Personal risk management and control decision-making method

Abstract

The invention relates to a personal risk management and control decision-making method. Target text data sent by a terminal and obtained through recognition and conversion based on user voice are obtained; performing decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data by adopting a personal risk management and control decision model to obtain a risk management and control decision result; and finally, the risk management and control decision result is sent to the terminal. According to the method, the user voice is acquired in real time and converted into the text, so that the request or the problem of the user can be quickly responded, and the information processing time is remarkably shortened. A personal risk management and control decision model is adopted, potential risks and user intentions can be identified through deep analysis of target text data, so that accurate decision suggestions are generated, and errors of subjective judgment are reduced.

Description

Technical Field

[0001] The present application relates to the technical field of risk management and control, and in particular to a personal risk management decision-making method. Background Art

[0002] With the rapid development of intelligent technology, voice recognition and processing technology has gradually become an important part of human-computer interaction. In many application scenarios, personal risk control has gradually received attention. In all walks of life, the technology of risk control combined with voice interaction has been widely used.

[0003] However, some voice interaction scenarios rely on human judgment and experience, which leads to certain subjectivity and errors. Summary of the invention

[0004] Based on this, it is necessary to provide a personal risk management decision-making method that can improve accuracy in response to the above technical problems.

[0005] In a first aspect, the present application provides a personal risk management decision-making method, the method comprising:

[0006] Obtain target text data sent by the terminal; the target text data is obtained by the terminal through recognition and conversion based on the user's voice;

[0007] Adopting the personal risk management decision model, decision suggestion processing, intention recognition processing and strategy deduction processing are carried out based on the target text data to obtain the risk management decision results; the risk management decision results are used to characterize the potential impact of the risk;

[0008] Send risk management and control decision results to the terminal.

[0009] In one embodiment, the personal risk management decision model includes a decision suggestion model, an intention judgment model and a strategy deduction model. The personal risk management decision model is used to perform decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain a risk management decision result, including:

[0010] Acquire multi-channel data from a pre-established knowledge graph; the multi-channel data includes at least one of market data, internal business indicator data, and competitor behavior data;

[0011] Input multi-channel data into the decision-making recommendation model for decision-making recommendation processing to obtain decision-making recommendation results; the decision-making recommendation results represent the specific action plan taken and its expected effect;

[0012] Input the target text data into the intention judgment model to perform user intention recognition processing to obtain the intention judgment result; the intention judgment result is used to represent the information currently required by the user;

[0013] The decision recommendation results and intention judgment results are input into the strategy deduction model for strategy deduction processing to obtain the risk management decision results; the risk management decision results include at least one of personnel mobilization, resource allocation and emergency response plan.

[0014] In one embodiment, the method further comprises:

[0015] Obtain user feedback information sent by the terminal;

[0016] The risk management decision results are adjusted based on user feedback information, and the personal risk management decision model is updated based on the adjusted risk management decision results.

[0017] In a second aspect, the present application also provides a personal risk management decision-making method, the method comprising:

[0018] Acquire user voice, and use the voice processing model to recognize and convert the user voice to obtain target text data;

[0019] Send the target text data to the server so that the server adopts the personal risk management decision model, performs decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data, and obtains the risk management decision result; the risk management decision result is used to characterize the potential impact of the risk;

[0020] The risk management and control decision results returned by the server are received, and the risk management and control decision results are converted into speech using a speech synthesis model to obtain speech output results.

[0021] In one embodiment, the speech processing model includes an acoustic model and a language model, and the speech processing model is used to perform recognition and conversion processing on the user's speech to obtain target text data, including:

[0022] Perform feature extraction processing on the user's voice to obtain feature extraction data;

[0023] Input the feature extraction data into the acoustic model for feature mapping processing to obtain a probability distribution result; the probability distribution result includes phoneme probability distribution and word probability distribution;

[0024] The probability distribution result is input into the language model for probability analysis and processing to obtain multiple candidate word sequences;

[0025] Each candidate word sequence is decoded to obtain the target text data.

[0026] In one embodiment, the acoustic model includes an input layer, a hidden layer, a recurrent layer, and an output layer. The feature extraction data is input into the acoustic model for feature mapping processing to obtain a probability distribution result, including:

[0027] Input the feature extraction data into the input layer for data conversion processing to obtain an input vector;

[0028] The input vector is input into the nonlinear activation function in the hidden layer for mapping to obtain sequence data;

[0029] Input the sequence data into the recurrent layer for sequence prediction processing to obtain the predicted sequence;

[0030] The prediction sequence is input into the output layer for probability distribution conversion to obtain the probability distribution result.

[0031] In one embodiment, the probability distribution result is input into the language model for probability analysis to obtain multiple candidate word sequences, including:

[0032] The probability distribution result is input into the language model for probability calculation processing to obtain the probability result of each sequence in the probability distribution result;

[0033] According to each probability result, each sequence is screened to obtain multiple candidate word sequences.

[0034] In one embodiment, the above decoding process of each candidate word sequence to obtain target text data includes:

[0035] Input each candidate word sequence and the corresponding probability into the decoding model to perform candidate sequence expansion processing, and obtain the probability of multiple updated sequences;

[0036] Determine the update sequence with the highest probability as the target text sequence;

[0037] The target text sequence is converted to obtain target text data.

[0038] In one embodiment, the method further comprises:

[0039] Perform visual conversion processing on the risk management and control decision results to obtain visual results;

[0040] Generate a user interface based on the visualization results.

[0041] In one embodiment, the method further comprises:

[0042] Based on the user interface, obtain user feedback information;

[0043] Send user feedback information to the server, which is used by the server to adjust the risk management decision results and update the personal risk management decision model based on the adjusted risk management decision results.

[0044] In a third aspect, the present application also provides a personal risk management decision-making device, which includes:

[0045] The data acquisition module is used to acquire the target text data sent by the terminal; the target text data is obtained by the terminal through recognition and conversion based on the user's voice;

[0046] The decision-making suggestion module is used to adopt the personal risk management decision-making model to perform decision-making suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain the risk management decision results; the risk management decision results are used to characterize the potential impact of the risk;

[0047] The result sending module is used to send the risk management and control decision results to the terminal.

[0048] In a fourth aspect, the present application also provides a personal risk management and decision-making device, which includes:

[0049] The voice acquisition module is used to acquire the user's voice and use the voice processing model to recognize and convert the user's voice to obtain the target text data;

[0050] The data sending module is used to send the target text data to the server so that the server adopts the personal risk management decision model to perform decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain the risk management decision result; the risk management decision result is used to characterize the potential impact of the risk;

[0051] The result receiving module is used to receive the risk management and control decision results returned by the server, and use the speech synthesis model to perform speech conversion processing on the risk management and control decision results to obtain speech output results.

[0052] In a fifth aspect, the present application further provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned methods of the first and second aspects when executing the computer program.

[0053] In a sixth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the methods of the first and second aspects described above are implemented.

[0054] In a seventh aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the method of the first aspect and the second aspect are implemented.

[0055] The above-mentioned personal risk management decision-making method obtains the target text data sent by the terminal through recognition and conversion based on the user's voice; adopts the personal risk management decision model to process decision suggestions, intention recognition and strategy deduction based on the target text data to obtain the risk management decision result; finally, the risk management decision result is sent to the terminal. This application can quickly respond to user requests or questions by acquiring user voice in real time and converting it into text, significantly shortening the information processing time. Enhanced decision accuracy: The personal risk management decision model can be used to identify potential risks and user intentions through in-depth analysis of the target text data, thereby generating accurate decision suggestions and reducing errors in subjective judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A diagram of an application environment of a personal risk management and decision-making method in an embodiment;

[0057] Figure 2 A flowchart of a personal risk management decision-making method in one embodiment;

[0058] Figure 3 A schematic diagram of a process for obtaining a risk management decision result in an embodiment;

[0059] Figure 4 A flowchart of a personal risk management decision-making method in another embodiment;

[0060] Figure 5 A flowchart of a personal risk management and control decision-making method applied to a terminal in one embodiment;

[0061] Figure 6 A schematic diagram of a process for obtaining target text data in one embodiment;

[0062] Figure 7 A schematic diagram of a process for obtaining a probability distribution result in one embodiment;

[0063] Figure 8 A schematic diagram of a process for obtaining multiple candidate word sequences in one embodiment;

[0064] Fig. 9 A schematic diagram of a process for obtaining target text data in another embodiment;

[0065] Fig.10 A flowchart of a personal risk management and control decision-making method applied to a terminal according to another embodiment;

[0066] Fig.11 A schematic diagram of a process for updating a personal risk management decision model according to an embodiment;

[0067] Fig.12It is a structural block diagram of a personal risk management and control decision-making device applied to a server in one embodiment;

[0068] Fig.13 It is a structural block diagram of a terminal personal risk management and control decision-making device applied in one embodiment;

[0069] Fig.14 is an internal structure diagram of a computer device in one embodiment;

[0070] Fig.15 FIG. 4 is a diagram showing the internal structure of a computer device in another embodiment. DETAILED DESCRIPTION

[0071] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0072] First of all, before specifically introducing the technical solution of the embodiments of the present application, the technical background based on the embodiments of the present application is introduced.

[0073] With the rapid development of science and technology and the widespread application of artificial intelligence technology, the intelligence of personal risk management and decision-making systems has become an inevitable trend in the development of the industry. In all walks of life, the technology of risk management through voice interaction has been widely used. For example, related technologies have introduced real-time audio into remote information review scenarios, and the response performance of the person being questioned in real-time questions and answers is used as an important consideration for review, which effectively improves the efficiency and accuracy of information review. Another related technology simulates users calling the target company, determines the risk category of the target company based on the outbound voice, and quantifies the risk category, making the company's financial risk more intuitive.

[0074] Based on this, this application provides a personal risk management decision-making method, aiming to solve the above-mentioned technical problems.

[0075] The personal risk management decision-making method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The server obtains the target text data sent by the terminal and obtained by recognition and conversion based on the user's voice; adopts a personal risk management decision model, and performs decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain the risk management decision result; finally, the risk management decision result is sent to the terminal. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones and tablets. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0076] In an exemplary embodiment, Figure 2 As shown, a personal risk management decision-making method is provided, which is applied to Figure 1 The server in the example is used for explanation, including the following S201 to S203. Among them:

[0077] S201, obtaining target text data sent by the terminal.

[0078] The target text data is obtained by the terminal through recognition and conversion based on the user's voice.

[0079] In the embodiment of the present application, obtaining target text data may include:

[0080] Voice input: Users use voice input through terminals (such as mobile phones, smart headphones, etc.) to express their needs or situations. Users can ask questions such as "What risks will I encounter on my way home at night?" or "How do I protect myself?"

[0081] Speech recognition: The terminal's built-in speech recognition module converts the user's voice signal into text data in real time. The module uses advanced speech recognition technology and natural language processing algorithms to ensure the accuracy and completeness of the conversion.

[0082] Target text data: After the conversion is completed, the generated text data is called target text data, which contains the specific intentions and situation descriptions expressed by the user.

[0083] S202, using the personal risk management decision model, performs decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain the risk management decision result.

[0084] Among them, risk management and control decision results are used to characterize the potential impact of risks.

[0085] In the embodiment of the present application, the following process is specifically included:

[0086] Decision making suggestion processing:

[0087] The acquired target text data is input into the personal risk management decision model. The system first analyzes the user's needs and potential risks and generates preliminary decision recommendations. This stage uses historical data and risk assessment algorithms to provide users with relevant information.

[0088] Intent recognition processing:

[0089] The system uses intent recognition technology to analyze the real needs behind the user's voice and determine the types of risks the user may face (such as personal safety, financial security, etc.). This step uses machine learning models to analyze large amounts of data to determine the context of the user's voice input.

[0090] Strategy deduction processing:

[0091] Based on the identified intentions and potential risks, the system implements strategy deduction, evaluates the potential impact of different decision paths, and comes up with the best risk management strategy. This includes simulating the effects of decisions under different scenarios to ensure that the most effective and feasible risk management solution is selected.

[0092] Risk management decision results:

[0093] The final risk management decision results characterize the potential impact of risks through specific indicators, including expected losses, possible consequences and recommended response measures.

[0094] S203, sending the risk management and control decision result to the terminal.

[0095] In the embodiment of the present application, the server sends the risk management decision result back to the user's terminal. This process may be carried out through application notification, text message, or voice feedback to ensure that the user can obtain information in a timely manner. Furthermore, the user can take corresponding actions based on the received decision results, and the system can also be designed to allow users to provide feedback on the decision results, thereby further optimizing subsequent decision models and algorithms.

[0096] The above-mentioned personal risk management decision-making method obtains the target text data sent by the terminal through recognition and conversion based on the user's voice; adopts the personal risk management decision model to perform decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain the risk management decision result; finally, the risk management decision result is sent to the terminal. The embodiment of the present application can quickly respond to user requests or questions by acquiring user voice in real time and converting it into text, which significantly shortens the information processing time. Enhanced decision accuracy: By adopting the personal risk management decision model, potential risks and user intentions can be identified through in-depth analysis of the target text data, thereby generating accurate decision suggestions and reducing errors in subjective judgment.

[0097] In an exemplary embodiment, based on the above embodiment, see Figure 3 The personal risk management decision model of the embodiment of the present application includes a decision suggestion model, an intention judgment model and a strategy deduction model. The personal risk management decision model is adopted. The embodiment of the present application involves a process of performing decision suggestion processing, intention recognition processing and strategy deduction processing based on target text data to obtain a risk management decision result, including the following S301 to S304. Among them:

[0098] S301, obtaining multi-channel data from a pre-established knowledge graph.

[0099] The multi-channel data includes at least one of market data, internal business indicator data and competitor behavior data.

[0100] In the embodiment of the present application, data is extracted from a pre-established knowledge graph. It is ensured that the extracted data includes at least one of market data, internal business indicator data, and opponent behavior data. The acquisition of multi-channel data will provide multi-dimensional information support for subsequent decision-making analysis.

[0101] S302, inputting the multi-channel data into the decision suggestion model for decision suggestion processing to obtain a decision suggestion result.

[0102] Among them, the decision-making recommendation results represent the specific action plan taken and its expected effects.

[0103] In the embodiment of the present application, the cleaned multi-channel data is input into the decision-making recommendation model. In the model, the data is analyzed and combined with the existing historical data and market rules.

[0104] Generate decision suggestions: The model generates decision suggestions based on the analysis results, including but not limited to: specific marketing action plans, possible future growth forecasts, and resource allocation suggestions.

[0105] Output decision recommendation results: Decision recommendation results will be output in a structured or unstructured form, representing specific action plans and their expected effects.

[0106] S303, inputting the target text data into the intention judgment model to perform user intention recognition processing to obtain an intention judgment result.

[0107] Among them, the intention judgment result is used to represent the information currently required by the user.

[0108] The natural language text data input by the user through the terminal device (such as mobile application, web page, etc.) is input into the intent judgment model.

[0109] Intent judgment processing: The intent judgment model uses natural language processing technology to analyze user intentions and needs. Identify the type of information required by users (such as product inquiries, service consultations, market trends, etc.).

[0110] Generate intent judgment results: This result will be used to accurately characterize the information currently required by the user and ensure the targetedness of subsequent strategies.

[0111] S304, inputting the decision suggestion results and the intention judgment results into the strategy deduction model for strategy deduction processing to obtain the risk management decision results.

[0112] Among them, the risk management decision results include at least one of personnel mobilization, resource allocation and emergency response plan.

[0113] In the embodiment of the present application, the decision recommendation result generated in step S302 and the intention judgment result in step S303 are jointly input into the strategy deduction model.

[0114] Strategy deduction analysis: The model combines the results of the first two steps to deduce the most suitable risk management decision. Analyze the possible impact of different strategies, including estimated effects, required resources and potential risks. Generate risk management decision results:

[0115] The results of risk management decisions may include: Personnel transfer: recommending that some employees of the marketing department be transferred to high-demand areas. Resource allocation: such as optimizing inventory allocation to cope with the upcoming sales peak. Emergency response plan: quickly formulate emergency measures in response to market changes to prevent potential risks.

[0116] The embodiment of the present application provides a systematic, data-driven risk management solution by integrating multi-channel data and intelligent decision-making models, which can effectively support accurate decision-making in a dynamic market, reduce potential risks, and improve overall operational efficiency.

[0117] In an exemplary embodiment, based on the above embodiment, see Figure 4The method of the embodiment of the present application further includes the following S401 to S402. Among them:

[0118] S401, obtaining user feedback information sent by a terminal.

[0119] In the embodiment of the present application, the system collects user feedback information through a user terminal (such as a mobile application, website or customer service system), and the feedback information may include: user satisfaction evaluation of the service (such as a score or comment). Specific service requests or complaint content (such as opinions on risk control measures). New demands or suggestions put forward by users. The collected user feedback information will be parsed and stored in a feedback database for subsequent analysis.

[0120] S402: Adjust the risk management decision result according to the user feedback information, and update the personal risk management decision model according to the adjusted risk management decision result.

[0121] In the embodiment of the present application, the collected user feedback information is input into the feedback adjustment processing module. The feedback adjustment processing module analyzes the content of the user feedback and identifies the relevance to the existing risk management decision results. The main steps include:

[0122] Sentiment analysis: Use natural language processing (NLP) technology to identify the emotional tendency (positive, negative, or neutral) of user feedback.

[0123] Content analysis: Extract key information (such as specific risks, issues, or needs mentioned by users) and categorize this information.

[0124] Based on the feedback analysis results, adjust the original risk management and decision-making results. For example:

[0125] If user feedback shows dissatisfaction with a certain risk control measure, the system can consider adjusting the measure to a solution that better meets user needs.

[0126] If users raise new potential risk points, the system should quickly update existing decisions to incorporate this information into the risk assessment.

[0127] Update the risk management decision model:

[0128] Update the risk management decision model based on the adjusted risk management decision results. The decision model update module analyzes the new decision results and optimizes the model parameters to adapt to user feedback information. Regularly record user feedback and adjusted results into the model database to form a feedback history to facilitate future decision adjustments and optimizations. This module can also support self-learning, and gradually improve the accuracy and effectiveness of risk management decisions through historical data accumulation and feedback information analysis.

[0129] In one embodiment, Figure 5 As shown, the present application also provides a personal risk management decision-making method, which is applied to Figure 1 The terminal shown in the figure comprises the following S501 to S503. Wherein:

[0130] S501, obtaining user speech, and using a speech processing model to perform recognition and conversion processing on the user speech to obtain target text data.

[0131] In the embodiment of the present application, the user issues a voice command through a terminal device (such as a smart phone or a smart speaker) equipped with a voice recognition function. For example, the user can say: "What should I pay attention to when I go home at night?".

[0132] The voice processing model integrated in the device processes the user's voice signal. The main steps include: Voice signal preprocessing: Use noise elimination and echo suppression technology to improve voice quality and ensure clear audio input. Feature extraction: Extract features from clear voice signals, including Mel-frequency cepstral coefficients (MFCCs), pitch, and other acoustic features to form data that can be used for model training.

[0133] Text conversion: The extracted features are mapped to text output through a trained deep learning model (such as CTC-RNN or Transformer architecture). The model is trained on a large amount of speech data to improve recognition accuracy.

[0134] Generate target text data:

[0135] After processing, the user's voice is converted into text data, such as "What do I need to pay attention to when I go home at night?" This text data will become the basis for subsequent decision-making processing.

[0136] S502, sending the target text data to the server so that the server adopts a personal risk management decision model, performs decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data, and obtains a risk management decision result.

[0137] Among them, risk management and control decision results are used to characterize the potential impact of risks.

[0138] In the embodiment of the present application, the generated target text data is sent to a remote server via a secure network protocol (such as HTTPS). This process ensures the security and integrity of the data.

[0139] After receiving the target text data, the server performs the following processing steps:

[0140] Application of personal risk management decision model: Utilize the built-in personal risk management decision model to make decision recommendations on the target text data, involving professional knowledge in key risk management areas.

[0141] Intent recognition processing: Natural language processing technology is used to deeply analyze user needs and identify their intentions. This process includes conversation context analysis and sentiment analysis to better understand the user's specific needs.

[0142] Strategy deduction processing: Based on the evaluation of user intentions, it combines historical data and existing knowledge base to generate specific response strategies. For example, it recommends users to choose a safe way to travel, such as "It is recommended that you choose public transportation and avoid walking alone."

[0143] Generation of risk management decision results: The system integrates the above analysis results to generate a risk management decision result, which usually includes: specific suggestions (such as avoiding remote areas, staying vigilant, etc.) and potential impact assessment of risks (such as the possible consequences of the behavior).

[0144] S503, receiving the risk management and control decision result returned by the server, and using the speech synthesis model to perform speech conversion processing on the risk management and control decision result to obtain a speech output result.

[0145] In the embodiment of the present application, the terminal device receives the risk management decision result returned by the server through the network. When receiving data, a data unpacking and verification mechanism is used to ensure that the information is accurate.

[0146] Speech synthesis model: After obtaining the risk management decision results, the system calls the built-in speech synthesis model to perform speech conversion on the results. The process steps include:

[0147] Text-to-speech conversion:

[0148] Use deep learning-based speech synthesis technology (such as Tacotron or WaveNet) to convert text information into natural and fluent speech output. The model is trained in a variety of scenarios to ensure that the generated speech has a natural intonation and speed.

[0149] Generate voice output results: The final output voice results will be played through the device speakers. For example, the generated voice may be: "To ensure safety, it is recommended that you choose public transportation and avoid remote roads."

[0150] In the embodiment of the present application, users can obtain timely risk management suggestions through voice input, and improve their safety awareness and ability in daily life. At the same time, the system can continuously optimize its decision-making model based on user feedback to provide users with more personalized and accurate services.

[0151] In an exemplary embodiment, based on the above embodiment, see Figure 6The speech processing model of the embodiment of the present application includes an acoustic model and a language model. The embodiment of the present application involves a process of using the speech processing model to recognize and convert the user's speech to obtain target text data, including the following S601 to S604.

[0152] in:

[0153] S601, performing feature extraction processing on the user's voice to obtain feature extraction data.

[0154] In the embodiment of the present application, the terminal uses a microphone or other voice collection device to collect the user's voice input. The voice data needs to be properly preprocessed, such as denoising and normalization, to improve the accuracy of subsequent analysis.

[0155] Feature extraction processing: Use acoustic feature extraction algorithms (such as Mel-frequency cepstral coefficients MFCC, filter bank energy, etc.) to analyze speech signals. Convert speech signals into a set of feature vectors, called feature extraction data. Feature vectors include time domain and frequency domain information, which can reflect the main characteristics of audio signals.

[0156] S602: Input the feature extraction data into the acoustic model for feature mapping processing to obtain a probability distribution result.

[0157] The probability distribution results include phoneme probability distribution and word probability distribution.

[0158] In an embodiment of the present application, the feature extraction data is input into a constructed acoustic model (such as a deep neural network model, DNN or a convolutional neural network, CNN) for processing.

[0159] Probability mapping processing: The acoustic model calculates the phoneme probability distribution and word probability distribution based on the feature data. The phoneme probability distribution indicates the possibility of each phoneme appearing in a specific feature vector group, and the word probability distribution indicates the probability of a word appearing in the speech signal.

[0160] Output probability distribution results: The returned results are two key probability distributions: phoneme probability distribution and word probability distribution, which provide the data basis for the subsequent steps of the language model.

[0161] S603: Input the probability distribution result into the language model for probability analysis and processing to obtain multiple candidate word sequences.

[0162] In an embodiment of the present application, the probability distribution result obtained in step S602 is input into a language model (such as an N-gram model, an LSTM model, or a Transformer model) for further analysis.

[0163] Candidate word sequence generation: Based on the probability distribution results, the language model analyzes and generates multiple candidate word sequences. This process usually involves ranking possible word sequences to evaluate their suitability in the context.

[0164] Output candidate word sequence: The generated candidate word sequence is sorted and output according to probability, providing multiple possible text options for subsequent decoding processing.

[0165] S604: Decode each candidate word sequence to obtain target text data.

[0166] In the embodiment of the present application, the decoding module analyzes and evaluates multiple candidate word sequences one by one, and finally determines the target text by aggregating relevant context information.

[0167] Output target text data: The system selects the candidate word sequence with the highest probability as the final output, forming clear and accurate text data to be presented to the user or used for subsequent applications.

[0168] In the embodiments of the present application, efficient and accurate speech-to-text conversion is achieved through precise speech feature extraction, acoustic model analysis, language model reasoning and decoding process.

[0169] In an exemplary embodiment, based on the above embodiment, see Figure 7 The acoustic model of the embodiment of the present application includes an input layer, a hidden layer, a recurrent layer and an output layer. The embodiment of the present application involves a process of inputting feature extraction data into the acoustic model for feature mapping processing to obtain a probability distribution result, including the following S701 to S704. Among them:

[0170] S701, input the feature extraction data into the input layer for data conversion processing to obtain an input vector.

[0171] In the embodiment of the present application, original feature extraction data is obtained from a data source (such as a sensor, a database, etc.), and the data is input into the input layer for conversion processing.

[0172] The conversion includes:

[0173] Data normalization: Standardize the feature values ​​to ensure that they are in the same range (such as 0 to 1). Feature selection: Select the features related to the task to eliminate redundant information and improve model performance. The data after the above processing generates the input vector (X) for calculation in the subsequent steps.

[0174] S702, inputting the input vector into the nonlinear activation function in the hidden layer for mapping processing to obtain sequence data.

[0175] In the embodiment of the present application, the input vector (X) is input to the hidden layer for nonlinear activation. Common activation functions include ReLU (rectified linear unit), sigmoid or tanh, etc. Each neuron processes the input value through the activation function to generate a set of hidden layer outputs to form sequence data (H). This sequence data (H) will be passed to the recurrent layer as input for deeper processing.

[0176] S703, inputting the sequence data into the recurrent layer for sequence prediction processing to obtain a predicted sequence.

[0177] In the embodiment of the present application, the sequence data (H) output by the hidden layer is input to the recurrent layer, and the commonly used recurrent neural network (RNN) architecture includes LSTM or GRU. In the recurrent layer, the input of each time step is processed and the output of the previous step is taken into account to generate the next prediction. This process is achieved through state transfer: the context information of the sequence is learned using a recurrent neural network to obtain a predicted sequence (P).

[0178] For example: For a given time series, the model will learn the time-dependent features in the sequence, so the output of the predicted sequence will be generated based on historical data.

[0179] S704, input the prediction sequence into the output layer for probability distribution conversion processing to obtain a probability distribution result.

[0180] In the embodiment of the present application, the prediction sequence (P) output by the recurrent layer is input to the output layer for probability distribution conversion processing. The prediction result is usually converted into probability distribution using a softmax function.

[0181] The output probability distribution results represent the probability of each category or possible outcome for decision making or subsequent analysis.

[0182] For example, in classification tasks, the prediction distribution can indicate the probability of each category being predicted, which helps in decision making.

[0183] In the embodiments of the present application, a complete prediction process is formed through feature extraction, activation function, sequence modeling and probability conversion, which provides strong support for practical applications in related fields.

[0184] In an exemplary embodiment, based on the above embodiment, see Figure 8 The embodiment of the present application relates to a process of inputting a probability distribution result into a language model for probability analysis and processing to obtain multiple candidate word sequences, including the following S801 to S802. Among them:

[0185] S801, input the probability distribution result into the language model for probability calculation processing to obtain the probability result of each sequence in the probability distribution result.

[0186] In the embodiment of the present application, the terminal first generates a preliminary probability distribution result from the input text or voice input. This process may include preliminary calculations based on the context, and the probability distribution generated by the pre-trained language model (such as GPT, BERT, etc.) can be represented as an N-dimensional vector, where N represents the number of possible words.

[0187] Probability calculation process:

[0188] The obtained probability distribution results are input into the selected language model for further probability calculation. The main process of this step includes:

[0189] Sequence modeling: The probability distribution of the input is used as context information and combined with the existing context information of the language model for processing. The language model used here (such as conditional random fields, LSTM, or Transformer, etc.) will evaluate the sequence association in the language and calculate the probability of each word sequence.

[0190] After being processed by the language model, the probability results of each candidate word sequence are generated to form a new probability distribution. This data structure will contain mappings such as sequences and their corresponding probabilities, for example:

[0191] Sequence 1: "The weather is good", probability: 0.75, Sequence 2: "The weather is very good", probability: 0.15.

[0192] Sequence 3: "Today's weather", probability: 0.10.

[0193] S802: Screen each sequence according to each probability result to obtain a plurality of candidate word sequences.

[0194] In the embodiment of the present application, the criteria for screening candidate word sequences are determined. This can be based on a set probability threshold or a specific screening algorithm. Common screening strategies include:

[0195] Fixed threshold method: Set a minimum probability threshold and filter out sequences with a probability lower than this value.

[0196] Top N method: select the top N sequences with the highest probability to ensure diversity.

[0197] Sequence screening process:

[0198] According to the set screening criteria, the obtained probability results are processed to screen out the candidate word sequences that meet the conditions. The main steps include: traversing the probability results: traversing each sequence and its corresponding probability, and comparing to determine whether it meets the screening conditions. Forming a candidate set: adding the sequences that meet the conditions to the candidate set. The multiple candidate word sequences that have been screened are sorted and output for use by subsequent processing modules (such as intent recognition, text generation, etc.).

[0199] In the embodiment of the present application, by combining probability calculation and language model processing, high-quality, high-confidence candidate word sequences can be generated in real time, thereby improving the performance and accuracy of downstream application systems.

[0200] In an exemplary embodiment, based on the above embodiment, see Fig. 9 The embodiment of the present application relates to a process of decoding each candidate word sequence to obtain target text data, including the following S901 to S903. Among them:

[0201] S901, input each candidate word sequence and the corresponding probability into the decoding model to perform candidate sequence expansion processing, and obtain the probability of multiple updated sequences.

[0202] In the embodiment of the present application, multiple candidate word sequences and their corresponding probability values ​​are obtained from the language model. These candidate sequences are usually composed based on the probability distribution of phonemes and words, and may contain multiple similar or repeated sequences.

[0203] Input decoding model:

[0204] The received candidate word sequence and the corresponding probability value are input into the decoding model (such as long short-term memory network LSTM, transformer Transformer, etc.).

[0205] Candidate sequence expansion analysis:

[0206] The decoding model uses contextual information to expand candidate word sequences. The expansion can include in-depth analysis of the candidate word's grammatical structure, contextual relevance, and possible spelling corrections. Multiple new update sequences are generated that are closer to the target text in terms of semantics and structure. The decoding model assigns a corresponding probability to each expanded update sequence to form a list of update candidate sequences for the subsequent selection process.

[0207] S902: Determine the update sequence with the highest probability as the target text sequence.

[0208] In the embodiment of the present application, the probability values ​​of each sequence are compared from the updated sequence output by the decoding model to find the sequence with the highest probability. A certain threshold mechanism (such as setting a minimum probability requirement) is adopted to ensure the rationality and effectiveness of the selected sequence.

[0209] Determine the target text sequence: Confirm the update sequence with the highest probability as the target text sequence and identify it as the text content that is most likely to represent the user's intention.

[0210] S903, converting the target text sequence to obtain target text data.

[0211] In the embodiment of the present application, after the target text sequence is confirmed, necessary conversion processing is applied to the sequence, which may include:

[0212] Text normalization: formatting text, such as case conversion, punctuation adjustment, etc.

[0213] Semantic Error Correction: Automatically correct potential spelling errors or grammatical issues to improve text quality.

[0214] Generate target text data: Output the processed text sequence in a predetermined format to form the final target text data (such as text files, dialog box contents, etc.). You can further choose to output in audio form, or send the text to other applications for further use.

[0215] In the embodiments of the present application, efficient conversion of speech to text is achieved by expanding the analysis of candidate sequences, selecting the best probability sequence and performing final text processing.

[0216] In an exemplary embodiment, based on the above embodiment, see Fig.10 The method of the embodiment of the present application further includes the following S1001 to S1002. Among them:

[0217] S1001, perform visualization conversion processing on the risk management and control decision results to obtain visualization results.

[0218] In the embodiment of the present application, after completing the prediction process, the risk management and control decision results generated by the system are stored in the database.

[0219] Visualize the decision results, including:

[0220] Data chart generation: Generate line graphs, bar graphs or pie charts based on decision results to show the proportions of different risk levels and their changing trends.

[0221] Indicator display: Visualize important indicators, such as risk level, potential loss, etc., to ensure that users can make decisions quickly.

[0222] Interactive function: Add interactive functions to the visualization results, allowing users to select different time periods, regions or other variables for in-depth analysis. Finally, the processed visualization results are presented to users through a graphical interface to ensure that the information is conveyed intuitively and easily.

[0223] S1002, generating a user interface according to the visualization result.

[0224] In the embodiment of the present application, based on the visualization results, the user interface module generates a user-friendly interface, and the different modules include:

[0225] Visualization display area: displays the visualization results of risk management and control.

[0226] Decision information area: displays detailed information of risk management decisions, including influencing factors, parameter settings, etc.

[0227] Feedback entry: Provide users with a feedback input box to allow users to enter their opinions and suggestions on decision results. On the user interface, users can intuitively see changes in risk data and their corresponding decisions.

[0228] The embodiments of the present application realize visualization of risk management decision results, making the decision-making process more transparent.

[0229] In an exemplary embodiment, based on the above embodiment, see Fig.11 The method of the embodiment of the present application further includes the following S1101 to S1102. Among them:

[0230] S1101, obtaining user feedback information based on the user interface.

[0231] In the embodiment of the present application, after seeing the visualization results, the user can provide feedback on the risk management decision results through the user interface. The specific feedback forms include: selecting the feedback type (such as "satisfied", "unsatisfied", "there are problems"). Enter detailed feedback information to explain the considerations for its decision or suggested improvements. A scoring mechanism can also be provided to give a score of 1 to 5. User feedback information can be recorded in real time for unified processing later.

[0232] S1102, sending user feedback information to the server, the user feedback information is used by the server to adjust the risk management decision results, and update the personal risk management decision model according to the adjusted risk management decision results.

[0233] In the embodiment of the present application, after the user submits feedback, the system will automatically package and format the user feedback information. Send the user feedback information to the server through a secure API interface. After receiving the feedback information, the server stores the information in the database for future analysis and model adjustment. The server will check the user feedback and adjust the risk management decision results. The main steps include: analyzing the differences between the user feedback information and the existing decision model. Update the risk management decision model based on the adjusted information, and retrain the model to adapt to actual needs.

[0234] In the embodiments of the present application, a more efficient solution is provided for risk management and control decisions through a visualization method and a user feedback mechanism to adapt to the rapidly changing market demands and risk environment.

[0235] In an exemplary embodiment, based on the above embodiment, the method of the embodiment of the present application further includes the following steps:

[0236] Step 1: The terminal obtains the user's voice, performs feature extraction processing on the user's voice, and obtains feature extraction data;

[0237] Step 2: The terminal inputs the feature extraction data into the input layer for data conversion processing to obtain an input vector; the input vector is input into the nonlinear activation function in the hidden layer for mapping processing to obtain sequence data; the sequence data is input into the recurrent layer for sequence prediction processing to obtain a predicted sequence; the predicted sequence is input into the output layer for probability distribution conversion processing to obtain a probability distribution result;

[0238] Step 3: The terminal inputs the probability distribution result into the language model for probability calculation processing to obtain the probability result of each sequence in the probability distribution result; based on each probability result, each sequence is screened to obtain multiple candidate word sequences;

[0239] Step 4: The terminal inputs each candidate word sequence and the corresponding probability into the decoding model to perform candidate sequence expansion processing, and obtains the probabilities of multiple update sequences; determines the update sequence with the highest probability as the target text sequence; and converts the target text sequence to obtain target text data;

[0240] Step 5: The server obtains the target text data sent by the terminal;

[0241] Step 6: The server obtains multi-channel data from the pre-established knowledge graph; inputs the multi-channel data into the decision suggestion model for decision suggestion processing to obtain the decision suggestion result; inputs the target text data into the intention judgment model for user intention recognition processing to obtain the intention judgment result; the intention judgment result is used to characterize the information currently required by the user; inputs the decision suggestion result and the intention judgment result into the strategy deduction model for strategy deduction processing to obtain the risk management decision result; and sends the risk management decision result to the terminal;

[0242] Step 7: The terminal performs visualization conversion processing on the risk management and control decision results to obtain visualization results; and generates a user interface based on the visualization results;

[0243] Step 8: The terminal obtains user feedback information based on the user interface; the terminal sends the user feedback information to the server, and the user feedback information is used by the server to adjust the risk management decision results, and update the personal risk management decision model according to the adjusted risk management decision results.

[0244] Based on the same inventive concept, the embodiment of the present application also provides a personal risk management decision device for implementing the personal risk management decision method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more personal risk management decision device embodiments provided below can refer to the limitations of the personal risk management decision method above, and will not be repeated here.

[0245] In one embodiment, Fig.12 As shown, a personal risk management and decision-making device 1200 is provided, including:

[0246] The data acquisition module 1201 is used to acquire the target text data sent by the terminal; the target text data is obtained by the terminal through recognition and conversion based on the user's voice;

[0247] The decision suggestion module 1202 is used to adopt the personal risk management decision model to perform decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain the risk management decision result; the risk management decision result is used to characterize the potential impact of the risk;

[0248] The result sending module 1203 is used to send the risk management and control decision results to the terminal.

[0249] In one embodiment, the personal risk management decision model includes a decision suggestion model, an intention judgment model and a strategy deduction model. The decision suggestion module includes:

[0250] A data acquisition unit, used to acquire multi-channel data from a pre-established knowledge graph; the multi-channel data includes at least one of market data, internal business indicator data, and competitor behavior data;

[0251] The decision-making unit is used to input multi-channel data into the decision-making recommendation model for decision-making recommendation processing to obtain decision-making recommendation results; the decision-making recommendation results represent the specific action plan taken and its expected effect;

[0252] The intention recognition unit is used to input the target text data into the intention judgment model to perform user intention recognition processing and obtain the intention judgment result; the intention judgment result is used to represent the information currently required by the user;

[0253] The strategy deduction unit is used to input the decision recommendation results and intention judgment results into the strategy deduction model for strategy deduction processing to obtain risk management decision results; the risk management decision results include at least one of personnel mobilization, resource allocation and emergency response plan.

[0254] In one embodiment, the above device further comprises:

[0255] An information acquisition module is used to acquire user feedback information sent by the terminal;

[0256] The adjustment module is used to adjust the risk management decision results according to user feedback information, and update the personal risk management decision model according to the adjusted risk management decision results.

[0257] In one embodiment, see Fig.13 The present application also provides a personal risk management and decision-making device 1300, which includes:

[0258] The speech acquisition module 1301 is used to acquire the user's speech and use the speech processing model to perform recognition and conversion processing on the user's speech to obtain target text data;

[0259] The data sending module 1302 is used to send the target text data to the server, so that the server adopts the personal risk management decision model, performs decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data, and obtains the risk management decision result; the risk management decision result is used to characterize the potential impact of the risk;

[0260] The result receiving module 1303 is used to receive the risk management and control decision results returned by the server, and use the speech synthesis model to perform speech conversion processing on the risk management and control decision results to obtain a speech output result.

[0261] Each module in the above-mentioned personal risk management and decision-making device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0262] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.14As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a personal risk management decision method is implemented.

[0263] In one embodiment, a computer device is provided. The computer device may also be a terminal. The internal structure diagram thereof may be as follows: Fig.15 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a personal risk management decision method is realized. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0264] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0265] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A personal risk management decision-making method, characterized in that: The method comprises: Acquire target text data sent by the terminal; the target text data is obtained by the terminal through recognition and conversion based on user voice; Using a personal risk management decision model, decision suggestion processing, intention recognition processing and strategy deduction processing are performed based on the target text data to obtain a risk management decision result; the risk management decision result is used to characterize the potential impact of the risk; The risk management and control decision result is sent to the terminal.

2. The method according to claim 1, characterized in that The personal risk management and control decision model includes a decision suggestion model, an intention judgment model and a strategy deduction model. The personal risk management and control decision model is used to perform decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data to obtain a risk management and control decision result, including: Acquire multi-channel data from a pre-established knowledge graph; the multi-channel data includes at least one of market data, internal business indicator data, and competitor behavior data; Inputting the multi-channel data into the decision suggestion model for decision suggestion processing to obtain a decision suggestion result; the decision suggestion result represents the specific action plan taken and its expected effect; Inputting the target text data into the intention judgment model to perform user intention recognition processing to obtain an intention judgment result; the intention judgment result is used to represent the information currently required by the user; The decision recommendation result and the intention judgment result are input into the strategy deduction model for strategy deduction processing to obtain the risk management decision result; the risk management decision result includes at least one of personnel mobilization, resource allocation and emergency response plan.

3. The method according to any one of claims 1 to 2, characterized in that: The method further comprises: Obtain user feedback information sent by the terminal; The risk management decision result is adjusted according to the user feedback information, and the personal risk management decision model is updated according to the adjusted risk management decision result.

4. A personal risk management decision-making method, characterized in that: The method comprises: Acquire user voice, and use a voice processing model to perform recognition and conversion processing on the user voice to obtain target text data; The target text data is sent to a server, so that the server adopts a personal risk management decision model, performs decision suggestion processing, intention recognition processing and strategy deduction processing based on the target text data, and obtains a risk management decision result; the risk management decision result is used to characterize the potential impact of the risk; The risk management and control decision result returned by the server is received, and a speech synthesis model is used to perform speech conversion processing on the risk management and control decision result to obtain a speech output result.

5. The method according to claim 4, characterized in that The speech processing model includes an acoustic model and a language model. The speech processing model is used to perform recognition and conversion processing on the user speech to obtain target text data, including: Performing feature extraction processing on the user voice to obtain feature extraction data; Inputting the feature extraction data into the acoustic model for feature mapping processing to obtain a probability distribution result; the probability distribution result includes a phoneme probability distribution and a word probability distribution; Inputting the probability distribution result into the language model for probability analysis and processing to obtain multiple candidate word sequences; Decoding is performed on each of the candidate word sequences to obtain the target text data.

6. The method according to claim 5, characterized in that The acoustic model includes an input layer, a hidden layer, a recurrent layer and an output layer. The feature extraction data is input into the acoustic model for feature mapping processing to obtain a probability distribution result, including: Inputting the feature extraction data into the input layer for data conversion processing to obtain an input vector; Inputting the input vector into the nonlinear activation function in the hidden layer for mapping processing to obtain sequence data; Inputting the sequence data into the recurrent layer for sequence prediction processing to obtain a predicted sequence; The predicted sequence is input into the output layer for probability distribution conversion processing to obtain the probability distribution result.

7. The method according to claim 6, characterized in that The probability distribution result is input into the language model for probability analysis to obtain multiple candidate word sequences, including: Inputting the probability distribution result into the language model for probability calculation processing to obtain the probability result of each sequence in the probability distribution result; According to each of the probability results, each of the sequences is screened to obtain a plurality of candidate word sequences.

8. The method according to claim 7, characterized in that The decoding process of each candidate word sequence to obtain the target text data includes: Inputting each of the candidate word sequences and the corresponding probabilities into a decoding model to perform candidate sequence expansion processing to obtain the probabilities of multiple updated sequences; Determine the update sequence with the highest probability as the target text sequence; The target text sequence is converted to obtain the target text data.

9. The method according to any one of claims 4 to 8, characterized in that: The method further comprises: Performing visualization conversion processing on the risk management and control decision result to obtain a visualization result; A user interface is generated according to the visualization result.

10. The method according to claim 9, characterized in that The method further comprises: Based on the user interface, obtaining user feedback information; The user feedback information is sent to the server, and the user feedback information is used by the server to adjust the risk management decision result, and update the personal risk management decision model according to the adjusted risk management decision result.

Images