clarifies the case from their perspective. This complements the first professional element and helps restore the truth of the case. The second professional element primarily focuses on safeguarding the legitimate rights and interests of the target's opponent, thereby improving the accuracy of the provided legal reference information.

[0054] Missing elements: Elements that are identified based on legal reference information (such as virtual judgment documents) and affect the judgment outcome, but are not required to be provided by the target party or the content provided is insufficient.

[0055] Disputed points: These are points where the evidence is insufficient and / or the reasoning is insufficient in the aforementioned general key elements, first professional elements, and second professional elements, which are usually controversial and will affect the quantitative description of the content in the final judgment.

[0056] The first intelligent agent: An intelligent agent built with AI intelligent technology, which can be understood as the defense lawyer of the target object. Standing from the perspective of the target object, the first intelligent agent can use its own knowledge and experience to dig out the truth of the case, with the goal of safeguarding the legitimate rights and interests of the target object as much as possible.

[0057] The second intelligent agent: The intelligent agent constructed by AI intelligent technology can be understood as the defense lawyer of the opponent of the target object. It is used to stand in the perspective of the opponent of the target object, use its own knowledge and experience to dig out the truth of the case, and aim to protect the legitimate rights and interests of the opponent of the target object as much as possible.

[0058] The third intelligent agent is an intelligent agent built by AI intelligent technology (such as a legal assistant), whose main responsibility is to clarify the basic facts of the case and retrieve case-related information.

[0059] Judge agent: An agent built with AI technology, used to infer virtual judgment documents based on simulation results during simulated trials.

[0060] It should be noted that the functions of the aforementioned agents are primary, and the capabilities of each agent can be designed based on specific needs. For example, the capabilities of agents can be combined into a single agent. For example, a first agent can have the capabilities of both the first and third agents, while a judge agent can have the capabilities of all the aforementioned agents.

[0061] Figure 2 An application scenario architecture 200 is shown to which the legal reference information providing method disclosed in the present disclosure can be applied.

[0062] like Figure 2 As shown, scenario architecture 200 may include terminal devices 201, 202, and 203, a network 204, and a server 205. To provide legal reference information, in the disclosed embodiment, a middleware 2051 may be provided in the server 205; the middleware 2051 may interact and collaborate with multiple agents 2052. Each agent 2052 may be provided in the same server 205 as the middleware 2051, or may be installed in another server 205. Network 204 is used to provide a medium for communication links between terminal devices 201, 202, and 203 and server 205. Network 204 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0063] Users can use terminal devices 201, 202, and 203 to interact with server 205 via network 204 to receive or send messages, etc. Various applications for enabling information communication between the terminal devices 201, 202, and 203 and server 205 can be installed, such as complex task processing applications, browser applications, and instant messaging applications.

[0064] Terminal devices 201, 202, 203 and server 205 can be either hardware or software. When terminal devices 201, 202, 203 are hardware, they can be various electronic devices with display screens, including but not limited to smartphones, tablet computers, laptop computers, and desktop computers. When terminal devices 201, 202, 203 are software, they can be installed in the electronic devices listed above. They can be implemented as multiple software programs or software modules, or as a single software program or software module, and are not specifically limited here. When server 205 is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software programs or software modules, or as a single software program or software module, and are not specifically limited here.

[0065] The server 205 can provide various services through various built-in applications. Taking the provision of complex task processing applications as an example, the server 205 can achieve the following effects when running the complex task processing applications: First, the middle platform 2051 receives the natural language input transmitted by the user through the terminal devices 201, 202, and 203 through the network 204, and then the middle platform 2051 coordinates the capabilities of multiple intelligent agents based on the natural language input to interact with the user based on the capabilities of multiple intelligent agents, and further generates legal reference information based on the interaction results.

[0066] Furthermore, the server 205 may also transmit the legal reference information back to the terminal devices 201 , 202 , and 203 via the network 204 , so that the terminal devices 201 , 202 , and 203 display the received legal reference information to the user.

[0067] It should be noted that, in addition to being obtained from terminal devices 201, 202, and 203 via network 204, natural language input can also be pre-stored locally on server 205 in various ways. Therefore, when server 205 detects that such data is already stored locally (for example, when starting to process a previously reserved task), it can choose to directly obtain such data locally. In this case, exemplary scenario architecture 200 may also not include terminal devices 201, 202, 203 and network 204.

[0068] Because processing complex tasks requires significant computational resources and high computing power, the legal reference information providing methods provided in the subsequent embodiments of this disclosure are generally executed by a server 205 possessing significant computational power and resources. Accordingly, the legal reference information providing apparatus is generally also located within server 205. However, it should also be noted that, if terminal devices 201, 202, and 203 also possess sufficient computational power and resources, terminal devices 201, 202, and 203 may also utilize the complex task processing applications installed thereon to complete the aforementioned computations delegated to server 205, thereby outputting the same results as server 205. In particular, in the presence of multiple terminal devices with varying computational capabilities, if the complex task processing application determines that the terminal device it is in possession of possesses significant computational power and abundant remaining computational resources, it may allow the terminal device to perform the aforementioned computations, thereby appropriately alleviating the computational burden on server 205. Accordingly, the legal reference information providing apparatus may also be located within terminal devices 201, 202, and 203. In this case, the exemplary scenario architecture 200 may also not include the server 205 and the network 204 .

[0069] It should be noted that the middle platform 2051 and each intelligent agent 2052 can be installed on the server 205 at the same time, and each intelligent agent 2052 called and controlled by the middle platform 2051 can also be installed on other servers or terminal devices different from the server 205. No specific limitation is made here.

[0070] It should be understood that Figure 2 The numbers of terminal devices, networks, servers, middle platforms, and agents in the examples are only for reference. The corresponding numbers can be set based on the implementation requirements.

[0071] like Figure 3 FIG. 1 is a flow chart of a method for providing legal reference information according to an embodiment of the present disclosure. The flow chart is as follows: Figure 2 The method is executed by the middle station 2051 in the process, and includes the following contents:

[0072] S301 , in response to any form of natural language input provided by a target object, determining a target legal field that the target object needs to consult.

[0073] The target user can be any user seeking advice. This could be an ordinary user who is not a legal professional. The method provided by the disclosed embodiments can provide legal reference information to ordinary users based on their professional legal knowledge, helping them quickly understand the relevant legal circumstances of the case. The target user can also be a lawyer with legal knowledge, helping them understand the possible verdicts of the case and conduct research and learning.

[0074] During implementation, the target user can use any form of natural language input to express the target legal field for consultation. For example, this can be voice input, text input, or input through options provided in the human-computer interaction interface.

[0075] Option input can provide specialized field divisions for the target audience to select. Voice and text input allow the target audience to describe the area of consultation they require based on their own expression habits. When natural language input is provided via voice or text, ERNIE3-medium (Enhanced Representation through Knowledge Integration3-medium) can be used to identify the target legal field covered by the target audience's natural language input.

[0076] S302: Determine at least one intelligent agent for the target object based on the target legal field.

[0077] In the disclosed embodiments, a corresponding intelligent agent can be created based on a real lawyer. Alternatively, a virtual person with certain case handling experience can be created based on the case handling process, and an intelligent agent with corresponding professional knowledge can be created based on this virtual person. Based on this, a suitable intelligent agent can be determined for the target legal field that the target person needs to consult, providing legal reference information services. This determined intelligent agent possesses the corresponding professional experience and can provide relatively reliable legal reference information services.

[0078] S303, based on the target knowledge associated with the target legal field, at least one intelligent agent interacts with the target object to mine case details to obtain case description information.

[0079] The interaction method in the embodiments of the present disclosure can be voice interaction, or a graphical interface and user interaction can be provided.

[0080] During implementation, the middle platform can control the corresponding intelligent agent to interact with the target object, or the intelligent agent can send the content that needs to be interacted to the middle platform, and the middle platform will complete the interaction with the target object.

[0081] S304: Based on the case description information, provide legal reference information to the target object.

[0082] The legal reference information may include possible judgments and action recommendations based on prior knowledge and legal knowledge. It can be generated by the corresponding intelligent agent controlled by the middle platform or by calling the corresponding service.

[0083] In the disclosed embodiment, the needs of the target object are accurately located and the corresponding intelligent agent is matched to interact with the target object based on the professional knowledge and experience of the intelligent agent, so as to guide the target object to provide a relevant description of the case. In the interactive process, the details of the case are fully excavated to obtain case description information, thereby improving the accuracy and rationality of the legal reference information. Through the interactive process, the target object can understand the information elements that need to be paid attention to in the case, and provide the target object with corresponding important legal knowledge through legal reference information. Therefore, the entire consultation process can improve the target object's further understanding and sorting of the case from the perspective of relevant laws, provide guidance for the target object to handle the case offline, and thus improve the user experience of the target object.

[0084] In some embodiments, the legal reference information includes at least one of the following: a virtual legal opinion, a virtual judgment document.

[0085] Among them, the virtual legal opinion refers to a legal opinion issued by at least one of the aforementioned intelligent entities, which provides professional opinions and action suggestions for the legal issues consulted by the target object.

[0086] Virtual legal opinions can provide professional solutions and action recommendations for specific legal issues faced by the target object, helping the target object to clarify what to do next and avoiding the target object from taking wrong or passive actions due to lack of understanding of the law.

[0087] Among them, the virtual judgment document includes the possible judgment results inferred from the case description information of the target object, as well as the probabilities of various judgment results. These judgment results are quantitative descriptions.

[0088] Virtual judgment documents can allow the target party to understand the possible judgment results of similar cases in advance, so that the target party has a clearer understanding of the risks and possible results of handling the case, thereby improving the user experience.

[0089] In the disclosed embodiments, the virtual legal opinions and virtual judgment documents are interpretations and applications of relevant legal provisions and legal principles, which can improve the accuracy of the ultimately generated legal reference information and provide effective legal guidance for the target objects.

[0090] In some embodiments, the elements of the at least one agent's interaction with the target object include at least one of the following:

[0091] 1) Common key elements generated based on public domain legal knowledge in the target knowledge.

[0092] Among them, the common key elements are the essential basic elements of the entire case.

[0093] In the disclosed embodiment, the universal key elements generated based on public domain legal knowledge provide the intelligent agent with a professional knowledge base, enabling it to effectively explore case details when interacting with the target object, thereby improving the accuracy of legal reference information.

[0094] 2) A first professional element generated based on the first private domain legal knowledge in the target knowledge; the first professional element satisfies a first preset condition, and the first preset condition is used to indicate that defending the target object based on the first professional element can be beneficial to safeguarding the legitimate rights and interests of the target object.

[0095] Among them, the first intelligent agent can stand from the perspective of the target object and focus on safeguarding the legitimate rights and interests of the target object through the first professional element.

[0096] In the embodiment of the present disclosure, since the first professional element satisfies the first preset condition, that is, when defending the target object based on this element, it can be beneficial to safeguard its legitimate rights and interests, the knowledge and experience of the first intelligent agent can be effectively utilized in the process of interaction with the target object to improve the effectiveness of legal services.

[0097] 3) A second professional element generated based on the second private domain legal knowledge in the target knowledge; the second professional element satisfies a second preset condition, and the second preset condition is used to indicate that when defending the opponent of the target object based on the second professional element, it can be beneficial to safeguard the legitimate rights and interests of the opponent of the target object.

[0098] Among them, the second intelligent agent can stand from the perspective of the opponent of the target object, focusing on safeguarding the legitimate rights and interests of the opponent of the target object through the second professional elements.

[0099] In the disclosed embodiment, since the second professional element satisfies the second preset condition, that is, when defending the opponent of the target object based on this element, it can be beneficial to safeguard the legitimate rights and interests of the opponent, therefore, in the process of interaction with the target object, the knowledge and experience of the second intelligent agent can be more effectively utilized to restore the truth of the case for the opponent of the target object and improve the service effect.

[0100] As previously explained, target knowledge can include both public and private legal knowledge. Public legal knowledge encompasses factors that influence the outcome of various cases. Therefore, common key elements for interacting with target objects can be determined based on public legal knowledge. Private legal knowledge, as the name suggests, refers to legal knowledge not publicly available and can include the case handling experience of the agent's corresponding management object.

[0101] During implementation, public domain legal knowledge is pre-constructed. The public domain legal knowledge in the target knowledge can be implemented as follows:

[0102] Step A1: Generate a knowledge tree to be optimized based on the target legal provisions in the target legal field and the authoritative interpretation information of the target legal provisions; the knowledge tree to be optimized includes a root node created by the target legal provisions and at least one first-level child node formed by the authoritative interpretation information of the target legal provisions.

[0103] Step A2 supplements the last-level first child node in the knowledge tree to be optimized based on relevant cases of the target legal provision to generate at least one second-level child node corresponding to the last-level first child node, thereby obtaining public legal knowledge represented by the public knowledge tree. The last-level second child node in the public knowledge tree represents the judgment outcome; the path from the root node to the last-level second child node includes the common key elements for obtaining the judgment outcome.

[0104] Legal provisions are the codified, written form of legal norms. A single legal provision may express one or several imperative norms, while a complete, logically structured norm, encompassing both regulatory and protective norms, often consists of several interconnected legal provisions. Legal provisions are the fundamental building blocks of normative legal documents, which are collections of legal provisions. Legal provisions are the basis of law. Legal provisions refer to the original text of specific provisions of relevant laws and regulations that are directly cited during litigation debates.

[0105] Authoritative legal interpretation refers to the explanation of the meaning, content, concepts, terminology and applicable conditions of current legal provisions made by a certain authoritative organization in order to apply and comply with the law based on relevant legal provisions, policies, concepts of fairness and justice, legal theories and practices.

[0106] Legal provisions are the concrete manifestation of legal norms, while legal interpretation further clarifies and explains the meaning of legal provisions. Legal interpretation is of great significance in the application of law, helping legal professionals to understand and apply the law more accurately.

[0107] The target legal provisions in the embodiments of the present disclosure are legal provisions corresponding to the target legal field among numerous legal provisions.

[0108] To facilitate the reasonable and accurate provision of legal reference information, in the disclosed embodiments, the target legal domain can be a subdivided domain, i.e., a subdivided legal domain that cannot be further categorized based on the legal provisions. For example, the target legal domain can be divided based on the legal provisions corresponding to specific crimes, so as to reasonably determine the elements (including common key elements, first professional elements, and second professional elements) that the intelligent agent interacts with the target object.

[0109] Taking the field of traffic accidents as an example, except for the crime of causing traffic accidents, each branch of the knowledge tree that can be constructed for the crime of causing traffic accidents includes corresponding common key elements. Figure 4 As shown in , the root node is traffic accident crime, and its associated element classification includes multiple types. Figure 4 As shown, the first child node of the first level includes: whether there are casualties, whether there are serious injuries, whether there are major losses, and whether it is the main responsibility, based on which a knowledge tree to be optimized is formed.

[0110] On the basis of the knowledge tree to be optimized, the first child node at the last level in the knowledge tree to be optimized is supplemented in combination with relevant cases of the target legal provisions to generate at least one second child node corresponding to the first child node at the last level, such as Figure 4 The "primary responsibility", "secondary responsibility", etc. shown in the figure are all second child nodes. The second child node at the last level corresponds to the judgment result of the second child node of the previous level.

[0111] During implementation, knowledge can be extracted from relevant legal provisions and corresponding authoritative interpretations to form a knowledge tree to be optimized. However, some conclusions in the knowledge tree to be optimized are quite broad. For example, when a sentence of more than three years but less than seven years is possible, the specific sentencing must be determined based on the details of the case. Therefore, when the corresponding authoritative interpretation cannot provide further detailed quantitative descriptions, the knowledge tree to be optimized can be further supplemented and improved through case studies to obtain a public domain knowledge tree.

[0112] On the basis of the public domain knowledge tree of public domain legal knowledge, a semantic understanding model can be adopted, whose input includes the public domain legal knowledge established based on the knowledge tree and the natural language input of the target object. Through the semantic understanding model, the natural language input of the target object is mapped to the corresponding target node to locate the root node and the first child node from the public domain knowledge tree, and then accurately locate the child node (including the first child node and / or the second child node) that needs to be further mined based on the mined case details, and extract the associated knowledge elements that affect the judgment results in the public domain knowledge tree, so as to obtain the common key elements.

[0113] Of course, the target legal field can also be mapped to the public domain knowledge tree through keyword matching, and the node corresponding to the target legal field is obtained as the root node, and the common key elements are mined from the root node to the last-level node.

[0114] In the disclosed embodiment, the target legal provisions and their authoritative interpretations are constructed into a knowledge tree to be optimized, which is then supplemented with relevant cases to generate a more complete public domain knowledge tree. This tree-like structure is intuitive and clear, making it easy to quickly locate and understand the key points of legal knowledge without having to search and reason word by word through large amounts of text. Furthermore, the knowledge tree format allows for rapid navigation from the legal provisions at the root node along different paths to the judgment results relevant to the specific case, providing an effective knowledge foundation for subsequent reasoning of legal reference information, thereby improving the accuracy and rationality of legal reference information.

[0115] Similarly, a private domain knowledge tree can also be created. For example, in the case of the first intelligent agent, as the defense lawyer of the target object, it has certain case handling experience, and this experience can be described using the first private domain legal knowledge. Information extraction can be performed on the first private domain legal knowledge to generate a first private domain knowledge tree containing the first professional factor that affects the judgment result. Similarly, information extraction can be performed on the second private domain legal knowledge of the second intelligent agent to generate a second private domain knowledge tree containing the second professional factor that affects the judgment result. Thus, the first private domain knowledge tree and the second private domain knowledge tree respectively contain professional knowledge points, so that during the consultation process, based on these professional knowledge points, the target object can interact and thus mine personalized information about the case. It can be understood that the root node of the first private domain knowledge tree and the second private domain knowledge tree can be established with the crime name in the same way as the public domain knowledge tree, and multiple levels of child nodes of the root node can be established according to the corresponding private domain knowledge.

[0116] In some embodiments, the aforementioned multiple elements (including the common key elements, the first professional elements, and the second professional elements) can be aggregated by the middle platform to the same intelligent agent, which can interact with the target object under the control of the middle platform. Alternatively, multiple intelligent agents can interact with the target object separately under the control of the middle platform, generating elements generated by each intelligent agent. For example, a third intelligent agent, acting as a paralegal, can interact with the target object using the common key elements under the control of the middle platform; a first intelligent agent can interact with the target object using the first professional elements under the control of the middle platform; and a second intelligent agent can interact with the target object using the second professional elements under the control of the middle platform.

[0117] Accordingly, in some embodiments, the at least one agent determined by the middle station for the target object includes at least one of the following:

[0118] The first intelligent agent is used to generate a first professional element based on the first private domain legal knowledge.

[0119] Among them, the first intelligent agent can act as the defense lawyer of the target object. From the perspective of the target object, the first intelligent agent focuses on using professional knowledge to safeguard the legitimate rights and interests of the target object, thereby completing the service of providing legal reference information.

[0120] The first intelligent agent infers personalized elements based on experience, which serve as a supplement to the universal key elements, and thus obtains the first professional elements.

[0121] Among them, the second intelligent agent is used to generate the second professional elements based on the second private domain legal knowledge.

[0122] The second intelligent agent can act as a defense lawyer for the opponent of the target object. The second intelligent agent stands in the perspective of the opponent of the target object and focuses on using professional knowledge to safeguard the legitimate rights and interests of the opponent of the target object, thereby assisting the target object to complete the case review more comprehensively and complete the service of providing legal reference information.

[0123] The second intelligent agent draws personalized elements based on experience, which serve as a supplement to the universal key elements and the first professional elements, thus obtaining the second professional elements.

[0124] Among them, the third intelligent agent is used to generate common key elements.

[0125] The third agent can act as a paralegal. Based on the public legal knowledge within the target knowledge, it generates common key factors that influence the decision outcome and retrieves relevant knowledge (including legal provisions and relevant cases) to provide to the first and second agents. Common key factors are fundamental to the case and are essential for forming legal reference information. Common key factors often influence the virtual decision outcome in the legal reference information. They are often related to the corresponding legal provisions and can ultimately influence the decision outcome.

[0126] A judge agent is used to generate a virtual judgment document in legal reference information based at least on case description information.

[0127] Among them, the judge agent can play the role of a judge in the court (also known as the presiding judge), and is used to generate virtual judgment documents in the legal reference information based on the statements of the first agent and the second agent in the mock court session.

[0128] Accordingly, generating elements that interact with the target object can be implemented as follows:

[0129] Inputting the first private domain legal knowledge and the public domain legal knowledge into the first intelligent agent to obtain a first professional element generated by the first intelligent agent;

[0130] Inputting the second private domain legal knowledge and the public domain legal knowledge into the second intelligent agent to obtain a second professional element generated by the second intelligent agent;

[0131] Input public domain legal knowledge into the third agent to generate common key elements;

[0132] The middle platform can summarize the first professional elements, the second professional elements and the common key elements to obtain the elements for interacting with the target object.

[0133] In this disclosed embodiment, by controlling the first, second, and third agents to identify elements requiring interaction with the target object from different perspectives, a more comprehensive consideration of case details is achieved, minimizing omissions and deviations, and improving the comprehensiveness and accuracy of the case description. Furthermore, by controlling each agent to perform its specific functions, it fully explores and analyzes the case based on its corresponding legal knowledge, thereby improving the quality of legal reference information and, in turn, enhancing the user experience of the target object.

[0134] In some embodiments, based on target knowledge associated with a target legal field, at least one agent interacts with a target object to mine case details to obtain case description information, which may include the following implementations:

[0135] 1) Based on the interaction between the third agent and the target object, the case description information is obtained.

[0136] The middle platform can summarize common key elements, first professional elements, and second professional elements. Then it controls the third agent and the target object to interact with these summarized elements to generate case description information.

[0137] During implementation, the third agent can generate case description information based on the interaction results and send it to the middle station. Alternatively, the middle station can obtain the interaction results of the third agent and then generate case description information based on the interaction results.

[0138] 2) Based on the interaction between the third agent, the first agent and the target object, case description information is obtained.

[0139] During implementation, the middle platform can control the third agent and the target object to interact with common key elements to generate initial case information. This initial case information can be generated by the third agent based on the interaction results of the common key elements and sent to the middle platform. Alternatively, the middle platform can obtain the common key element interaction results of the third agent and call the corresponding large language model to generate initial case information that is smooth and meets certain formatting requirements. The initial case information can include a case introduction and case reference information retrieved by the third agent based on the case introduction.

[0140] Based on the initial case information, the middle platform controls the first agent to interact with the target object on the basis of the initial case information to exchange the first professional element and the second professional element to generate case description information. Similarly, the first case statement information generated based on the interaction result of the first professional element can be generated by the first agent to better utilize the knowledge and experience of the first agent to accurately generate case description information. The second case statement information generated based on the interaction result of the second professional element can be generated by the second agent. Finally, the middle platform summarizes the first case statement information and the second case statement information to obtain the case description information.

[0141] During implementation, taking the generation of the first case statement as an example, the middle platform controls the first agent to interact with the target object based on the first professional factor, and obtains the first result output by the first agent. The first result is the response obtained by querying the target object based on the first professional factor. The first agent then uses the reasoning and summary model to summarize and analyze the first result, generating a first case statement with clear logic and smooth writing.

[0142] For example, in the case of generating the second case statement, the middle platform controls the first agent to interact with the target object based on the second professional factor, obtaining the second result output by the first agent. The second result is the response obtained by querying the target object based on the second professional factor. The second agent then uses the reasoning summary model to summarize and analyze the second result, generating a second case statement with clear logic and smooth writing.

[0143] 3) Based on the interaction between the third agent, the second agent and the target object, case description information is obtained.

[0144] During implementation, the middle platform can control the third agent and the target object to interact with common key elements to generate initial case information. This initial case information can be generated by the third agent based on the interaction results of the common key elements and sent to the middle platform. Alternatively, the middle platform can obtain the common key element interaction results of the third agent and call the corresponding large language model to generate initial case information that is smooth and meets certain formatting requirements. The initial case information can include a case introduction and case reference information retrieved by the third agent based on the case introduction.

[0145] Based on the initial case information, the middle platform controls the second agent to interact with the target object based on the initial case information, using the first and second professional elements to generate a first case statement corresponding to the first professional element and a second case statement corresponding to the second professional element. Similar second case statements can be generated by the second agent to better leverage the second agent's knowledge and experience to accurately generate case descriptions. The first case statement can be generated by the first agent.

[0146] 4) Based on the interaction between the first agent, the second agent, the third agent and the target object.

[0147] In some embodiments, when the first agent, the second agent, and the third agent interact with the target object, based on the target knowledge associated with the target legal field, the interaction with the target object to mine case details to obtain case description information can be implemented as follows:

[0148] Step B1, control the third agent to interact with the target object based on the common key elements to obtain the initial case information generated by the third agent; the initial case information includes the case introduction and case reference information retrieved based on the third agent.

[0149] Step B2: inputting the initial case information into the first agent to obtain the first case statement information generated by the first agent after interacting with the target object through the first professional element based on at least the initial case information.

[0150] The method for obtaining the first case statement information has been described above and will not be repeated here in the embodiment of the present disclosure.

[0151] Step B3: input the initial case information into the second agent to obtain the second case statement information generated by the second agent after interacting with the target object on the basis of at least the initial case information and the second professional element.

[0152] The case description information includes first case statement information and second case statement information.

[0153] The method for obtaining the second case statement information has been described above and will not be repeated here in the embodiment of the present disclosure.

[0154] During implementation, after obtaining initial case information and the first agent interacting with the target object before the second agent, the middle platform controls the first agent to interact with the target object using the first professional element based on the initial case information to generate the first case statement. The middle platform controls the second agent to interact with the target object using the second professional element based on the initial case information and the first case statement to generate the second case statement, ultimately obtaining the case description information.

[0155] During implementation, after obtaining initial case information and the second agent interacting with the target object before the first agent, the middle platform controls the second agent to interact with the target object based on the initial case information to generate the second case statement. The middle platform controls the first agent to interact with the target object based on the initial case information and the second case statement to generate the first case statement, ultimately obtaining the case description information.

[0156] In the disclosed embodiment, by controlling the interaction between the third agent and the target object, the initial information of the case can be quickly collected and organized, laying the foundation for the subsequent generation of case description information. In addition, by controlling the first agent and the second agent to interact with the target object from different angles, respectively, based on the initial information of the case, it is possible to more comprehensively consider the various possibilities of the case, minimize omissions and deviations, and improve the accuracy of the case statement. In summary, the middle station dispatches multiple agent roles to aggregate the professional experience of multiple agents and sort out a more comprehensive case situation, thereby improving the accuracy of the legal reference information provided to the target object.

[0157] In some embodiments, for any one of the first agent and the second agent, the agent generates corresponding professional elements, which can be specifically implemented as steps C1-C2, including:

[0158] Step C1, mapping the information currently provided by the target object to the set of professional knowledge points corresponding to the private domain legal knowledge corresponding to any intelligent agent.

[0159] The information currently provided by the target subject may only include the target legal field, which is a non-subdividable field and can be used by the first and second agents to mine relevant professional elements based on legal knowledge. Furthermore, the currently provided information may also include the target legal field and / or a supplementary description provided by the target subject, which may be a text or voice segment.

[0160] When any agent is a first agent, the private legal knowledge corresponding to the agent is the first private legal knowledge. The first private legal knowledge includes a first precedent input by a first management object corresponding to the first agent and / or case analysis information of the first precedent by the first management object.

[0161] Each agent can be maintained by a corresponding lawyer or lawyer team to reflect and simulate the professional experience of the lawyer or lawyer team. For a first agent, its first management object is the lawyer or lawyer team corresponding to the first agent.

[0162] Among them, the first case is a specific legal case input by the first management object. The first case may be a case that has been tried and judged through judicial procedures, and usually contains information such as the basic facts of the case, the litigation process, the court's reasons for the judgment, and the final judgment result. The first management object constructs the first intelligent entity based on the representative cases encountered in its legal practice, so that the first intelligent entity can have professional legal knowledge, so that the legal reference information provided in the end is more accurate. The first case exemplarily records in detail the time, place, cause, process, result, and factors of concern of the case. Of course, the processing of relevant information is desensitized in accordance with laws and regulations.

[0163] Case analysis information represents the insights and opinions formed by the first managed entity after an in-depth analysis of the first precedent. This information can include an understanding of the court's reasoning in the precedent, an interpretation of the application of legal provisions to a specific case, and a summary of the key factors that led to the case's success or failure. Thus, through the first private domain legal knowledge, the first agent possesses the first managed entity's case handling experience, knowledge, and unique characteristics.

[0164] In this disclosed embodiment, the first precedent and case analysis information input by the first managed object is equivalent to transferring the first managed object's experience in legal practice to the first agent. This allows the first agent to refer to more actual case details and professional insights when conducting legal analysis, thereby improving the first agent's professional level and further enhancing the target object's interaction experience with the first agent.

[0165] Similarly, when any of the intelligent agents is a second intelligent agent, its corresponding private domain legal knowledge is the second private domain legal knowledge.

[0166] The second private domain legal knowledge corresponding to the second intelligent agent includes the second precedent input by the second management object corresponding to the second intelligent agent, and / or the case analysis information of the second precedent by the second management object.

[0167] For the second intelligent entity, its second management object is the lawyer or lawyer team corresponding to the second intelligent entity.

[0168] The second precedent is a specific legal case input by the second management object. This second precedent can be a case that has been heard and judged through judicial procedures, typically containing information such as the basic facts of the case, the litigation process, the court's reasons for the ruling, and the final judgment. The second management object can be a lawyer or a corresponding legal team. The second management object constructs a second intelligent agent based on representative precedents encountered in their legal practice. The second precedent exemplifies a detailed record of the time, location, cause, process, and outcome of the case. Of course, this information can be desensitized in accordance with laws and regulations.

[0169] Case analysis information is the insights and opinions formed by the second management subject after an in-depth analysis of the second precedent. This information can include an understanding of the court's reasoning in the precedent, an interpretation of the application of legal provisions in a specific case, and a summary of the key factors that led to the case's success or failure.

[0170] In the disclosed embodiment, the second precedent and case analysis information included in the second private domain legal knowledge are based on the perspective of the combination of the second precedent and case analysis information input by the second management object, so that the subsequent second intelligent agent can refer to more actual case details and professional insights when conducting legal analysis, so as to improve the professional level of the second intelligent agent and enhance the user experience.

[0171] In some embodiments, for the target legal field, corresponding professional elements can be mined based on the aforementioned first private domain knowledge tree and second private domain knowledge tree. Based on this, the information currently provided by the target object is mapped to the set of professional knowledge points corresponding to the private domain legal knowledge corresponding to any agent. Specifically, it can be implemented as follows: in the knowledge tree corresponding to the private domain legal knowledge (such as the first private domain knowledge tree or the second private domain knowledge tree), the private domain subtree corresponding to the target legal field is searched to obtain a set of professional knowledge points. For example, the first agent is good at handling cases related to traffic accidents and cases related to marriage. Based on the target legal field provided by the target object, the target legal field can be mapped to the first private domain knowledge tree, and the knowledge tree part of the traffic accident case that matches it can be found as a set of professional knowledge points.

[0172] For the target legal field, corresponding professional elements can also be mined based on the aforementioned first and second private knowledge trees. Based on this, the information currently provided by the target object is mapped to a set of professional knowledge points corresponding to the private legal knowledge corresponding to any agent. Alternatively, this can be implemented as follows: any agent inputs the corresponding private legal knowledge and the information currently provided by the target object into a large inference and summary model. The large inference and summary model then mines knowledge points with similar case details within the private legal knowledge through inference analysis to obtain a set of professional knowledge points.

[0173] The currently provided information may include at least the interaction results of the common key elements. For the first agent, it may further include the second case statement information of the second agent. Similarly, for the second agent, it may further include the first case statement information of the first agent.

[0174] During implementation, for the supplementary information currently provided by the target object, the information currently provided by the target object is mapped to the set of professional knowledge points corresponding to the private domain legal knowledge corresponding to any intelligent agent. It can be implemented as follows: any intelligent agent performs reasoning and analysis on the supplementary information currently provided by the target object and the corresponding private domain legal knowledge based on the reasoning summary model, and maps it to the set of professional knowledge points corresponding to the private domain legal knowledge corresponding to any intelligent agent.

[0175] The reasoning summary large model may include a large model and a retrieval-augmented generation model (RAG).

[0176] Large models are, for example, large language models (LLMs). Large language models refer to a specific type of large model specifically designed for processing text data. These models are natural language processing models based on neural networks that can be used to generate, understand, and process text data. Large language models can have tens of billions of parameters, can generate high-quality text, and can be used for various natural language processing tasks such as question answering, text generation, and dialogue systems.

[0177] Large language models have excellent reasoning capabilities and the ability to learn from small samples. Large language models can be understood on a large scale or trained on a large number of samples. Based on these large models, accurate semantic understanding can be achieved.

[0178] Large models include, for example, Multimodal Large Language Models (MLLMs), which facilitate processing of image information during interaction with target objects.

[0179] The RAG model is a retrieval-augmented generative model that combines the language generation capabilities of the generative model and the external knowledge acquisition capabilities of the retrieval model.

[0180] During implementation, for the supplementary information provided by the target object, the RAG model can be used to query relevant cases and knowledge points from the corresponding private domain legal knowledge base based on the supplementary information, and then the query results and supplementary information are input into the large language model. The large language model performs reasoning and analysis according to the case handling style and logic of any intelligent agent, so as to determine the elements that need to be mined and obtain a set of professional knowledge points.

[0181] Step C2, from the set of professional knowledge points, screen out the knowledge points that match the position of any intelligent agent, and obtain the professional elements corresponding to any intelligent agent; wherein, when any intelligent agent is the first intelligent agent, the position of the first intelligent agent is to safeguard the legitimate rights and interests of the target object; when any intelligent agent is the second intelligent agent, the position of the second intelligent agent is to safeguard the legitimate rights and interests of the opponent of the target object.

[0182] Among them, when any intelligent agent is a first intelligent agent, when it analyzes the problem from the standpoint of the target object, the professional element is the first professional element; when any intelligent agent is a second intelligent agent, the professional element is the second professional element.

[0183] During implementation, the set of professional knowledge points may include the key elements that need to be questioned in the relevant case, but in order to clarify the facts and explore the key information of the positions of all parties, it is necessary to screen out the elements of different positions in order to interact with the target object for information mining.

[0184] In some embodiments, in a knowledge tree corresponding to private domain legal knowledge (such as a first private domain knowledge tree or a second private domain knowledge tree), when a set of professional knowledge points is obtained by searching a private domain subtree portion corresponding to a target legal field, knowledge points that match the position of any agent are screened from the set of professional knowledge points to obtain professional elements corresponding to any agent. This can be implemented as follows: knowledge points that match the position of any agent are screened from the set of professional knowledge points according to preset tags; wherein, when any agent is the first agent, a preset tag is determined based on the role category of the target object, and the preset tag is used to mark the screened knowledge points as satisfying a first preset condition;

[0185] When any agent is the second agent, a preset mark is determined based on the role category of the opponent of the target object, and the preset mark is used to mark that the screened knowledge points meet the second preset condition.

[0186] Among them, the role category is plaintiff or defendant.

[0187] During implementation, when establishing a private domain knowledge tree, the knowledge points that are beneficial to safeguarding the plaintiff's legitimate rights and interests can be marked with a first tag, and the knowledge points that are beneficial to safeguarding the defendant's legitimate rights and interests can be marked with a second tag. The first tag and the second tag can be distinguished by different tags, so as to facilitate the mining of knowledge points that match the position of any intelligent entity through tags.

[0188] For example, you can first determine whether the target object is the plaintiff or the defendant, and then filter the corresponding role's tag in the professional knowledge point set to obtain the knowledge point corresponding to the tag.

[0189] In the disclosed embodiment, based on this method, knowledge points that match the standpoint of any intelligent agent can be quickly screened out to construct a knowledge point set.

[0190] In some embodiments, the information currently provided by the target object and the corresponding private domain legal knowledge can be reasoned and analyzed based on the reasoning summary model. When mapped to the professional knowledge point set corresponding to the private domain legal knowledge corresponding to any intelligent agent, the knowledge points that match the position of any intelligent agent are screened from the professional knowledge point set to obtain the professional elements corresponding to any intelligent agent. Specifically, it can be implemented as follows: based on the reasoning summary model, the knowledge points in the professional knowledge point set are classified to obtain the classification results; the knowledge points that match the position of any intelligent agent are screened from the classification results to obtain the professional elements corresponding to any intelligent agent.

[0191] During implementation, the inference and summary model analyzes the set of professional knowledge points corresponding to each agent and the role of the target object (plaintiff or defendant) to obtain a classification result. The classification results include two categories: those that are beneficial to protecting the legitimate rights and interests of the target object and those that are beneficial to protecting the legitimate rights and interests of the target object's opponent.

[0192] For example, when the target object is the plaintiff, the knowledge points in the classification results that are beneficial to safeguarding the legitimate rights and interests of the target object are knowledge points that safeguard the legitimate rights and interests of the plaintiff, and are determined as professional elements that match the position of any intelligent entity.

[0193] In the disclosed embodiment, a large model of reasoning and summarization is used to classify a set of professional knowledge points, and then the knowledge points that match the position of the intelligent agent are screened out, so that the intelligent agent can process and utilize knowledge more intelligently and efficiently, and improve the automation and intelligence level of knowledge processing.

[0194] In some embodiments, to further improve the effect and quality of interaction with the target object, the case description information in the embodiments of the present disclosure includes information on disputed points that affect the judgment result. The disputed point information includes at least one of the following:

[0195] A) Elements used for qualitative description in common key elements.

[0196] Among them, the elements of the qualitative description are conditions or standards that cannot be measured by numerical values. The qualitative description has the same meaning in the following text, and the embodiments of the present disclosure will not be described one by one in the following text.

[0197] Qualitative descriptive elements within the general key elements: Qualitative descriptive elements are non-quantitative descriptive elements. For example, in some cases, factors influencing the outcome of a judgment need to be proven to be positive or negative. However, there is no clear quantitative standard for determining the positive or negative of these factors. Whether these factors are positive or negative requires specific analysis of the specific case, and is a qualitative, not a quantitative, description. Therefore, these qualitative descriptive elements are often controversial points in relevant judgments.

[0198] B) Elements of the Common Key Elements that are marked as controversial based on historical cases.

[0199] For example, in the process of constructing a public domain knowledge tree, some elements can be marked based on the experience of relevant cases, and elements that are often controversial points can be marked in the public domain knowledge tree to facilitate the extraction of controversial points from the common key elements of the public domain knowledge tree.

[0200] Elements in the common key elements that are marked as controversial points based on historical cases: These refer to common key elements that have often been controversial in similar cases in the past. By analyzing and marking historical cases, these potential controversial points can be identified and paid attention to in advance. In the process of constructing the public domain knowledge tree, the elements that are often controversial points will be marked in the public domain knowledge tree to provide a reference for the analysis of the current case.

[0201] C) Elements used for qualitative description in the first professional element;

[0202] D) the elements marked as controversial in the first professional element;

[0203] Elements marked as controversial points in the first professional elements refer to professional elements that are marked as prone to controversy or require special attention within the professional knowledge of the first intelligent agent. These elements may involve complex legal issues or have the possibility of multiple interpretations.

[0204] E) Elements used for qualitative description in the second professional element;

[0205] F) the elements marked as controversial in the second professional element;

[0206] Elements marked as controversial points in the second professional elements refer to professional elements that are marked as prone to controversy or require special attention within the professional knowledge of the second intelligent agent. These elements may involve complex legal issues or have the possibility of multiple interpretations.

[0207] G) Controlling the first agent and / or the second agent to extract elements constituting dispute points from the supplementary description of the target object.

[0208] During the interaction with the target object, there may be inconsistencies in the description of the target object before and after for the same focus point. This focus point can be used to construct elements of the dispute point.

[0209] During implementation, the description provided by the target object can be analyzed and summarized through the reasoning summary model to extract the focus points to constitute the elements of the controversial points.

[0210] In the disclosed embodiments, by clarifying the dispute point information, the focus of the dispute in the case can be comprehensively and accurately located, providing a clear direction for subsequent legal analysis and judgment prediction, thereby laying a strong foundation for obtaining more accurate legal reference information in the future.

[0211] In some embodiments, for any of the first and second agents, based on the interaction between any of the agents and the target object, it can be implemented as follows: any of the agents inputs the elements of the required questions into the questioning model to generate follow-up questions, and the follow-up questions are used to guide the target object to provide descriptive information of the elements of the required questions; any of the agents inputs the follow-up questions into the questioning option model so that the questioning option model generates multiple follow-up options for the target object to choose based on the follow-up questions; based on any of the agents, the follow-up questions and multiple follow-up options are provided to the target object to obtain the target object's response to the follow-up questions.

[0212] In some embodiments, for any of the first, second, and third agents, based on the interaction between any of the agents and the target object, it can be implemented as follows: any of the agents inputs the elements to be asked into the questioning model to generate follow-up questions, and the follow-up questions are used to guide the target object to provide descriptive information of the elements to be asked; any of the agents inputs the follow-up questions into the questioning option model so that the questioning option model generates multiple follow-up options for the target object to choose based on the follow-up questions; based on any of the agents, the follow-up questions and multiple follow-up options are provided to the target object to obtain the target object's response to the follow-up questions.

[0213] For the first intelligent agent, the elements that need to be questioned are the first professional elements; for the second intelligent agent, the elements that need to be questioned are the second professional elements; for the third intelligent agent, the elements that need to be questioned are the general key elements.

[0214] The follow-up questions generated by the follow-up model can guide the target audience to accurately understand the question, allowing them to provide accurate responses and describe the required elements. The questions generated by the follow-up model can guide the target audience to provide information from a professional perspective and an understanding angle to restore the facts of the case.

[0215] The follow-up model can be the ERNIE SPEED PRO (ERNIE Speed Professional) model. The ERNIE SPEED PRO model analyzes the target subject's input and contextual information, understands the target subject's expression habits, and then generates relevant follow-up questions based on the required elements, naturally guiding the target subject to provide more information.

[0216] During implementation, any intelligent agent can input the required question elements into the question model to generate question questions, and then display the question questions to the target object. The target object interacts with the intelligent agent based on the actual situation.

[0217] Because open-ended responses can lead to biased responses or inability to effectively answer follow-up questions due to differences in expressive ability, follow-up options can be provided to ensure more accurate and efficient responses. This allows participants to make relatively accurate choices based on their memory and understanding of the given options, avoiding inaccurate responses due to individual differences in expression or understanding.

[0218] The large model for follow-up question options can be the ERNIE LITE model. The ERNIE LITE (ERNIE Lightweight) model generates follow-up question options, providing clear choices for the target audience. The ERNIE LITE model analyzes the follow-up questions generated by the large model and generates relevant question options for the target audience to choose from. This helps guide the target audience to provide more information while reducing the complexity of their consultation and interaction.

[0219] For example, when the follow-up question is a binary question, the options provided may be "yes" or "no." The target object may select one of the options based on the actual situation.

[0220] The follow-up question option model can intelligently generate multiple options. For example, for the follow-up question "In a traffic accident, how is the accident liability determined?", the model could generate multiple options such as "1. Plaintiff fully responsible; 2. Defendant fully responsible; 3. Plaintiff and defendant each share 50% responsibility; 4. Plaintiff primarily responsible, defendant secondarily responsible; 5. Plaintiff secondarily responsible, defendant primarily responsible." The target audience can then select based on their specific circumstances.

[0221] In the disclosed embodiment, follow-up questions can be generated based on the follow-up question model, and these questions can accurately target the key information that the target object needs to provide. The follow-up questions are input into the follow-up question option model to generate multiple follow-up question options, providing the target object with a clear range of choices. From generating follow-up questions to providing follow-up question options, and then to obtaining the target object's response, a clear and orderly interaction process is formed. Compared with open-ended questions, follow-up question options in the form of multiple-choice questions allow the target object to respond more quickly. At the same time, for the intelligent agent, it can also obtain the required information more quickly, reducing the time waiting for the target object to answer and improving communication efficiency.

[0222] In some embodiments, when at least one agent includes a first agent and a second agent, determining at least one agent for the target object based on the target legal field can be implemented as follows: based on the target legal field of the target object and / or the location information of the target object, recommending multiple first candidate agents to the target object; wherein each first candidate agent is associated with a professional feature description; determining the agent selected by the target object from the first candidate agents as the first agent; and automatically matching the second agent, the third agent and the judge agent for the target object based on the first agent.

[0223] During implementation, the target legal field to which the case belongs can be determined from the target object's conversation content. Based on the target object's location information (which the target object can choose to provide with knowledge), the intelligent agents corresponding to lawyers who are good at handling the target legal field and located in the area where the target object's location information is located are screened out as the first candidate intelligent agents. Each first candidate intelligent agent can be set with its professional characteristics, including an anthropomorphic image, and its virtual profile. The virtual profile may include case handling style, professional experience, etc. The target object can screen from at least one first candidate intelligent agent and use the selected intelligent agent as the first intelligent agent. During implementation, the target object can select an intelligent agent as the first intelligent agent according to its own needs. Since the second intelligent agent is a lawyer standing on the opposite side of the target object, the second intelligent agent can be automatically matched based on the selected first intelligent agent. For example, an intelligent agent with similar case handling experience as the first intelligent agent can be selected as the second intelligent agent.

[0224] Similarly, a suitable third agent and a judge agent can be automatically matched based on the matched first agent.

[0225] In this disclosed embodiment, multiple first candidate agents are recommended to a target subject based on their target legal field and / or location information. The agent selected by the target subject from the candidate agents is then designated as the first agent. This enables personalized and intelligent allocation of legal service resources, fully respects the target subject's right to choose independently, and improves the pertinence and satisfaction of the service. A second agent is then automatically matched, ensuring the integrity and collaboration of case handling.

[0226] In other embodiments, when the at least one agent includes a first agent, a second agent, and a judge agent, providing legal reference information to a target object based on the case description information includes:

[0227] Step D1, simulate the trial process based on the first agent, the second agent and the judge agent.

[0228] Once at least one agent interacts with the target object and obtains case reference information, the core factors influencing the decision are explored from various perspectives. Disputed points can be identified and then simulated in a trial. This simulated trial primarily involves multiple agents debating these points. For example, a first agent and a second agent can debate these points.

[0229] Step D2: Generate legal reference information based on the simulation results of the mock trial.

[0230] During implementation, a mock trial may include court preparation, court investigation, court debate, and closing statements. For example, during court preparation, it is necessary to verify the presence of the plaintiff, defendant, plaintiff's lawyer, and defendant's lawyer. When the judge inquires about the plaintiff's lawyer's presence, the state machine controls the plaintiff's lawyer's corresponding agent to respond. When the judge inquires about the defendant's lawyer's presence, the state machine controls the defendant's lawyer's corresponding agent to respond. A mock trial can be understood as a virtual trial process. Once the judge's agent outputs a virtual judgment, the corresponding first agent can generate a corresponding virtual legal opinion.

[0231] In the disclosed embodiment, the first intelligent agent, the second intelligent agent and the judge intelligent agent respectively analyze the case from different professional perspectives and standpoints, and can comprehensively consider all aspects of the case. Therefore, during the simulated trial process, each intelligent agent can conduct in-depth discussions and debates on the case, further clarify the core elements of the case description information, and then generate legal reference information based on the simulation results of the simulated trial link, which can provide users with more valuable legal advice.

[0232] In some embodiments, legal reference information is generated based on the simulation results of the simulated trial process, which can be implemented as follows: obtaining similar cases retrieved by the judge agent; inputting the similar cases and simulation results into the judge agent to obtain the virtual judgment documents in the legal reference information output by the judge agent.

[0233] Therefore, for virtual judgment documents, the judge intelligent agent can refer to the judgments of similar cases and conduct reasoning analysis, thereby improving the quality of virtual judgment documents and enhancing user experience.

[0234] During implementation, in order to improve the effectiveness and quality of consultation, a background picture of the mock court is output in the mock trial phase, and the background picture includes at least the character images corresponding to the first intelligent agent, the second intelligent agent and the judge intelligent agent respectively; when any of the first intelligent agent, the second intelligent agent and the judge intelligent agent performs a corresponding operation, a prompt is given based on the corresponding character image in the background picture.

[0235] The background image of the moot court is as follows: Figure 5As shown, the system includes defendant 51, defendant's lawyer 52, plaintiff 53, plaintiff's lawyer 54, and judge 55, and of course, a recorder (not shown). In the case of defendant 51 being the target, defendant's lawyer 52 is the first agent, plaintiff's lawyer 54 is the second agent, plaintiff 53 can be the plaintiff agent, and judge 55 corresponds to the judge agent. When plaintiff's lawyer 54 speaks, a session marker can be displayed near plaintiff's lawyer 54, as shown in Figure 5, to indicate that the second agent corresponding to plaintiff's lawyer 54 is performing the corresponding simulated trial session.

[0236] In the disclosed embodiment, a background image of a simulated courtroom is output, and the background image includes the first agent, the second agent, and the judge agent. This can provide users with an intuitive and realistic legal scenario experience, enhancing their understanding and trust in the case handling process. When any agent performs a corresponding task, prompts are provided in the background image, helping users clearly understand the progress of the current case handling and the working status of each agent, thereby improving user engagement and interactivity. This visual and scenario-based presentation makes complex legal processes more vivid and easy to understand.

[0237] In some embodiments, after obtaining legal reference information, further analysis of the legal reference information can be performed to help further optimize the quality of consultation. This can be specifically implemented as follows:

[0238] Step E1: Analyze the legal reference information and case description information to obtain analysis results.

[0239] The middle platform or any other intelligent agent can analyze the legal reference information using a large inference and summary model to determine the analysis results. The intelligent agent can be at least one of the first intelligent agent, the second intelligent agent, and the judge intelligent agent.

[0240] Step E2: When the analysis result indicates that the case description information includes missing elements that do not interact with the target object, at least one intelligent agent is controlled to interact with the target object based on the missing elements to update the legal reference information.

[0241] Missing elements, as the name suggests, are elements that are not interacted with in at least one link where the intelligent agent interacts with the target object.

[0242] For example, in a traffic accident, if the target is the defendant and the interaction process doesn't require them to provide evidence that the person struck didn't comply with traffic regulations, then this evidence is a missing element and is included in the legal reference information. The first agent can then analyze this missing element and further interact with the target to ask them if they can provide the missing element, thereby updating the legal reference information.

[0243] If the analysis results do not include missing elements, the case interaction process with the target object is completed, that is, the target object obtains legal reference information, and depending on the situation, a corresponding lawyer can be recommended to the target object, or the case can be finally ended.

[0244] In the disclosed embodiments, by analyzing legal reference information, missing elements can be promptly identified. Once these missing elements are identified, interaction with the target object can be conducted based on these missing elements to specifically supplement and improve the legal reference information. The target object can then provide relevant supplementary information, which can then be used to regenerate the legal reference information, thereby improving the completeness and accuracy of the legal reference information.

[0245] The method for providing legal reference information proposed in the embodiment of the present disclosure provides a virtual think tank for the target object. The think tank may include a first agent, a second agent, and a judge agent. Based on the interaction between the think tank and the target object, the middle platform can implement the following Figure 6 As shown:

[0246] S601 : In response to any form of natural language input provided by a target object, determine a target legal field that the target object needs to consult.

[0247] S602 : Recommend multiple first candidate agents to the target object based on the target legal field of the target object and / or the location information of the target object.

[0248] S603: In response to the selection operation of the target object, an agent selected from the first candidate agents is determined as the first agent.

[0249] S604: Automatically match a second agent, a third agent, and a judge agent to the target object based on the first agent.

[0250] Among them, the matched second agent, third agent and judge agent can be output and displayed to the target object for the target object to understand.

[0251] S605, controlling the first intelligent agent to call the third intelligent agent to obtain the common key factors affecting the judgment result generated by the third intelligent agent based on the public domain legal knowledge in the target knowledge.

[0252] Among them, the third intelligent agent can further retrieve relevant cases based on the target legal field and generate corresponding retrieval reports to display to the target object, so that the target object can understand the judgment distribution of similar cases.

[0253] S606, controlling the first intelligent agent to generate a first professional element based on the target legal field and the first private domain legal knowledge, and controlling the second intelligent agent to generate a second professional element based on the second private domain legal knowledge.

[0254] S607, summarize the common key elements, the first professional elements and the second professional elements to obtain an element set, and control the third intelligent agent to interact with the target object based on the element set to sort out the case description information.

[0255] Among them, the third intelligent agent can call the follow-up question model to process the elements required for questioning into follow-up questions described in natural language, and process the follow-up questions into follow-up options based on the follow-up question option model, and then output the follow-up questions and follow-up options to the target object, so that the target object completes the interaction by selecting relevant options, and finally obtains the case description information.

[0256] Before the mock trial, the target can provide additional information about the case via voice or text. The first and / or second agents can then process this information using a large inference and summarization model to identify key elements and identify points of contention.

[0257] When all elements required for questioning have been interacted, the case description information is updated.

[0258] S608: Sort out the case description information to determine the dispute points.

[0259] S609, based on the dispute point information, controls the first agent, the second agent and the judge agent to simulate the trial process; based on the simulation results of the simulated trial process, generates legal reference information.

[0260] S610, when it is determined that there are missing elements based on the legal reference information, the first intelligent agent can be controlled to re-initiate a round of questioning based on the missing elements to supplement the situation of the missing elements, and then re-simulate the trial process to generate legal reference information.

[0261] Of course, it should be noted that the mock trial session can be used as an optional operation. The target object can skip this session and not observe it, and can still obtain legal reference information.

[0262] The method for providing legal reference information proposed in the embodiment of the present disclosure may include three stages: matching think tank, pre-court preparation, and court trial. During the interaction between the target object and the intelligent agent, the stage of progress can be displayed through a visual interface. For example, the visual interface diagram of the matching think tank stage is as follows: Figure 7 As shown in ac. Figure 7The a portion of the image is used to obtain the target legal area of the target subject, such as "marriage issues." The target subject selects the corresponding area to determine the target subject's location information. Furthermore, if the automatically filled location information contains errors, the location information can be corrected to obtain accurate location information. Figure 7 Part b of the figure shows the recommendation of the first candidate agent for the target object, such as the first candidate agents including lawyer Wang, lawyer Zhang, lawyer Chen, etc. In response to the selection of the target object, the first agent is determined, for example, the first agent corresponding to lawyer Wang is selected. Then based on Figure 7 Figure c shows the automatic matching with other agents (such as the paralegal agent, the second agent, and the judge agent in the first agent). After the lawyer think tank is matched, the role of the lawyer think tank and the potential recommendations that may be given can be introduced to the target object.

[0263] When the lawyers are matched, the pre-court preparation phase begins. When interacting with the first agent during the pre-court preparation phase, the visual interface diagram for this phase is as follows: Figure 8 As shown in ad. Figure 8 Part a of the figure shows the interaction between the first agent and the target object, and the responses obtained are the common key elements, the first professional elements and the second professional elements. Figure 8 The paralegal assistant corresponding to the first agent will compile the information obtained above and provide preliminary analysis results and qualitative description of the tendency conclusion. The statistical distribution of similar cases can be displayed to the target object in the form of a chart. Figure 8 The c part of the figure shows that the information obtained above is sorted by the paralegal (i.e. the third agent) corresponding to the first agent and presented to the target object in the form of a pie chart. Figure 8 As shown in Figure d, you can also ask the target party to provide additional information and / or upload relevant evidence documents. This concludes the pre-court preparation phase and allows you to enter the mock trial phase. Once the aforementioned interactions are complete, you will enter the on-site trial phase.

[0264] At the trial stage Figure 9 The process of this link can be carried out according to the actual trial process, or the corresponding links can be simplified. There is no limitation on this during implementation. For example, it can include Figure 9 The formal opening of the court is shown in Figure a. After simulating the real court process, Figure 9 As shown in Figure b, legal reference information is obtained, including virtual legal opinions and virtual judgment documents. The AI prediction report can include possible predicted outcomes and the probabilities of various outcomes.

[0265] The architecture diagram in the embodiment of the present disclosure is as follows: Figure 10 As shown, compared with the related art that only uses the retrieval method, the disclosed embodiment adds operations such as general key element questioning and analysis, personalized information (such as the first professional element, the second professional element) questioning and clarification, and adds a private domain legal knowledge base (a database for storing private domain legal knowledge) and a public domain legal knowledge base (a database for storing public domain legal knowledge). Figure 10 The judge agent is used to perform reasoning, which generates legal reference information to provide quantitative judgment conclusions. In addition, based on this legal reference information, relevant services or reports can be retrieved and recommended to the target object.

[0266] During implementation, a knowledge base can be built and the corresponding agent roles can be optimized using different management objects to provide services to the target objects. These agents can be assigned specific character traits, such as gentle or serious, based on their needs. They can also configure corresponding thinking and logic methods for each management object. For example, for the same case, some agents may prioritize question A, while others may prioritize question B. By customizing the agent's image and leveraging its private and public legal knowledge, professional services can be provided.

[0267] For example, the platform for applying the method for providing legal reference information proposed in the embodiment of the present disclosure can be as follows: Figure 11 As shown, it includes four categories: model base, knowledge management, personalization, and platform support. Among them, at least one intelligent agent corresponding to a lawyer or firm is established in the model base. Figure 11 The model base in can be used for operations such as knowledge retrieval, setting basic workflows, and fine-tuning models. Knowledge retrieval involves retrieving relevant content using the retrieval model during its interaction with the target object.

[0268] The basic workflow is the multiple agents involved in the entire interaction and the execution order of the agents, which controls the execution process of each stage.

[0269] The models targeted in model fine-tuning can be iteratively optimized for follow-up question models, follow-up question option models, reasoning summary models, intent recognition models, etc.

[0270] Figure 11 Knowledge management in this system is used to enable management of corresponding agents by management objects. For example, the first and second management objects can upload cases they have handled to the system / platform, such as the aforementioned private knowledge of lawyers, to optimize the corresponding agents. Another example is accessing publicly available knowledge and case materials online to build public legal knowledge.

[0271] Personalization is used to represent each intelligent agent role, and each can be assigned a corresponding personality. The knowledge, image, and thinking / logic of the personality can be set.

[0272] Platform support is used to place multiple agents on a single platform / system for use. This is equivalent to using a state machine for large model base management, plugin management, and command management.

[0273] In some embodiments, a first agent, a second agent, and a third agent each interact with a target object. To effectively control each agent's interaction with the target object at the appropriate time, the disclosed embodiment may be implemented by initiating a state machine based on the target legal domain to control the timing of the interaction between the first agent, the second agent, and the third agent and the target object.

[0274] During implementation, the platform can control each agent to perform related tasks based on workflows and state machines. For example, the state machine can first schedule a third agent to interact with a target object based on common key elements. The first agent can then interact with the target object based on the first specialized element, and then initiate a second agent to interact with the target object based on the second specialized element. Finally, the judge agent can participate in the simulated trial and generate legal reference information, ensuring a smooth and orderly process.

[0275] In this disclosed embodiment, after determining the target legal domain, a state machine is activated. By controlling the timing of interactions between the first, second, and third agents and the target object, the entire legal consultation process can be optimized. The state machine can rationally arrange the order and content of interactions between the agents based on different legal domains and consultation stages, making the consultation process smoother and more efficient.

[0276] In some embodiments, based on the target legal field, a state machine is started to control the interaction timing of the first intelligent agent and the second intelligent agent with the target object through the state machine. It can be implemented as follows: for any intelligent agent among the first intelligent agent and the second intelligent agent, any intelligent agent marks the interaction progress of the elements that any intelligent agent needs to interact with during the interaction with the target object; when the state machine queries that the interaction progress of any intelligent agent is completed, it is determined that the interaction between any intelligent agent and the target object is completed, and according to the workflow pre-set by the state machine, the interaction timing between the next intelligent agent and the target object is determined.

[0277] For example, according to the pre-set intelligent agent interaction process in the simulated trial phase, the interaction timing of the next intelligent agent to interact with the target object is determined.

[0278] During implementation, the elements that any agent needs to ask about can be stored in the form of a list. When any element is asked about, it will be deleted from the list. When there are no elements to be asked about in the list, the agent interaction progress is determined to be completed.

[0279] In the disclosed embodiment, by starting a state machine based on the legal field and using the state machine to control the timing of the interaction between the first and second agents and the target object, it is possible to achieve automated and standardized management of the legal process. Marking the interaction progress of each agent helps ensure that the completion of tasks in each link is accurately recorded, avoids the omission of important information, and improves the efficiency and quality of legal services. After an agent completes the interaction with the target object, it automatically switches to the next agent according to the process preset by the state machine, ensuring the consistency and orderliness of the entire case handling process, reducing delays or confusion caused by human factors, and improving the user experience.

[0280] Among them, the state machine controls the entire process of legal consultation. Based on the workflow model, the state machine controls the interaction timing between each intelligent agent and the target object, as well as the flow of work nodes within each intelligent agent.

[0281] In some embodiments, when it is determined based on the state machine that case description information has been obtained, the state machine switches to the simulated trial phase.

[0282] During implementation, all elements that need to interact with the target object are interacted with. Once the dispute points and case description information are determined, the state machine is used to switch to the simulated trial phase, that is, the state machine drives the judge agent, the first agent, and the second agent to simulate the trial phase.

[0283] In the disclosed embodiment, when the state machine determines that the case description information has been obtained, it switches to the moot court session, ensuring the smooth progress of the entire legal process and improving the efficiency and quality of legal consultation.

[0284] Based on the same technical concept, such as Figure 12 As shown, the embodiment of the present disclosure further provides a legal reference information providing system 1200, including:

[0285] The middle platform 1201 is configured to respond to any form of natural language input provided by a target subject and determine a target legal field that the target subject needs to consult; determine at least one intelligent agent for the target subject based on the target legal field; interact with the target subject based on the target knowledge associated with the target legal field by the at least one intelligent agent to mine case details to obtain case description information; and provide legal reference information to the target subject based on the case description information;

[0286] At least one intelligent agent 1202 is used to generate elements required to interact with the target object.

[0287] In some embodiments, the at least one agent includes a first agent, a second agent, and a third agent;

[0288] A third agent is configured to interact with the target object based on the common key elements to obtain initial case information generated by the third agent; the initial case information includes a case introduction and case reference information retrieved by the third agent;

[0289] a first intelligent agent configured to interact with a target subject based on at least initial case information to generate first case statement information, wherein the first professional element satisfies a first preset condition, the first preset condition indicating that defending the target subject based on the first professional element is beneficial to safeguarding the target subject's legitimate rights and interests;

[0290] a second agent configured to interact with the target object based on at least the initial case information to generate second case statement information using a second professional element; wherein the second professional element satisfies a second preset condition, the second preset condition indicating that defending the target object's opponent based on the second professional element is beneficial to safeguarding the legitimate rights and interests of the target object's opponent;

[0291] The case description information includes first case statement information and second case statement information.

[0292] In some embodiments, the system further comprises a judge agent;

[0293] The first agent, the second agent, and the judge agent are used to simulate the trial process based on the case description information;

[0294] The judge agent generates virtual judgment documents in legal reference information based on the simulation results of similar cases and simulated trial sessions.

[0295] In some embodiments, the first agent is configured to generate a first professional element based on the first private domain legal knowledge and the public domain legal knowledge;

[0296] A second intelligent agent is used to generate a second professional element based on the second private domain legal knowledge and the public domain legal knowledge;

[0297] For any of the first agent and the second agent, any of the agents generates corresponding professional elements, including:

[0298] Map the information currently provided by the target object to the set of professional knowledge points corresponding to the private domain legal knowledge of any intelligent agent;

[0299] From the set of professional knowledge points, the knowledge points that match the position of any intelligent agent are screened out to obtain the professional elements corresponding to any intelligent agent; wherein, when any intelligent agent is the first intelligent agent, the position of the first intelligent agent is to safeguard the legitimate rights and interests of the target object; when any intelligent agent is the second intelligent agent, the position of the second intelligent agent is to safeguard the legitimate rights and interests of the opponent of the target object.

[0300] In some embodiments, any agent performs the following operations: filtering knowledge points that match the position of any agent from a set of professional knowledge points to obtain professional elements corresponding to any agent, including:

[0301] Filter the knowledge points that match the position of any agent from the set of professional knowledge points according to the preset tags;

[0302] Wherein, when any of the agents is the first agent, a preset mark is determined based on the role category of the target object, and the preset mark is used to mark that the selected knowledge points meet the first preset condition;

[0303] When any agent is the second agent, a preset mark is determined based on the role category of the opponent of the target object, and the preset mark is used to mark that the screened knowledge points meet the second preset condition.

[0304] In some embodiments, any agent performs the following operations: filtering knowledge points that match the position of any agent from a set of professional knowledge points to obtain professional elements corresponding to any agent, including:

[0305] Classify the knowledge points in the professional knowledge point set based on the reasoning summary model to obtain the classification results;

[0306] The knowledge points that match the position of any intelligent agent are filtered out from the classification results to obtain the professional elements corresponding to any intelligent agent.

[0307] It can be understood that the corresponding operations performed by the middle platform and intelligent agent in the above system have been explained in the previous article and will not be repeated here.

[0308] During implementation, the middle platform of the aforementioned system controls the entire process of interaction based on the state machine as follows Figure 13 As shown, when the middleware receives the first round of dialogue from the target object, it analyzes the first round of dialogue to determine the target legal field. If the target legal field is a field that cannot be handled collaboratively by multiple agents, the consultation process is terminated by interacting with the target object based on a specific copy. An example of the specific copy may be "This field is not currently involved. Please explain it again."

[0309] When the target legal field is a field that can be handled by multiple agents, the middle platform determines the role of each agent, the document format of the legal reference information in the field, and the case summary of historical cases in the field.

[0310] The cache can store the roles of each agent, the document format of legal reference information in the field, and historical case summaries in the field to facilitate the coordination of the various agents. The document format can be provided to the judge agent, and the historical case summaries can be used by the reasoning and summarization model to extract the required elements, such as common key elements, primary professional elements, and secondary professional elements.

[0311] Based on the current state of the state machine, the third agent is first controlled to interact with the target object, and in the follow-up phase, the target object is asked to supplement the common key elements. Each agent has a corresponding state monitoring strategy to monitor whether the current state needs to be modified in order to enter the next work node of the workflow. Figure 13 As shown in Figure 2, the working mechanism of each agent is as follows:

[0312] 1) For the third agent:

[0313] exist Figure 13 The role is a paralegal. The middle platform checks whether the status has changed based on the state machine to confirm whether to enter the next workflow. Figure 13 In the process, the middle platform first determines, based on the state machine, that the current task is for the third agent to interact with the target object to clarify the case. Based on the aforementioned solution, the third agent summarizes and organizes common key elements. Figure 13 The "questioning" in the questioning includes calling the questioning model to generate questioning questions for each common key element.

[0314] In the follow-up question phase, after the follow-up question is generated, Figure 13 The third agent in the game uses the "text processing" to call the follow-up option model to process the corresponding follow-up questions into options. After saving the historical dialogue, the follow-up options and follow-up questions generated by the follow-up option model are output to the target object. This operation is performed by Figure 13 It is realized by the "on screen" corresponding to the third intelligent agent.

[0315] Figure 13In the third agent's "modify status" during the follow-up phase, it can be understood as invoking a state modification strategy to determine whether the state of the third agent's follow-up phase needs to be modified. For example, if the follow-up phase has been completed for all common key elements, responses to the common key elements have been obtained, and a case description has been generated, the third agent's follow-up phase is complete. The middle platform detects this status through a check query and controls the third agent to enter the phase of obtaining case reference information.

[0316] The process of obtaining case reference information, such as Figure 13 The paralegal agent operates from the link of "supplementing pictures"... "going to screen".

[0317] The "Supplementary Image" function can be used to request additional evidence from the target; the "Thought Chain Rendering" function displays the current actions of the third agent, such as searching for relevant literature; the "Document Interface" function is used to call the corresponding interface for search queries; the "Image Rendering" function displays the search results; and the "Text Summary" function allows the third agent to organize the search results and generate case reference information. The "Modify Status" function then indicates that the third agent has completed the process of obtaining case reference information.

[0318] The "turn_break" corresponding to the third agent can indicate waiting for the target object to respond or confirm. For example, in the follow-up question phase, each follow-up question requires the target object to respond.

[0319] 2) For the first agent:

[0320] exist Figure 13 The role is exemplified as the opposing lawyer. The middle platform checks whether the status has changed based on the state machine to confirm whether to enter the next workflow. Figure 13 In this example, based on the state machine, the middle platform determines that the current task is for the first agent to interact with the target object to fully explore the details of the case. Based on the aforementioned solution, the middle platform controls the first agent to organize the first professional element. Figure 13 The "follow-up questions" corresponding to the first intelligent agent include calling the follow-up question model to generate follow-up questions for each first professional element.

[0321] In the first agent's follow-up questioning phase, after the follow-up question is generated, Figure 13 The first agent in the "text processing" corresponds to the first agent calling the question option model to process the corresponding question into options. After saving the historical dialogue, the question options and question questions generated by the question option model are output to the target object. This operation is performed by Figure 13 It is realized by the "on screen" corresponding to the first intelligent agent.

[0322] Figure 13 In the "Modify Status" of the first agent during the follow-up phase, it can be understood as invoking the corresponding state modification strategy to determine whether the state of the first agent's follow-up phase needs to be modified. For example, if the follow-up of the first professional element is completed and the first result of the first professional element has been obtained, it indicates that the first agent's follow-up phase has ended. The middle platform checks this status and controls the first agent to enter the phase of generating the case statement information corresponding to the first agent.

[0323] The step of generating the first case statement information corresponding to the first agent, such as Figure 13 The first agent performs the operation of the link from "summary (history + documents)"... "on screen".

[0324] Among them, "Summary (History + Documents)" can be used to summarize historical conversations and related documents, and "Text Summary" can be used by the first intelligent agent to summarize using the reasoning summary model to generate the first case statement information corresponding to the first intelligent agent, and use "Modify Status" to express that the link of the first intelligent agent generating the first case statement information has ended.

[0325] The "turn_break" corresponding to the first agent can indicate waiting for the target object to respond or confirm. For example, in the follow-up question phase, each follow-up question requires the target object to respond.

[0326] 3) For the second agent:

[0327] Continuing with the previous example, Figure 13 The role is exemplified as a lawyer for the party. The middle platform checks whether the status has changed based on the state machine to confirm whether to enter the next workflow. Figure 13 In this scenario, the middleware first determines, based on the state machine, that the second agent's current task is interacting with the target object. This second agent, by exploring the case from the perspective of the target object's opponent, can compensate for the first agent's shortcomings and further clarify the case. Based on the aforementioned solution, the second agent then summarizes and organizes the second professional elements. Figure 13 The "follow-up questions" corresponding to the second intelligent agent include the second intelligent agent calling the follow-up question model to generate follow-up questions for each second professional element.

[0328] In the follow-up question phase, after the follow-up question is generated, Figure 13 The second agent in the game uses the "text processing" to call the follow-up option model to process the corresponding follow-up questions into options. After saving the historical dialogue, the follow-up options and follow-up questions generated by the follow-up option model are output to the target object. This operation is performed by Figure 13 It is realized by the "on screen" corresponding to the second intelligent agent.

[0329] Figure 13 In the second agent's "modify status" during the follow-up phase, it can be understood as invoking a state modification strategy to determine whether the state of the second agent's follow-up phase needs to be modified. For example, if the second professional element has been questioned and the second result has been obtained, the second agent's follow-up phase is complete. The middle platform checks this status and controls the second agent to generate the second case statement.

[0330] The step of generating the second case statement information corresponding to the second agent, such as Figure 13 The second agent operates the link from "summary (history + documents)"... "on screen".

[0331] Among them, "Summary (History + Documents)" can be used to summarize historical conversations (supplementary information of the target object) and related documents, and "Text Summary" can be used to summarize using the reasoning summary model to generate the case statement information corresponding to the second intelligent agent. The "Modify Status" is used to indicate that the second intelligent agent has completed the process of generating the second case statement information.

[0332] The second agent's "turn_break" can indicate waiting for the target's response or confirmation. For example, in the follow-up question phase, each follow-up question requires a response from the target.

[0333] It should be noted that the first case statement information and the second case statement information in the embodiment of the present disclosure can be generated by the corresponding intelligent agent after the defense in the simulated trial is completed, and the judge intelligent agent makes a judgment.

[0334] 4) For the judge agent:

[0335] Continuing with the previous example, Figure 13 The role is exemplified as a judge. The case statement information of the first agent and the second agent is summarized as case description information. The state machine checks whether the state has changed to confirm whether to enter the next workflow, that is, the simulated trial phase. Figure 13 First, the middleware determines the current phase of the work as a mock trial based on the state machine. Specifically, based on the aforementioned solution, the judge agent summarizes and organizes the first case statement corresponding to the first agent and the second case statement corresponding to the second agent to conduct the mock trial. After the mock trial, a case summary is conducted. Figure 13The "case summary" in the example includes the judge agent calling the reasoning summary model to summarize the debate content corresponding to the first agent to obtain a case summary. The "judgment result" includes the judge agent using the RAG model to search the historical judgment information based on the case summary, and obtaining the judgment result based on the search results and the aforementioned case summary. Legal reference information is the legal reference information generated based on the judgment result and the case summary. Text processing can be used for the judge agent to select a suitable template, generate corresponding legal reference information, and associate corresponding legal provisions, etc. "Clear status" is used to clear the status of the target object's consultation in the cache to free up network resources. After saving the historical conversation, the legal reference information is displayed on the screen to the target object.

[0336] Based on the same technical concept, the embodiment of the present disclosure also proposes a legal consulting device 1400, such as Figure 14 As shown, including:

[0337] A first determination module 1401 is configured to determine a target legal field that the target subject needs to consult in response to any form of natural language input provided by the target subject;

[0338] A second determination module 1402 is configured to determine at least one agent for a target object based on the target legal field;

[0339] Interaction module 1403, configured to interact with a target object based on target knowledge associated with a target legal field by at least one agent to mine case details and obtain case description information;

[0340] The processing module 1404 is used to provide legal reference information to the target object based on the case description information.

[0341] In some embodiments, the legal reference information includes at least one of the following:

[0342] Virtual legal opinions and virtual judgment documents.

[0343] In some embodiments, the elements of the at least one agent's interaction with the target object include at least one of the following:

[0344] Generic key elements based on the generation of public domain legal knowledge in target knowledge;

[0345] A first professional element generated based on the first private domain legal knowledge in the target knowledge; the first professional element satisfies a first preset condition, and the first preset condition is used to indicate that defending the target object based on the first professional element is beneficial to safeguarding the legitimate rights and interests of the target object;

[0346] A second professional element is generated based on the second private domain legal knowledge in the target knowledge; the second professional element satisfies a second preset condition, and the second preset condition is used to indicate that when defending the opponent of the target object based on the second professional element, it can be beneficial to safeguard the legitimate rights and interests of the opponent of the target object.

[0347] In some embodiments, a construction module is further included for constructing public domain legal knowledge in the target knowledge based on the following method:

[0348] Generate a knowledge tree to be optimized based on the target legal provision in the target legal field and the authoritative interpretation information of the target legal provision; the knowledge tree to be optimized includes a root node created by the target legal provision and at least one first-level child node formed by the authoritative interpretation information of the target legal provision;

[0349] Supplementing the first child node at the last level in the knowledge tree to be optimized based on relevant cases of the target legal provision to generate at least one second child node at the last level corresponding to the first child node at the last level, so as to obtain public domain legal knowledge represented by the public domain knowledge tree;

[0350] The second child node at the last level in the public domain knowledge tree is used to represent the judgment result;

[0351] The path from the root node to the second child node at the last level includes the common key elements for obtaining the judgment result.

[0352] In some embodiments, the method for generating elements for interacting with a target object is based on the following method:

[0353] Inputting the first private domain legal knowledge and the public domain legal knowledge into the first intelligent agent to obtain a first professional element generated by the first intelligent agent;

[0354] Inputting the second private domain legal knowledge and the public domain legal knowledge into the second intelligent agent to obtain a second professional element generated by the second intelligent agent;

[0355] Input public domain legal knowledge into the third agent to generate common key elements;

[0356] Summarize the first professional elements, the second professional elements and the common key elements to obtain the elements that interact with the target object.

[0357] In some embodiments, the interaction module 1403 includes:

[0358] a first control unit configured to control the third agent to interact with the target object based on the common key elements to obtain initial case information generated by the third agent; the initial case information includes a case introduction and case reference information retrieved by the third agent;

[0359] The second control unit is configured to input the initial case information into the first agent to obtain first case statement information generated by the first agent after interacting with the target object on at least the basis of the initial case information and the first professional element;

[0360] a third control unit, configured to input the initial case information into the second agent, so as to obtain second case statement information generated by the second agent after interacting with the target object for a second professional element based on at least the initial case information;

[0361] The case description information includes first case statement information and second case statement information.

[0362] In some embodiments, the first private domain legal knowledge includes a first case input by a first management object corresponding to the first agent, and / or case analysis information of the first case by the first management object;

[0363] The second private domain legal knowledge includes the second precedent input by the second management object corresponding to the second intelligent agent, and / or the case analysis information of the second precedent by the second management object.

[0364] In some embodiments, the case description information includes information about disputed points that affect the judgment result. The disputed point information includes at least one of the following:

[0365] Elements of the Common Key Elements used for qualitative description;

[0366] Elements of the Common Key Elements that are marked as controversial based on historical cases;

[0367] Elements used for qualitative description in the first professional element;

[0368] The elements marked as controversial points in the first professional element;

[0369] The elements used for qualitative description in the second professional element;

[0370] Elements marked as controversial points in the second professional element;

[0371] Among them, the elements of qualitative description are conditions or standards that cannot be measured by numerical values;

[0372] Control the first agent and / or the second agent to extract elements constituting dispute points from the supplementary description of the target object.

[0373] In some embodiments, when the at least one agent includes a first agent, a second agent, and a judge agent, the processing module 1404 includes:

[0374] A simulation unit, configured to simulate a court trial based on the first agent, the second agent, and the judge agent;

[0375] The generation unit is used to generate legal reference information based on the simulation results of the simulated trial.

[0376] In some embodiments, the generating unit is specifically configured to:

[0377] Obtain similar cases retrieved by the judge agent;

[0378] Similar cases and simulation results are input into the judge agent to obtain the virtual judgment document in the legal reference information output by the judge agent.

[0379] In some embodiments, it further includes:

[0380] The analysis module is used to analyze legal reference information and case description information to obtain analysis results;

[0381] An updating module is used to control at least one intelligent agent to interact with the target object based on the missing elements in the case description information to update the legal reference information when the analysis result indicates that the case description information includes missing elements that do not interact with the target object.

[0382] In some embodiments, the second determining module 1402 includes:

[0383] A recommendation unit, configured to recommend a plurality of first candidate agents to the target object based on the target legal field of the target object and / or the location information of the target object; wherein each first candidate agent is associated with a professional characteristic description;

[0384] a determining unit, configured to determine an agent selected by the target object from the first candidate agents as the first agent;

[0385] The matching unit is used to automatically match the second agent, the third agent and the judge agent for the target object based on the first agent.

[0386] For the description of specific functions and examples of each module and submodule of the device in the embodiment of the present disclosure, please refer to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0387] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0388] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0389] Figure 15A schematic block diagram of an example electronic device 1500 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0390] like Figure 15 As shown, device 1500 includes a computing unit 1501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1502 or a computer program loaded from a storage unit 1508 into a random access memory (RAM) 1503. Various programs and data required for the operation of device 1500 can also be stored in RAM 1503. Computing unit 1501, ROM 1502, and RAM 1503 are connected to each other via a bus 1504. An input / output (I / O) interface 1505 is also connected to bus 1504.

[0391] Various components in device 1500 are connected to I / O interface 1505, including an input unit 1506, such as a keyboard and mouse; an output unit 1507, such as various types of displays and speakers; a storage unit 1508, such as a magnetic disk and optical disk; and a communication unit 1509, such as a network card, a modem, a wireless communication transceiver, etc. Communication unit 1509 allows device 1500 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0392] Computing unit 1501 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of computing unit 1501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Computing unit 1501 performs the various methods and processes described above, such as the legal reference information providing method. For example, in some embodiments, the legal reference information providing method is implemented as a computer software program tangibly embodied in a machine-readable medium, such as storage unit 1508. In some embodiments, part or all of the computer program can be loaded and / or installed onto device 1500 via ROM 1502 and / or communication unit 1509. When the computer program is loaded into RAM 1503 and executed by computing unit 1501, one or more steps of the legal reference information providing method described above can be performed. Alternatively, in other embodiments, the computing unit 1501 may be configured to execute the legal reference information providing method in any other appropriate manner (eg, by means of firmware).

[0393] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system comprising at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0394] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0395] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0396] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0397] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0398] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0399] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0400] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A method for providing legal reference information, comprising: In response to any form of natural language input provided by a target object, determining a target legal field that the target object needs to consult; determining at least one agent for the target object based on the target legal field; Based on the target knowledge associated with the target legal field, the at least one intelligent agent interacts with the target object to mine case details to obtain case description information; Based on the case description information, legal reference information is provided to the target object.

2. The method according to claim 1, wherein The legal reference information includes at least one of the following: Virtual legal opinions and virtual judgment documents.

3. The method according to claim 1, wherein The elements of the interaction between the at least one intelligent agent and the target object include at least one of the following: Generic key elements generated based on public domain legal knowledge in the target knowledge; A first professional element generated based on the first private domain legal knowledge in the target knowledge; The first professional element satisfies a first preset condition, where the first preset condition is used to indicate that defending the target object based on the first professional element is beneficial to safeguarding the legitimate rights and interests of the target object; A second professional element is generated based on the second private domain legal knowledge in the target knowledge; the second professional element satisfies a second preset condition, and the second preset condition is used to indicate that when defending the opponent of the target object based on the second professional element, it can be beneficial to safeguard the legitimate rights and interests of the opponent of the target object.

4. The method according to claim 3, wherein: Constructing public domain legal knowledge in the target knowledge includes: generating a knowledge tree to be optimized based on the target legal provision in the target legal field and the authoritative interpretation information of the target legal provision; wherein the knowledge tree to be optimized includes a root node created by the target legal provision and at least one first-level child node formed by the authoritative interpretation information of the target legal provision; Supplementing the first child node at the last level of the knowledge tree to be optimized based on relevant cases of the target legal provision to generate at least one second child node at the first level corresponding to the first child node at the last level, so as to obtain the public domain legal knowledge represented by the public domain knowledge tree; The second child node at the last level in the public domain knowledge tree is used to represent the judgment result; The path from the root node to the second child node at the last level includes common key elements for obtaining the judgment result.

5. The method according to claim 3, wherein: Generate elements for interacting with the target object, including: Inputting the first private domain legal knowledge and the public domain legal knowledge into a first intelligent agent to obtain the first professional element generated by the first intelligent agent; Inputting the second private domain legal knowledge and the public domain legal knowledge into a second intelligent agent to obtain the second professional element generated by the second intelligent agent; Inputting the public domain legal knowledge into a third intelligent agent to generate the universal key elements; The first professional elements, the second professional elements, and the universal key elements are aggregated to obtain elements for interacting with the target object.

6. The method according to claim 5, wherein: The at least one intelligent agent interacts with the target object based on the target knowledge associated with the target legal field to mine case details to obtain case description information, including: Controlling the third agent to interact with the target object based on the universal key elements to obtain initial case information generated by the third agent; the initial case information includes a case introduction and case reference information retrieved by the third agent; Inputting the initial case information into the first agent to obtain first case statement information generated by the first agent after interacting with the target object with the first professional element based on at least the initial case information; Inputting the initial case information into the second agent to obtain second case statement information generated by the second agent after interacting with the target object with the second professional element based on at least the initial case information; The case description information includes the first case statement information and the second case statement information.

7. The method according to claim 5 or 6, wherein: The first private domain legal knowledge includes a first case input by a first management object corresponding to the first agent, and / or case analysis information of the first case by the first management object; The second private domain legal knowledge includes the second precedent input by the second management object corresponding to the second intelligent agent, and / or the case analysis information of the second precedent by the second management object.

8. The method according to claim 3, wherein: The case description information includes information on disputed points that affect the judgment result, and the disputed points information includes at least one of the following: Elements used for qualitative description in the general key elements; Elements of the general key elements marked as controversial based on historical cases; The elements used for qualitative description in the first professional element; The elements marked as disputed points in the first professional element; The elements used for qualitative description in the second professional element; The elements marked as disputed points in the second professional element; The elements of the qualitative description are conditions or standards that cannot be measured numerically; Control the first agent and / or the second agent to extract elements constituting dispute points from the supplementary description of the target object.

9. The method according to claim 8, wherein In a case where the at least one intelligent agent includes a first intelligent agent, a second intelligent agent, and a judge intelligent agent, providing legal reference information to the target object based on the case description information includes: A simulated court trial based on the first agent, the second agent, and the judge agent; The legal reference information is generated based on the simulation results of the simulated trial.

10. The method according to claim 9, wherein: The generating of the legal reference information based on the simulation result of the simulated trial includes: Obtaining similar cases retrieved by the judge agent; The similar cases and the simulation results are input into the judge agent to obtain a virtual judgment document in the legal reference information output by the judge agent.

11. The method according to any one of claims 1 to 10, further comprising: Analyzing the legal reference information and the case description information to obtain an analysis result; In a case where the analysis result indicates that the case description information includes missing elements that do not interact with the target object, the at least one agent is controlled to interact with the target object based on the missing elements to update the legal reference information.

12. The method according to claim 5, wherein: The determining of at least one intelligent agent for the target object based on the target legal field includes: Recommending a plurality of first candidate agents to the target object based on the target legal field of the target object and / or the location information of the target object; wherein each of the first candidate agents is associated with a professional characteristic description; Determine an agent selected by the target object from the first candidate agents as the first agent; Based on the first agent, the second agent, the third agent and the judge agent are automatically matched for the target object.

13. A legal reference information providing system comprising: The middle platform is used to determine the target legal field that the target object needs to consult in response to any form of natural language input provided by the target object; determining at least one agent for the target object based on the target legal field; Based on the target knowledge associated with the target legal field, the at least one intelligent agent interacts with the target object to mine case details to obtain case description information; Based on the case description information, provide legal reference information to the target object; At least one intelligent agent is used to generate elements required to interact with the target object.

14. The system according to claim 13, wherein the at least one agent comprises a first agent, a second agent, and a third agent; The third agent is configured to interact with the target object based on the common key elements to obtain initial case information generated by the third agent; The initial case information includes a case introduction and case reference information retrieved by the third agent; The first agent is configured to interact with the target object at least based on the initial case information to generate first case statement information; The first professional element satisfies a first preset condition, where the first preset condition is used to indicate that defending the target object based on the first professional element is beneficial to safeguarding the legitimate rights and interests of the target object; The second agent is configured to interact with the target object at least based on the initial case information to generate second case statement information; The second professional element satisfies a second preset condition, where the second preset condition is used to indicate that defending the opponent of the target object based on the second professional element can be beneficial to safeguarding the legitimate rights and interests of the opponent of the target object; The case description information includes the first case statement information and the second case statement information.

15. The system according to claim 13, further comprising a judge agent; The first agent, the second agent, and the judge agent are used to simulate a court trial based on the case description information; The judge agent generates a virtual judgment document in the legal reference information based on similar cases and the simulation results of the simulated trial.

16. The system according to claim 13, wherein the first agent is configured to generate a first professional element based on first private domain legal knowledge and public domain legal knowledge; The second agent is configured to generate a second professional element based on the second private domain legal knowledge and the public domain legal knowledge; For any one of the first agent and the second agent, the agent generates corresponding professional elements, including: Mapping the information currently provided by the target object to a set of professional knowledge points corresponding to the private domain legal knowledge corresponding to any of the intelligent agents; From the set of professional knowledge points, knowledge points that match the position of any intelligent agent are screened out to obtain the professional elements corresponding to any intelligent agent; wherein, when any intelligent agent is the first intelligent agent, the position of the first intelligent agent is to safeguard the legitimate rights and interests of the target object; when any intelligent agent is the second intelligent agent, the position of the second intelligent agent is to safeguard the legitimate rights and interests of the opponent of the target object.

17. The system according to claim 16, wherein: The step of filtering, from the professional knowledge point set, the knowledge points that match the position of the any agent to obtain the professional elements corresponding to the any agent includes: Filtering, from the professional knowledge point set, knowledge points that match the position of any of the intelligent agents according to preset tags; Wherein, when any of the agents is the first agent, the preset mark is determined based on the role category of the target object, and the preset mark is used to mark that the selected knowledge point meets the first preset condition; When any of the intelligent agents is the second intelligent agent, the preset mark is determined based on the role category of the opponent of the target object, and the preset mark is used to mark that the screened knowledge point meets the second preset condition.

18. The system according to claim 16, wherein: The step of filtering, from the professional knowledge point set, the knowledge points that match the position of the any agent to obtain the professional elements corresponding to the any agent includes: Classifying the knowledge points in the professional knowledge point set based on the reasoning summary model to obtain a classification result; The knowledge points that match the standpoint of any of the intelligent agents are screened out from the classification results to obtain the professional elements corresponding to any of the intelligent agents.

19. A device for providing legal reference information, comprising: A first determination module is configured to determine a target legal field that the target object needs to consult in response to any form of natural language input provided by the target object; a second determination module, configured to determine at least one agent for the target object based on the target legal field; an interaction module, configured to interact with the target object based on the target knowledge associated with the target legal field by the at least one intelligent agent, so as to mine case details and obtain case description information; A processing module is used to provide legal reference information to the target object based on the case description information.

20. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 12.

21. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-12.

22. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 12.

Images