Personnel work assistance method and system based on ai assistant

Abstract

The present application provides a personnel work assistance method and system based on an AI assistant, wherein the method comprises: when a person is working, assisting the person in work based on an AI assistant; wherein the AI assistant is an artificial intelligence model trained by a large amount of work experience related to the work profile of the person; wherein the assisting the person in work based on the AI assistant comprises: acquiring real-time multi-modal work records of the person; generating first assistance content based on the AI assistant according to the multi-modal work records; pushing the first assistance content to the person; and / or receiving work assistance requirements input by the person; generating second assistance content based on the AI assistant according to the work assistance requirements; and pushing the second assistance content to the person. The present application realizes intelligent work flow support and personalized assistance functions, greatly improves the efficiency and accuracy of personnel work, and helps workers better cope with increasingly complex work requirements.

Description

A method and system for assisting personnel in their work based on an AI assistant Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for assisting personnel in their work based on an AI assistant. Background Technology

[0002] Currently, some staff members have heavy and complex workloads, covering multiple aspects such as case handling, intelligence analysis, patrol and inspection, and emergency response. Each task requires staff members to process large amounts of information in a short period of time and make accurate judgments and decisions, thus requiring work assistance.

[0003] Traditional work assistance tools mainly focus on information recording, storage, and retrieval functions. They are usually just simple task management systems, lacking intelligent workflow support and personalized assistance functions.

[0004] Therefore, there is an urgent need for a more intelligent and efficient solution that can improve the efficiency and accuracy of people's work and help them better cope with increasingly complex work requirements. Summary of the Invention

[0005] One objective of this invention is to provide a work assistance method based on an AI assistant. When using the AI ​​assistant to assist staff, it is divided into two modes: proactive assistance and passive assistance. In proactive assistance, the AI ​​assistant generates and pushes first-level assistance content based on the staff's multimodal work records. In passive assistance, the staff inputs their work assistance needs, and the AI ​​assistant generates and pushes corresponding second-level assistance content, achieving intelligent workflow support and personalized assistance functions. Specifically, it can provide staff with emergency response plans or optimal handling suggestions. Furthermore, it can help staff analyze cases, predict and warn of risks, greatly improving the efficiency and accuracy of staff work and helping them better cope with increasingly complex work requirements.

[0006] This invention provides a method for assisting personnel in their work based on an AI assistant, comprising:

[0007] When personnel are working, AI assistants provide assistance; the AI ​​assistant is an artificial intelligence model trained using a large amount of work experience related to the personnel's work profile.

[0008] The AI ​​assistant-based work assistance includes:

[0009] Real-time acquisition of multimodal work records of personnel;

[0010] Based on the AI ​​assistant and multimodal work records, the first auxiliary content is generated.

[0011] Push the first auxiliary content to the personnel;

[0012] And / or,

[0013] Receive work assistance requests input by personnel;

[0014] Based on the AI ​​assistant, generate secondary auxiliary content according to work assistance needs;

[0015] Push the second auxiliary content to the personnel.

[0016] Optionally, the generation of second auxiliary content based on the AI ​​assistant and according to work assistance needs includes:

[0017] The work support requirements are vectorized to obtain the requirement content vector;

[0018] Calculate the similarity between the demand content vector and the document vector in the vector database;

[0019] Secondary auxiliary content is generated based on the document content corresponding to the document vector with the highest similarity.

[0020] The steps for obtaining document vectors from the vector database are as follows:

[0021] The recipient shall upload documents in multiple formats; the formats of the documents shall include at least: TXT, DOCX, and PDF.

[0022] Perform vector conversion on the content of multi-format documents to obtain document vectors;

[0023] The step of converting the document content of multi-format documents into vectors to obtain document vectors includes:

[0024] The document content of multi-format documents is text-segmented and converted into document vectors using an embedding model.

[0025] Optionally, the embedding model includes at least the bge-large-zh-v1.5 Chinese embedding model.

[0026] Optionally, the vector database uses the PGVector vector library for vector storage.

[0027] Optionally, after generating the second auxiliary content based on the document content corresponding to the document vector with the highest similarity, the generated second auxiliary content is further optimized through the chatGLM API.

[0028] Optionally, when performing vector conversion on the document content of multi-format documents, key information in the document content is automatically identified and extracted based on the element direction specified by the user.

[0029] Optional, AI assistant-based methods for assisting people in their work also include:

[0030] When personnel use a work simulation sandbox, the system determines whether to enter a standard auxiliary timing based on the real-time simulated content of the work simulation performed by the personnel in the work simulation sandbox.

[0031] When entering, the auxiliary evidence is extracted from the real-time simulated content based on the standard auxiliary timing extraction template;

[0032] Based on the AI ​​assistant and the supporting data, the content of the simulation is determined.

[0033] An interactive control timeline for generating auxiliary simulation content;

[0034] Based on the auxiliary simulation content, corresponding work simulations are carried out in the work simulation sandbox, and during the simulation process, corresponding interactive control is carried out with personnel based on the interactive control timeline.

[0035] Optionally, the interactive control timeline for generating auxiliary simulation content includes:

[0036] Obtain the simulation progress timeline of the auxiliary simulation content;

[0037] Based on the time interval division constraint, multiple first target time intervals are divided on the simulated progress time axis;

[0038] Iterate through each of the first target time intervals in chronological order;

[0039] Each time the simulation is traversed, multiple immersive perspectives are determined based on the changes in the actions and positions of multiple simulated subjects within the first target time interval traversed on the simulation progress timeline.

[0040] Generate immersion trigger constraints for each immersive viewpoint;

[0041] Determine the second target time interval corresponding to the first target time interval traversed from the blank timeline;

[0042] Set each immersive perspective and its respective immersive triggering constraints within the second target time interval;

[0043] After traversing each first target time interval, the blank timeline with all immersive perspectives and their respective immersive trigger constraints set is used as the interactive control timeline.

[0044] Optionally, the time interval division constraint includes:

[0045] Each of the different first target time intervals is a continuous simulation time interval that completely contains at least N simulation scenarios without overlap; where N is a positive integer greater than or equal to 2;

[0046] as well as,

[0047] There is at least one attribute relationship between each pair of simulation scenarios in consecutive simulation time periods contained in the same first target time interval.

[0048] Optionally, the determination of multiple immersive viewpoints and their respective immersive trigger constraints based on the action and position changes of multiple simulated subjects within the first target time interval traversed on the simulation progress timeline includes:

[0049] Key actions are selected from the changes in the actions of each simulated subject to obtain the key changes in the actions of each simulated subject.

[0050] The key action changes and position changes of each simulated subject are time-aligned to obtain an alignment sequence;

[0051] Iterate through and align multiple action change nodes in the sequence in the order they appear.

[0052] Each time a node with a change in action is encountered, the simulated subject of that node is taken as the first target, and the remaining simulated subjects are taken as the second target.

[0053] From the traversed action change nodes, determine the first position change node and the second position change node of the first target and the second target, respectively, from the alignment position change nodes in the alignment sequence;

[0054] In the work simulation sandbox, when the first target is at the first position change node, is the simulation view of the second target at the second position change node visible to the traversed action change nodes? The simulation view includes: first-person view and third-person view.

[0055] When the target is visible, the simulated viewpoint of the second target when it is at the second position change node is used as the candidate immersive viewpoint; otherwise, the third-person viewpoint of the first target when it is at the first position change node is used as the candidate immersive viewpoint.

[0056] Associate the candidate immersive viewpoint with the time of change of the traversed action change nodes;

[0057] After traversing all action change nodes, integrate all candidate immersive viewpoints to obtain a viewpoint set;

[0058] The view set is optimized based on set optimization constraints;

[0059] Based on the optimized view set, multiple immersive viewpoints are determined.

[0060] Optionally, the set optimization constraints include:

[0061] The optimized perspective sets different time intervals associated with the different candidate immersive perspectives;

[0062] as well as,

[0063] The optimized viewpoint set has the fewest categories of simulated subjects belonging to different candidate immersive viewpoints.

[0064] Optionally, the generation of immersion trigger constraints for each immersive viewpoint includes:

[0065] Personnel adjusting their viewing angle of the work simulation sand table, and the time difference between the adjustment and the time of change associated with the immersive view does not exceed the time difference threshold;

[0066] And / or,

[0067] The overlap of visible content between the viewing angle after personnel adjustment and the immersive view exceeds the overlap threshold.

[0068] This invention provides an AI-assisted work support system, comprising:

[0069] The first auxiliary module is used to assist personnel in their work based on an AI assistant; the AI ​​assistant is an artificial intelligence model trained using a large amount of work experience related to the personnel's work profile.

[0070] The first auxiliary module, based on an AI assistant, provides work assistance to personnel, including:

[0071] Real-time acquisition of multimodal work records of personnel;

[0072] Based on the AI ​​assistant and multimodal work records, the first auxiliary content is generated.

[0073] Push the first auxiliary content to the personnel;

[0074] And / or,

[0075] Receive work assistance requests input by personnel;

[0076] Based on the AI ​​assistant, generate secondary auxiliary content according to work assistance needs;

[0077] Push the second auxiliary content to the personnel.

[0078] Optional, AI assistant-based human work assistance systems also include:

[0079] The second auxiliary module is used for:

[0080] When personnel use a work simulation sandbox, the system determines whether to enter a standard auxiliary timing based on the real-time simulated content of the work simulation performed by the personnel in the work simulation sandbox.

[0081] When entering, the auxiliary evidence is extracted from the real-time simulated content based on the standard auxiliary timing extraction template;

[0082] Based on the AI ​​assistant and the supporting data, the content of the simulation is determined.

[0083] An interactive control timeline for generating auxiliary simulation content;

[0084] Based on the auxiliary simulation content, corresponding work simulations are carried out in the work simulation sandbox, and during the simulation process, corresponding interactive control is carried out with personnel based on the interactive control timeline.

[0085] Optionally, the second auxiliary module generates an interactive control timeline for the auxiliary simulation content, including:

[0086] Obtain the simulation progress timeline of the auxiliary simulation content;

[0087] Based on the time interval division constraint, multiple first target time intervals are divided on the simulated progress time axis;

[0088] Iterate through each of the first target time intervals in chronological order;

[0089] Each time the simulation is traversed, multiple immersive perspectives are determined based on the changes in the actions and positions of multiple simulated subjects within the first target time interval traversed on the simulation progress timeline.

[0090] Generate immersion trigger constraints for each immersive viewpoint;

[0091] Determine the second target time interval corresponding to the first target time interval traversed from the blank timeline;

[0092] Set each immersive perspective and its respective immersive triggering constraints within the second target time interval;

[0093] After traversing each first target time interval, the blank timeline with all immersive perspectives and their respective immersive trigger constraints set is used as the interactive control timeline.

[0094] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0095] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0096] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0097] Figure 1 is a schematic diagram of a personnel work assistance method based on an AI assistant in an embodiment of the present invention;

[0098] Figure 2 is a schematic diagram of a personnel work assistance system based on an AI assistant in an embodiment of the present invention. Detailed Implementation

[0099] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0100] This invention provides a method for assisting personnel in their work based on an AI assistant, as shown in Figure 1, including:

[0101] S1. When personnel are working, AI assistants provide work assistance; the AI ​​assistant is an artificial intelligence model trained using a large amount of work experience related to personnel's work profiles.

[0102] In S1, personnel can be staff members; the personnel's work profile includes: position, work experience, gender, etc.; work experience includes historical case handling processes, typical case handling procedures, historical case records, etc., related to the personnel's work profile; a large amount of this work experience is used as training samples to train the artificial intelligence model, so that it can combine the work experience used for training to provide adaptive assistance to personnel.

[0103] S1, which provides work assistance to personnel based on an AI assistant, includes:

[0104] S111. Real-time acquisition of multimodal work records of personnel;

[0105] In S111, the AI ​​assistant acquires multimodal work records of personnel in real time through work recorders or sensors worn by personnel. The recorded content includes various forms of data such as text input, voice communication, and image upload. For example, when staff are performing patrol tasks, the AI ​​assistant can acquire voice communication records and on-site images in real time through voice recognition and image recognition technologies.

[0106] S112. Based on the AI ​​assistant, generate the first auxiliary content according to the multimodal work record;

[0107] In S112, the AI ​​assistant analyzes the work situation based on multimodal work records and combines the work experience used for training to generate targeted auxiliary content. For example, if a person is handling a case and collecting case information, the AI ​​assistant can automatically provide suggestions and warnings related to the case based on existing case data and historical experience.

[0108] S113. Push the first auxiliary content to the personnel;

[0109] In S113, the AI ​​assistant will push the generated primary assistance content to the personnel in real time for their reference in decision-making. For example, in an emergency, the AI ​​assistant can push emergency response plans and early warning information of predicted risks to help personnel respond quickly.

[0110] And / or,

[0111] S121. Receiving work assistance requests input by personnel;

[0112] In S121, when a person encounters difficulties at work, they can send specific work assistance requests to the AI ​​assistant. For example, when a person is handling a complex task, they may request the AI ​​assistant to provide information support or operational suggestions in a certain related field.

[0113] S122. Based on the AI ​​assistant, generate second auxiliary content according to work assistance needs;

[0114] In S122, the AI ​​assistant generates secondary auxiliary content based on work assistance needs and training experience. For example, when a staff member asks the AI ​​assistant how to analyze the communication records of relevant personnel while handling a case, the AI ​​assistant generates targeted analysis suggestions based on historical case experience.

[0115] S123, Push the second auxiliary content to the personnel.

[0116] In S123, the AI ​​assistant will finally push the generated secondary assistance content to the person to help them complete their work more efficiently.

[0117] This application utilizes an AI assistant to assist staff in two ways: proactive and passive assistance. In proactive assistance, the AI ​​assistant generates and pushes primary assistance content based on the staff's multimodal work records. In passive assistance, the staff inputs their assistance needs, and the AI ​​assistant generates and pushes corresponding secondary assistance content, achieving intelligent workflow support and personalized assistance functions. Specifically, it can provide staff with emergency response plans or optimal handling suggestions. Furthermore, it can help staff analyze cases, predict and warn of risks, greatly improving the efficiency and accuracy of staff work and helping them better cope with increasingly complex work requirements.

[0118] In one embodiment, the AI ​​assistant-based method for assisting personnel in their work further includes:

[0119] S2. When personnel use the work simulation sandbox, determine whether to enter the standard auxiliary timing based on the real-time simulated content of the personnel's work simulation in the work simulation sandbox.

[0120] In S2, the work simulation sandbox is a virtual environment used to simulate a certain work process or task, specifically implemented in the form of a computer-simulated 3D environment. The real-time simulated content refers to the content that the personnel have completed in the work simulation sandbox in real time, such as: setting up simulated characters, simulated plots, simulated locations, etc. The standard assistance timing refers to the standard time when personnel need to provide work simulation assistance. For example, if the simulated content indicates that the personnel have completed the setting of all simulated characters, simulated plots, simulated locations, etc., then they need to start preparing for the simulation and thus need to provide work simulation assistance. Another example is if the simulated content indicates that the personnel have paused the simulation for a long time and may not know how to continue the simulation, then they need to provide work simulation assistance. Therefore, the decision to enter the standard assistance timing can be determined based on the real-time simulated content.

[0121] S3. When entering, extract auxiliary evidence from the real-time simulated content based on the standard auxiliary timing auxiliary evidence extraction template;

[0122] In S3, the standard timing for entering the simulation also reflects how to provide work simulation assistance to personnel. The template for extracting the auxiliary basis for the standard timing is to extract the auxiliary simulation content that can determine the auxiliary simulation content for personnel to provide work simulation assistance from the real-time simulated content. For example, if the standard timing for entering the simulation is when personnel have completed the setting of all simulated characters, simulated plots, simulated locations, etc., and begin to prepare for the simulation, then the template for extracting the auxiliary basis is to extract the setting parameters of all simulated characters, simulated plots, simulated locations, etc. from the real-time simulated content as the auxiliary basis.

[0123] S4. Based on the AI ​​assistant and the auxiliary data, determine the content of the auxiliary simulation;

[0124] In S4, the AI ​​assistant can determine the content of the auxiliary simulation based on the auxiliary data and the work experience used for training (focusing on the development process of historical cases). The content of the auxiliary simulation is the plot development that can be simulated in the work simulation sandbox.

[0125] S5. Generate an interactive control timeline for auxiliary simulation content;

[0126] In S5, the interactive control timeline is used to assist the simulation content in instructing the system and personnel to perform corresponding interactive controls during the work simulation sandbox process.

[0127] S6. Based on the auxiliary simulation content, conduct corresponding work simulations in the work simulation sandbox, and during the simulation process, conduct corresponding interactive control with personnel based on the interactive control timeline.

[0128] The embodiments of the present invention have achieved the following beneficial effects:

[0129] By analyzing real-time simulated content within the work simulation sandbox, the system can accurately identify standard assistance opportunities, ensuring that personnel receive timely and necessary support during work simulations. Once a standard assistance opportunity is entered, the system provides personalized assistance guidelines and intelligently generates supplementary simulation content through an AI assistant that combines historical cases and work experience. This not only enhances the realism and practicality of the work simulation but also effectively guides personnel to maintain efficient decision-making and responsiveness in complex work simulation scenarios. During the work simulation sandbox process, the interactive control timeline serves as an interactive control guide, making the interaction between the system and personnel smoother and more efficient. This not only improves the efficiency of personnel's work simulations but also enhances the user experience.

[0130] In one embodiment, S5, the interactive control timeline for generating auxiliary simulation content, includes:

[0131] S51. Obtain the simulation progress timeline of the auxiliary simulation content;

[0132] In S51, the simulation progress timeline is the timeline for the real-time control auxiliary simulation content to simulate progress within the work simulation sandbox.

[0133] S52. Based on the time interval division constraint, multiple first target time intervals are divided on the simulated progress time axis;

[0134] S53. Traverse each first target time interval in chronological order;

[0135] In S53, timing refers to the chronological order of the first target time interval on the simulation progress time axis;

[0136] S54. Each time the simulation progress time interval is reached, multiple immersive viewpoints are determined based on the action changes and position changes of multiple simulated subjects within the first target time interval traversed on the simulation progress time axis.

[0137] In S54, the first target time interval contains local simulation content that is simulated within the first target time interval. The local simulation content includes the action changes and position changes of multiple simulated subjects. The simulated subjects can be simulated characters set by personnel. The action changes refer to the sequence of simulated actions performed by the simulated subjects, and the position changes refer to the sequence of positional movements performed by the simulated subjects. When using the work simulation sandbox for work simulation, appropriate future work plans are mainly formulated by observing the simulation situation of the simulated subjects. Based on the action changes and position changes of these multiple simulated subjects, multiple immersive perspectives can be determined. Each immersive perspective can help personnel immerse themselves in observing the simulation situation of the simulated subjects within the first target time interval they have traversed.

[0138] S55. Generate immersion trigger constraints for each immersive viewpoint;

[0139] In S55, the immersion trigger constraint is a constraint condition that requires the use of an immersive perspective to guide people to immerse themselves in the viewing experience.

[0140] S56. Determine the second target time interval corresponding to the first target time interval traversed from the blank timeline;

[0141] In S56, the blank timeline is a timeline without any content set. The second target time interval corresponding to it has the same start time as the first target time interval being traversed. They are distinguished only because they are on different timelines.

[0142] S57. Set each immersive viewpoint and its respective immersive trigger constraint within the second target time interval;

[0143] In S57, when each immersive perspective and its respective immersive trigger constraint are set within the second target time interval, when interacting with personnel based on the interactive control timeline, if the time enters the second target time interval, and if the user's viewing of the work simulation sand table meets the immersive trigger constraint, then the immersive perspective will guide the personnel to immerse themselves in viewing the work simulation sand table.

[0144] S58. After traversing each first target time interval, use the blank timeline with all immersive views and their respective immersive trigger constraints set as the interactive control timeline.

[0145] In S58, after all traversals are completed, that is, after all immersive views and their respective immersive trigger constraints are set, the blank timeline is used as the interactive control timeline.

[0146] The time interval division constraints include:

[0147] Constraint A1: Each of the different first target time intervals must be a continuous simulation time interval that does not overlap with each other and completely contains at least N simulation scenarios; where N is a positive integer greater than or equal to 2.

[0148] as well as,

[0149] Constraint A2: There must be at least one attribute relationship between any two consecutive simulation time periods contained in the same first target time interval.

[0150] In constraint A1, when using auxiliary simulation content to simulate different simulation scenarios in different simulation time periods, setting constraint A1 can ensure that interactive control occurs within a complete and continuous simulation time period of at least N simulation scenarios, avoiding inappropriate interactive control timing, and also ensuring that the content of interactive control is different at different times, avoiding repetition of interactive control.

[0151] In constraint A2, the attribute relationships include: the simulated scenarios occur in the same location, there is a causal relationship between the simulated scenarios, and the simulated scenarios involve the same simulated characters. Setting constraint A2 can make the interactive control more appropriate for simulated scenarios with relationships.

[0152] The embodiments of the present invention have achieved the following beneficial effects:

[0153] By organically combining simulation progress with an interactive timeline, and through precise division of simulation time periods and intelligent allocation of immersive perspectives, users can observe and analyze various simulation scenarios in a time-series-oriented and targeted manner. Specifically, by dividing the simulation progress timeline into multiple first-target time intervals, the changes in the actions and positions of the simulated subject are analyzed within each first-target time interval, thereby determining multiple immersive perspectives. This allows users to not only track the evolution of the simulated subject in real time, but also flexibly enter immersive perspectives to obtain specific situational information based on immersive trigger constraints. Secondly, through the guidance of the interactive control timeline, it enables assistants to deeply participate in the simulation process, improving the accuracy of personnel conducting work simulations.

[0154] Furthermore, by using constraint A1, the system can avoid repetition and breaks in interactive control, ensuring that each simulation period provides users with a sufficient and effective interactive experience. By using constraint A2, the internal logical rigor of the simulation scenario is ensured, enabling users to not only focus on the development of individual scenarios when making decisions and analyses, but also to grasp the relationships between various scenarios from an overall perspective. The combination of constraints A1 and A2 ensures that the simulation environment is not only continuous in the time dimension, but also enriched in the content dimension, thereby further improving the suitability of interactive control and the user's immersion.

[0155] In one embodiment, S54, based on the motion and position changes of multiple simulated subjects within the first target time interval traversed on the simulation progress timeline, determines multiple immersive viewpoints and their respective immersive trigger constraints, including:

[0156] S5401. Select key actions for the changes in the actions of each simulated subject to obtain the key changes in the actions of each simulated subject.

[0157] In S5401, the key action screening process involves selecting changes in key actions that are valuable to the personnel's work simulation from the changes in actions. These changes are called key action changes. The value can be that viewing the key actions can enable the personnel to further develop future work plans, or that viewing the key actions can enable the personnel to quickly understand the key points of the simulation situation. For example, the key action is the simulation of a suspect escaping.

[0158] S5402. Time-align the key motion changes and position changes of each simulated subject to obtain an alignment sequence;

[0159] In S5402, when performing time alignment, the key action changes are first converted into action change sequences and the position changes are converted into position change sequences. Then, based on the change time of each sequence item in the two change sequences, the sequence items with the same change time in the two change sequences are aligned to obtain an aligned sequence. In the aligned sequence, there are action change nodes. An action change node is a change action, and the position change node aligned with it is the change position that occurs at the same time.

[0160] S5403. Traverse multiple action change nodes in the alignment sequence in sequence order;

[0161] In S5403, the sequence order refers to the order in which the action change nodes are positioned in the alignment sequence;

[0162] S5404. Each time the target is reached, the simulated subject of the action change node is taken as the first target, and the remaining simulated subjects are taken as the second target.

[0163] In S5404, the simulation subject of the motion change node is the simulation subject that generates the motion change node, and the other simulation subjects refer to the simulation subjects other than the first target.

[0164] S5405. Determine the first position change node and the second position change node of the first target and the second target respectively from the alignment position change nodes in the alignment sequence of the traversed action change nodes.

[0165] In S5405, the first position change node is generated by the first target, and the second position change node is generated by the second target;

[0166] S5406. When the first target is determined to be at the first position change node in the work simulation sandbox, is the simulation view of the second target at the second position change node visible to the traversed action change nodes? The simulation view includes: first-person view and third-person view.

[0167] In S5406, when determining whether something is visible, the judgment is based on the actual simulation environment in the working simulation sandbox when the first target is at the first position change node and the second target is at the second position change node.

[0168] S5407. When the target is visible, the simulated view of the second target when it is at the second position change node shall be used as the candidate immersive view; otherwise, the third-person view of the first target when it is at the first position change node shall be used as the candidate immersive view.

[0169] In S5406, when it is visible, it means that the person can see the traversed action change nodes through the simulated view when the second target is at the second position change node, and this view is taken as the candidate immersion view; otherwise, the third-person view when the first target is at the first position change node is taken as the candidate immersion view. The action change nodes can be clearly seen through the simulated view of others, and the surrounding environment of the first target can also be fully viewed, which is more suitable. Therefore, the visible simulated view is preferred as the candidate immersion view; if it is not visible, the first target's own third-person view can more fully view the action change nodes it has generated, so its own third-person view is taken as the candidate immersion view.

[0170] S5408. Associate the candidate immersive viewpoint with the change time of the traversed action change nodes;

[0171] In S5408, the action change node has a change time, which refers to the time when the change node of General Manager Dong occurs, and the candidate immersive viewpoint is associated with the change time;

[0172] S5409. After traversing all action change nodes, integrate all candidate immersive perspectives to obtain a perspective set.

[0173] In S5409, the view set includes all available immersive viewpoints;

[0174] S5410. Optimize the view set based on set optimization constraints;

[0175] S5411. Based on the optimized view set, determine multiple immersive viewpoints;

[0176] In S5411, all candidate immersive views in the optimized view set are used as immersive views;

[0177] The set optimization constraints include:

[0178] Constraint B1: The time changes associated with different candidate immersive viewpoints in the optimized viewpoint set are not the same.

[0179] as well as,

[0180] Constraint B2: The optimized viewpoint set has the fewest categories of simulated subjects belonging to different candidate immersive viewpoints.

[0181] In constraint B1, the change time associated with different candidate immersive perspectives is ensured to be different, so that the final determined immersive perspective repeatedly leads the personnel to immerse themselves in the work simulation sand table at the same time.

[0182] In constraint B2, the simulated subjects belonging to the candidate immersive viewpoints associated with the same change time in the viewpoint set may be different. When using different candidate immersive viewpoints in the viewpoint set to guide personnel to immerse themselves in the work simulation sand table, there will be multiple candidate immersive viewpoints belonging to the same simulated subject to choose from at the same time. Therefore, in the optimization process, ensuring that there is only one associated candidate immersive viewpoint for the same change time will result in multiple candidate viewpoint sets. Selecting the viewpoint set with the fewest categories of simulated subjects belonging to different candidate immersive viewpoints as the optimized viewpoint set can ensure that the number of viewpoint switching times is minimized when using different immersive viewpoints to guide personnel to immerse themselves in the work simulation sand table.

[0183] The embodiments of the present invention have achieved the following beneficial effects:

[0184] Key action screening ensures the value of key action changes, enabling personnel to obtain information closely related to work decisions through the key actions of the simulated subject; time alignment ensures that the action changes and position changes of different simulated subjects can be synchronized on the same timeline, thereby improving the realism and coherence of the simulation process; by optimizing the set of perspectives composed of selected immersive perspectives, it is possible to provide a diverse and efficient perspective experience while ensuring the fewest perspective choices and the smoothest switching, thereby enhancing the simulated personnel's overall understanding of the work situation and reducing visual fatigue and information overload during the simulation process.

[0185] By using constraint B1, it is ensured that different candidate immersive perspectives do not overlap in time, thus avoiding information duplication and visual redundancy, allowing personnel to experience complete and diverse work scenarios at different points in time. By using constraint B2, the category of the simulated subject to each immersive perspective at each point in time is minimized, which not only reduces the frequency of perspective switching but also improves the smoothness and immersion of the simulation. The combination of constraints B1 and B2 allows for rich information feedback when using the optimized perspectives to guide personnel through the work simulation sandbox, and can also effectively guide personnel to deeply analyze and understand the complex situations in the work simulation sandbox, thereby providing effective support for the decision-making of corresponding future work plans.

[0186] In one embodiment, S55, generating immersion trigger constraints for each immersive viewpoint, includes:

[0187] Constraint C1: When personnel adjust their viewing angle of the work simulation sand table, the time difference between the adjustment time and the time of change associated with the immersive view does not exceed the time difference threshold.

[0188] And / or,

[0189] Constraint C2: The overlap of visible content between the viewing angle after personnel adjustment and the immersive view exceeds the overlap threshold.

[0190] In constraint C1, when a person views the work simulation sand table, a viewing perspective is generated; the adjustment time refers to the time when the person adjusts their viewing perspective of the work simulation sand table; the time difference threshold can be 20 seconds; when the time difference between the whole time and the time of change associated with the immersive perspective does not exceed the time difference threshold, it means that the person adjusted the viewing perspective shortly before the triggerable time of the immersive perspective (associated change time), possibly wanting to view the work simulation sand table more deeply, and thus the immersive perspective can be triggered;

[0191] In constraint C2, the overlap threshold can be 65%. When the overlap of visible content between the user's adjusted viewing angle and the immersive view exceeds the overlap threshold, it indicates that the user intends to immerse themselves in the viewing experience by adjusting their viewing angle, and thus the immersive view can be triggered.

[0192] The embodiments of the present invention have achieved the following beneficial effects:

[0193] By constraining the coordination between C1 and C2, the timing for guiding personnel to view the work simulation sand table through an immersive perspective can be precisely determined, improving the user experience and making it more intelligent.

[0194] In one embodiment, the generation of second auxiliary content based on the AI ​​assistant and according to work assistance needs includes:

[0195] The work support requirements are vectorized to obtain the requirement content vector;

[0196] Calculate the similarity between the demand content vector and the document vector in the vector database;

[0197] Secondary auxiliary content is generated based on the document content corresponding to the document vector with the highest similarity.

[0198] The steps for obtaining document vectors from the vector database are as follows:

[0199] The recipient shall upload documents in multiple formats; the formats of the documents shall include at least: TXT, DOCX, and PDF.

[0200] Perform vector conversion on the content of multi-format documents to obtain document vectors;

[0201] The step of converting the document content of multi-format documents into vectors to obtain document vectors includes:

[0202] The document content of multi-format documents is text-segmented and converted into document vectors using an embedding model.

[0203] The embedding model includes at least the bge-large-zh-v1.5 Chinese embedding model.

[0204] The vector database uses the PGVector vector library for vector storage.

[0205] After generating the second auxiliary content based on the document content corresponding to the document vector with the highest similarity, the generated second auxiliary content is further optimized through the chatGLM API.

[0206] When performing vector conversion on the content of multi-format documents, the system automatically identifies and extracts key information from the document content based on the element orientation specified by the user.

[0207] The shortcomings of the prior art that this invention aims to address are as follows:

[0208] High reliance on manual operation: In the work process, from data collection and entry to archiving, a large amount of work relies on manual operation, which is not only inefficient, but also prone to errors due to human factors.

[0209] Low information processing efficiency: When handling cases, staff need to read and analyze a large number of documents and manually extract key information, which is time-consuming and affects the speed of case handling.

[0210] Information extraction accuracy is limited: When manually analyzing documents, important information is easily missed, or information extraction is inaccurate due to subjective judgment.

[0211] Cross-document information association is difficult: When handling complex business transactions, information is often linked between multiple documents. Manually sorting out these connections is quite challenging and may result in insufficient discovery of case leads.

[0212] The present invention aims to significantly improve the efficiency and quality of work by leveraging artificial intelligence technology. This objective is primarily achieved through the following advanced functions:

[0213] Through an automated document processing workflow, this invention can quickly and accurately extract key information, discover potential correlations and patterns, and help staff to explore clues in depth.

[0214] Employing cutting-edge natural language processing technology, it helps staff quickly grasp the essence of documents and effectively absorb relevant knowledge, thereby improving work efficiency;

[0215] Relying on advanced information integration technology, this invention can identify and merge relevant documents, sort out key elements in the documents, and construct complex relationship maps between elements; through precise algorithms and pattern recognition, this tool can reduce human error, ensure the accuracy of extracted information, and improve the reliability of case information.

[0216] The ultimate goal of these functions is to leverage artificial intelligence technology to reduce the workload of staff, comprehensively improve execution efficiency, and promote technological innovation to meet the challenges of the new era and provide solid technological support for the stability of social order and public safety.

[0217] The specific implementation aspects of this invention are as follows:

[0218] I. Document Analysis

[0219] Document upload and text parsing

[0220] Instructions: Upload TXT, DOCX, or PDF documents (PDF only supports text format). Each file must be no larger than 10MB and a maximum of 10 files can be uploaded.

[0221] Objective: To build a vector library so that corresponding files can be parsed.

[0222] Technology: After reading the document content and performing text segmentation, the text is vectorized using the same embedded model. (Text segmentation is a tool that breaks down continuous text data into smaller, more meaningful parts. In the field of Natural Language Processing (NLP), text segmentation is an important task that helps us better understand the structure and content of text, thereby enabling further text analysis and processing. The same embedded model refers to the bge-large-zh-v1.5 model.)

[0223] Objective: To convert documents of various formats into vector form to facilitate subsequent data processing and retrieval.

[0224] Data storage:

[0225] Operation: Save the vectorized text data in the PGVector vector library.

[0226] Objective: To ensure the consistency and searchability of knowledge base data.

[0227] II. User Questions and Answers

[0228] User input processing:

[0229] Operation: Vectorizing the user's input problem.

[0230] Objective: To transform user questions into vector form for data matching.

[0231] Determine the matching range:

[0232] Operation: The user selects the files they want to parse.

[0233] Objective: To determine the parsing range by matching the file names selected by the user with documents in the vector library.

[0234] Vector library matching:

[0235] Operation: Perform similarity matching between the user's question and the contents of the files in the selected vector library.

[0236] Logic: If a matching content is found, the corresponding content and the user's question are combined to generate a prompt, the chatGLM API is called, and the prompt returns the answer based on the matched vector library content.

[0237] Objective: To obtain accurate answers that are relevant to the parsed document.

[0238] Answer optimization and presentation:

[0239] This technical solution applies AI and NLP technologies to build an efficient information retrieval system that can answer user questions based on specified documents.

[0240] III. Element Extraction

[0241] Operation: Users can specify the direction of the elements to be extracted or let the AI ​​extract elements from the document autonomously.

[0242] Logic: Combine the document content selected by the user to construct a prompt, and call the chatGLM API to extract the corresponding elements from the document.

[0243] Objective: To help users quickly organize document information and extract key information points.

[0244] The embodiments of the present invention have achieved the following beneficial effects:

[0245] 1. Innovation in unified parsing and vectorization of multi-format documents

[0246] This innovative technical solution achieves unified parsing and vectorization of documents in various formats such as TXT, DOCX, and PDF. By using the bge-large-zh-v1.5 model, the content of documents in different formats is converted into vector form, facilitating storage and retrieval.

[0247] 2. Intelligent Question Answering Based on Vector Library

[0248] This invention allows users to upload documents and build a vector library. Then, by vectorizing the user's input questions, it achieves similarity matching with the content of documents in the vector library, thereby allowing users to find accurate answers in selected documents.

[0249] 3. Flexible element extraction function

[0250] This invention allows users to specify the direction of element extraction and uses chatGLM to quickly organize document information and extract key information points.

[0251] 4. High efficiency in data storage and retrieval

[0252] The technical solution uses the PGVector vector library to store vectorized text data, ensuring the consistency and efficient retrieval of knowledge base data. This data storage method not only optimizes the storage structure but also improves the speed of data retrieval.

[0253] This invention provides an AI assistant-based personnel work assistance system, as shown in Figure 2, including:

[0254] The first auxiliary module 1 is used to assist personnel in their work based on an AI assistant; wherein, the AI ​​assistant is an artificial intelligence model trained using a large amount of work experience related to the personnel's work profile.

[0255] The first auxiliary module, based on an AI assistant, provides work assistance to personnel, including:

[0256] Real-time acquisition of multimodal work records of personnel;

[0257] Based on the AI ​​assistant and multimodal work records, the first auxiliary content is generated.

[0258] Push the first auxiliary content to the personnel;

[0259] And / or,

[0260] Receive work assistance requests input by personnel;

[0261] Based on the AI ​​assistant, generate secondary auxiliary content according to work assistance needs;

[0262] Push the second auxiliary content to the personnel.

[0263] AI-assisted work support systems also include:

[0264] The second auxiliary module is used for:

[0265] When personnel use a work simulation sandbox, the system determines whether to enter a standard auxiliary timing based on the real-time simulated content of the work simulation performed by the personnel in the work simulation sandbox.

[0266] When entering, the auxiliary evidence is extracted from the real-time simulated content based on the standard auxiliary timing extraction template;

[0267] Based on the AI ​​assistant and the supporting data, the content of the simulation is determined.

[0268] An interactive control timeline for generating auxiliary simulation content;

[0269] Based on the auxiliary simulation content, corresponding work simulations are carried out in the work simulation sandbox, and during the simulation process, corresponding interactive control is carried out with personnel based on the interactive control timeline.

[0270] The second auxiliary module generates an interactive control timeline for the auxiliary simulation content, including:

[0271] Obtain the simulation progress timeline of the auxiliary simulation content;

[0272] Based on the time interval division constraint, multiple first target time intervals are divided on the simulated progress time axis;

[0273] Iterate through each of the first target time intervals in chronological order;

[0274] Each time the simulation is traversed, multiple immersive perspectives are determined based on the changes in the actions and positions of multiple simulated subjects within the first target time interval traversed on the simulation progress timeline.

[0275] Generate immersion trigger constraints for each immersive viewpoint;

[0276] Determine the second target time interval corresponding to the first target time interval traversed from the blank timeline;

[0277] Set each immersive perspective and its respective immersive triggering constraints within the second target time interval;

[0278] After traversing each first target time interval, the blank timeline with all immersive perspectives and their respective immersive trigger constraints set is used as the interactive control timeline.

[0279] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for assisting personnel in their work based on an AI assistant, characterized in that, include: When personnel are working, work assistance is provided based on an AI assistant; wherein the AI ​​assistant is an artificial intelligence model trained using a large amount of work experience related to the personnel's work profile; wherein, the work assistance based on the AI ​​assistant includes: real-time acquisition of the personnel's multimodal work records; generating first assistance content based on the multimodal work records using the AI ​​assistant; pushing the first assistance content to the personnel; and / or receiving work assistance requests input by the personnel; generating second assistance content based on the work assistance requests using the AI ​​assistant; pushing the second assistance content to the personnel; further including: when personnel use a work simulation sandbox, determining whether to use the simulation based on the real-time simulated content of the personnel's work simulation in the work simulation sandbox. Entering a standard auxiliary timing; upon entry, extracting auxiliary criteria from the real-time simulated content based on the auxiliary criteria extraction template of the standard auxiliary timing; determining the auxiliary simulation content based on the auxiliary criteria using the AI ​​assistant; generating an interactive control timeline for the auxiliary simulation content; conducting corresponding work simulations in the work simulation sandbox based on the auxiliary simulation content, and interacting with personnel during the simulation based on the interactive control timeline; wherein, generating the interactive control timeline for the auxiliary simulation content includes: obtaining the simulation progress timeline of the auxiliary simulation content; dividing multiple first target time intervals on the simulation progress timeline based on time interval division constraints; sequentially traversing each first target time interval according to the time sequence; each time the traversal reaches... In this process, multiple immersive perspectives are determined based on the motion and position changes of multiple simulated subjects within the first target time interval traversed on the simulation progress timeline. This determination includes: filtering key motions for each simulated subject's motion changes to obtain key motion changes for each subject; aligning the key motion changes with the position changes of each simulated subject in time to obtain an alignment sequence; traversing multiple motion change nodes in the alignment sequence sequentially; and each time a new motion change node is encountered, designating the simulated subject at the encountered motion change node as the first target, and the remaining simulated subjects as the second target. Two objectives; The process involves determining the first and second position change nodes for both the first and second targets from the alignment position change nodes in the alignment sequence of the traversed action change nodes; determining in the work simulation sandbox whether the simulation perspective of the second target when it is at the second position change node is visible to the traversed action change nodes when the first target is at the first position change node; where the simulation perspective includes: first-person perspective and third-person perspective; if visible, the simulation perspective of the second target when it is at the second position change node is used as the candidate immersion perspective; otherwise, the third-person perspective of the first target when it is at the first position change node is used as the candidate immersion perspective; and associating the candidate immersion perspective with the change time of the traversed action change nodes.After traversing all action change nodes, all candidate immersive viewpoints are integrated to obtain a viewpoint set. Based on set optimization constraints, the viewpoint set is optimized. Based on the optimized viewpoint set, multiple immersive viewpoints are determined. The set optimization constraints include: the change times associated with different candidate immersive viewpoints in the optimized viewpoint set are not the same; and the optimized viewpoint set has the fewest categories of simulated subjects belonging to different candidate immersive viewpoints. Immersion trigger constraints are generated for each immersive viewpoint. A second target time interval corresponding to the traversed first target time interval is determined from the blank time axis. Each immersive viewpoint and its respective immersion trigger constraint are set within the second target time interval. After traversing each first target time interval, the blank time axis with all immersive viewpoints and their respective immersion trigger constraints set is used as the interactive control time axis.

2. The AI ​​assistant-based personnel work assistance method as described in claim 1, characterized in that, The process of generating second auxiliary content based on an AI assistant and work assistance needs includes: vectorizing the work assistance needs to obtain a requirement content vector; calculating the similarity between the requirement content vector and document vectors in a vector database; and generating second auxiliary content based on the document content corresponding to the document vector with the highest similarity. The steps for obtaining document vectors from the vector database are as follows: receiving multi-format documents uploaded by personnel; wherein the formats of the multi-format documents include at least TXT, DOCX, and PDF; vectorizing the document content of the multi-format documents to obtain document vectors; wherein the vectorizing of the document content of the multi-format documents to obtain document vectors includes: performing text segmentation on the document content of the multi-format documents and applying an embedding model to convert it into document vectors.

3. The AI ​​assistant-based personnel work assistance method as described in claim 2, characterized in that, The embedding model includes at least the bge-large-zh-v1.5 Chinese embedding model.

4. The AI ​​assistant-based personnel work assistance method as described in claim 2, characterized in that, The vector database uses the PGVector vector library for vector storage.

5. The AI ​​assistant-based personnel work assistance method as described in claim 2, characterized in that, After generating the second auxiliary content based on the document content corresponding to the document vector with the highest similarity, the generated second auxiliary content is further optimized through the chatGLM API.

6. The AI ​​assistant-based personnel work assistance method as described in claim 2, characterized in that, When performing vector conversion on the content of multi-format documents, the system automatically identifies and extracts key information from the document content based on the element orientation specified by the user.

7. The AI ​​assistant-based personnel work assistance method as described in claim 1, characterized in that, The time interval division constraints include: each of the different first target time intervals completely contains at least N consecutive simulation time periods without overlap; where N is a positive integer greater than or equal to 2; and there is at least one attribute association relationship between each pair of simulation cases in the consecutive simulation time periods contained in the same first target time interval.

8. A personnel work assistance system based on an AI assistant, characterized in that, include: The first auxiliary module is used to assist personnel in their work based on an AI assistant. The AI ​​assistant is an artificial intelligence model trained using extensive work experience related to the personnel's work profile. The first auxiliary module assists personnel based on the AI ​​assistant by: acquiring the personnel's multimodal work records in real time; generating first auxiliary content based on the multimodal work records using the AI ​​assistant; pushing the first auxiliary content to the personnel; and / or receiving work assistance requests input by the personnel; generating second auxiliary content based on the work assistance requests using the AI ​​assistant; and pushing the second auxiliary content to the personnel. It also includes a second auxiliary module used to: when personnel use a work simulation sandbox, based on the personnel's work... The simulation module uses real-time simulated content in a work simulation sandbox to determine whether a standard auxiliary timing is required. If so, it extracts auxiliary criteria from the real-time simulated content based on the standard auxiliary timing's auxiliary criteria template. Using an AI assistant, it determines the auxiliary simulation content based on these criteria and generates an interactive control timeline for the auxiliary simulation content. Based on this auxiliary simulation content, it conducts corresponding work simulations in the work simulation sandbox and, during the simulation, interacts with personnel according to the interactive control timeline. The second auxiliary module generates the interactive control timeline for the auxiliary simulation content by: acquiring the simulation progress timeline of the auxiliary simulation content; and dividing the simulation progress timeline into multiple first objectives based on time interval constraints. Time intervals; sequentially traversing each first target time interval according to chronological order; each time it is traversed, multiple immersive perspectives are determined based on the action changes and position changes of multiple simulated subjects within the first target time interval traversed on the simulation progress time axis; wherein, determining multiple immersive perspectives based on the action changes and position changes of multiple simulated subjects within the first target time interval traversed on the simulation progress time axis includes: performing key action screening on the action changes of each simulated subject to obtain the key action changes of each simulated subject; aligning the key action changes and position changes of each simulated subject in time to obtain an alignment sequence; sequentially traversing multiple action change nodes in the alignment sequence according to the sequence order; each time it is traversed, the traversed... The simulated subject at the action change node is taken as the first target, and the remaining simulated subjects are taken as the second target. The first and second position change nodes for each target are determined from the alignment position change nodes in the alignment sequence of the traversed action change nodes. In the working simulation sandbox, it is determined whether the simulation perspective of the second target at the second position change node is visible to the traversed action change node when the first target is at the first position change node. The simulation perspective includes: first-person perspective and third-person perspective. If it is visible, the simulation perspective of the second target at the second position change node is taken as the candidate immersion perspective; otherwise, the third-person perspective of the first target at the first position change node is taken as the candidate immersion perspective.The process involves associating candidate immersive viewpoints with the time changes of traversed action change nodes; after traversing each action change node, integrating all candidate immersive viewpoints to obtain a viewpoint set; optimizing the viewpoint set based on set optimization constraints; determining multiple immersive viewpoints based on the optimized viewpoint set; wherein the set optimization constraints include: the time changes associated with different candidate immersive viewpoints in the optimized viewpoint set are not the same; and the optimized viewpoint set has the fewest categories of simulated subjects belonging to different candidate immersive viewpoints; generating immersion trigger constraints for each immersive viewpoint; determining the second target time interval corresponding to the traversed first target time interval from the blank time axis; setting each immersive viewpoint and its respective immersion trigger constraint within the second target time interval; after traversing each first target time interval, using the blank time axis with all immersive viewpoints and their respective immersion trigger constraints set as the interactive control time axis.

Images