Bank user-oriented interaction method and system and medium

Abstract

The invention relates to the technical field of software, and provides a bank user-oriented interaction method and system and a medium, and the method comprises the steps: obtaining a user session text; identifying whether the session text of the user contains sensitive information or not; if not, the session text is sent to a confusion model, and corresponding reply content is generated; and judging whether the reply content is compliant or not, and if yes, feeding back the corresponding reply content on the banking business interface. The reply content is generated through the confusion model, and the user experience is improved.

Description

Technical Field

[0001] The present application relates to the field of software technology, and in particular to an interaction method, system and medium for bank users. Background Art

[0002] In the existing technology, more and more bank services have enabled software virtual customer service to communicate with users. The software of the existing technology builds a database based on bank products, which better solves the user's question-based consultation and greatly improves the efficiency of sales consultation for bank products. However, the limitation of the existing technology is that it only interacts with product content, lacks risk identification ability during the interaction, and the interaction process is rigid, resulting in poor user experience. When the user's message is not within the preset question range, the conversation usually cannot proceed, and thus it is impossible to deeply explore customer needs.

[0003] Therefore, the prior art lacks a technical solution that is capable of establishing a dialogue scenario with users, and is a dialogue capability that further explores user needs within the scope of banking business scenarios. Summary of the invention

[0004] In order to solve or partially solve the problems existing in the related art, the present application provides an interactive method for bank users to improve the user experience.

[0005] The first aspect of the present application provides an interaction method for bank users, comprising:

[0006] Get user session text;

[0007] Identify whether the user's conversation text contains sensitive information; if not, send the conversation text to the obfuscation model to generate corresponding reply content;

[0008] Determine whether the reply content is compliant, and if so, feedback the corresponding reply content on the banking business interface.

[0009] Optionally, the obtaining of the user conversation text includes:

[0010] A session service selected by a user on a banking service interface is received, and a session in the session service is encrypted.

[0011] Optionally, before identifying whether the user's conversation text contains sensitive information, the step further includes:

[0012] Determining whether the conversation text contains personal privacy information;

[0013] If personal privacy information is included, the personal privacy content in the conversation text is replaced to obtain the conversation text after the privacy content is replaced.

[0014] Optionally, the method includes: determining to generate a language database corresponding to the reply content according to the conversation service selected by the user.

[0015] Optionally, sending the conversation text to the obfuscation model to generate corresponding reply content includes:

[0016] Performing semantic analysis on the conversation text, classifying the conversation text into tasks according to the semantic analysis result, and determining the task type corresponding to the conversation text;

[0017] According to the task type and preset priority rules, the conversation text is input into the corresponding confusion model to generate the corresponding reply content.

[0018] Optionally, inputting the conversation text into the corresponding confusion model to generate the corresponding reply content includes:

[0019] Inputting the conversation text into the corresponding obfuscation model to generate pre-reply content;

[0020] The reply content is generated using the language model according to the conversation service selected by the user and the pre-reply content.

[0021] Optionally, the determining whether the reply content is compliant includes:

[0022] According to the conversation service selected by the user in the banking service interface, it is determined whether the reply content exceeds the service scope of the conversation service.

[0023] Optionally, obtaining the user conversation text includes:

[0024] Obtain the user's voice conversation data;

[0025] Perform speech recognition on the speech conversation data to obtain a conversation text.

[0026] The second aspect of the present application provides an interactive system for bank users, comprising:

[0027] The acquisition module is used to obtain the user session text;

[0028] A generation module, used to identify whether the user's conversation text contains sensitive information; if not, the conversation text is sent to the obfuscation model to generate corresponding reply content;

[0029] The compliance judgment module is used to judge whether the reply content is compliant, and if so, to feedback the corresponding reply content on the banking business interface.

[0030] A third aspect of the present application provides an electronic device, including:

[0031] Processor; and

[0032] A memory stores executable codes thereon, and when the executable codes are executed by the processor, the processor is caused to execute the interaction method as described above.

[0033] A fourth aspect of the present application provides a computer-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor is caused to execute the interaction method as described above.

[0034] The technical solution provided by this application may have the following beneficial effects:

[0035] Compared with the existing technology, the present application utilizes a confusion model to achieve targeted and diverse user conversations. Therefore, the interaction process with the user is not limited to the preset scope of questions, thereby improving the user experience; improving the sales consulting efficiency of the bank's virtual customer service, making the interactive language richer and the user experience better; at the same time, based on the judgment of sensitive information, the risk identification ability is stronger during the user's interaction with the virtual customer service.

[0036] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and other objects, features and advantages of the present application will become more apparent by describing in more detail exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the present application.

[0038] Figure 1 It is a flow chart of an interactive method for bank users shown in an embodiment of the present application.

[0039] Figure 2 This is a schematic diagram of a virtual customer service user session of Example 1 shown in the embodiments of the present application.

[0040] Figure 3 This is a schematic diagram of a virtual customer service user session of Example 2 shown in the embodiments of the present application.

[0041] Figure 4 It is a structural diagram of an interactive system for bank users shown in an embodiment of the present application.

[0042] Figure 5 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application. DETAILED DESCRIPTION

[0043] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0044] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0045] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0046] In the related art, bank virtual dialogue customer service has poor risk identification of sensitive information, personal privacy, and business scope exceeding during user conversation interaction. In addition, the existing bank virtual customer service has rigid responses during user conversation interaction, resulting in poor user experience. For example, it only responds to the content of the conversation question, and when the user conversation exceeds the business scope, it can only rigidly respond with "This topic is beyond my business scope." There is no other humanized explanation to improve the user experience.

[0047] In view of the above problems existing in the process of interaction between bank virtual customer service and user conversation, the embodiment of the present application proposes an interaction method for bank users, which can be deployed in a computer system or a mobile terminal system and implemented through a corresponding website or application program, such as Figure 1 As shown, including:

[0048] Step S100: Obtain user conversation text.

[0049] The bank virtual customer service in the embodiment of the present application can be set on the bank website or the bank application of the mobile terminal. The user can find the corresponding bank virtual customer service on the bank website or the application business interface. The user can conduct a conversation by clicking the virtual customer service logo on the business interface. The user's conversation text in the embodiment of the present application refers to the text content corresponding to the current conversation entered by the user on the business interface where the bank virtual customer service is located. The bank website or application can be set up with at least one bank virtual customer service corresponding to a business. Different businesses can correspond to different bank virtual customer services. Specifically, the corresponding bank virtual customer service is set in different business interfaces. The bank virtual customer service on the current business interface can only reply to questions related to the current business; a bank virtual customer service with comprehensive business conversation capabilities can also be set up. The types of bank virtual customer service can be set up according to the bank's business needs. For example, there can be loan business, financial management business, deposit and withdrawal business, work card business and other types.

[0050] Step S200, identifying whether the user's conversation text contains sensitive information; if not, sending the conversation text to the obfuscation model to generate corresponding reply content.

[0051] When a user initiates a session with the virtual customer service of a bank website or application, the bank's software system first encrypts the session text at the application layer to ensure the security of the information during transmission. Subsequently, the current session text at the application layer is pushed to the security protection layer for user session identification and reply processing. The specific contents are as follows.

[0052] First, the encrypted conversation text content is decrypted. Then, the conversation content is preliminarily identified and processed using a locally deployed small model, including simple semantic recognition and sensitive information interception. The locally deployed small model is a semantic recognition model, which can be used for simple semantic recognition of conversation text. During the recognition process, once sensitive information is found, the system will immediately block the conversation and notify the user. When the conversation text involves sensitive information, the system directly extracts the corresponding sensitive information speech information from the database and feeds it back to the user. The system deployed in the embodiment of the present application includes a sensitive information corresponding speech database, which may contain multiple reply speech data for the same type of sensitive information.

[0053] After ensuring that the message content of the conversation text does not contain sensitive information, the current conversation text enters the model obfuscation stage. At this stage, the system will call different large models for each user's conversation text according to the preset strategy to give full play to the expertise of each model, while avoiding over-reliance on a single model and reducing the risk of service interruption due to model errors or unavailability. In addition, all calls are made through HTTPS encrypted channels to enhance the security of data transmission. Obfuscation models can include at least two AIGC (Artificial Intelligence Generated Content) models such as ChatGLM (Chat Generative Language Model), Wenxin Yiyan, Tongyi Qianwen, Hunyuan AI Large Model, ChatGPT (Chat Generative Pre-trained Transformer), etc.

[0054] Step S300: determine whether the reply content is compliant, and if so, feedback the corresponding reply content on the banking business interface.

[0055] In the embodiment of the present application, compliance means that the reply content needs to comply with the software specifications or laws and regulations of the bank. The compliance judgment function can be realized through corresponding codes and a database containing non-compliant content. For example, a non-compliant content database is established according to the contents of the bank software specifications and laws and regulations, and the reply content is divided into multiple semantic segments according to semantics. When a semantic segment corresponds to the data in the non-compliant content database, it means that the reply content contains non-compliant content, and the reply content is non-compliant. Otherwise, it is compliant. If it does not comply with the rules, directly reply with relevant non-compliant words. Bank software specifications can be industry specifications and virtual customer service business specifications, and the reply content of virtual customer service cannot exceed the specifications. The virtual customer service business specifications are specific. The bank software contains multiple virtual customer services, and different virtual customer services have different business scopes. Therefore, virtual customer services cannot reply to content that exceeds the corresponding business scope. For example, loan virtual customer service can only reply to loan-related issues.

[0056] Compared with the existing technology, this application uses the obfuscation model to achieve targeted and diverse user dialogues, so the interaction process with users is not limited to the preset question range, which improves the user experience; improves the sales consultation efficiency of bank virtual customer service, the interactive language is richer, and the user experience is better; at the same time, based on the judgment of sensitive information, the risk identification ability is stronger during the user's interaction with virtual customer service. In addition, through compliance judgment, the security risk of bank software is improved.

[0057] Optionally, the obtaining of the user conversation text includes:

[0058] A session service selected by a user on a banking service interface is received, and a session in the session service is encrypted.

[0059] In an embodiment of the present application, the banking software includes at least two types of conversation services, each conversation service has a corresponding virtual customer service, and the reply contents between the virtual customer services are not interoperable. When the user selects the conversation service corresponding to the bank's virtual customer service, that is, after the user initiates a conversation, the system automatically encrypts the conversation to ensure information security. After that, all user conversations sent by the user in the conversation are encrypted first before performing other subsequent operations.

[0060] Optionally, before identifying whether the user's conversation text contains sensitive information, the step further includes:

[0061] Determining whether the conversation text contains personal privacy information;

[0062] If personal privacy information is included, the personal privacy content in the conversation text is replaced to obtain the conversation text after the privacy content is replaced.

[0063] In the embodiment of the present application, the personal privacy information identification stage needs to be carried out in the system security protection layer. In the embodiment of the present application, the received user's conversation text is identified using a semantic recognition model to identify whether the conversation text contains personal privacy information. When it comes to the user's personal privacy information, personal privacy desensitization is performed. The rule of desensitization is content replacement, replacing all personal privacy information in the conversation text with meaningless content or symbols. For example, personal privacy information involves a phone number, and the replaced conversation text may be displayed in the original phone number as a meaningless string of numbers (00000000000, 0000029285810, etc.) or symbols (***********, ###########, etc.); personal privacy information involves a name, and the replaced conversation text may be displayed in the original phone number as a virtual name (Zhang San, Li Si, etc.) or a symbol (Zhang*, Li##, etc.). If the conversation text contains personal privacy information, after the above replacement, the system sends the replaced conversation text to step S200 for sensitive information identification; if the conversation text does not involve personal privacy information, the conversation text is directly fed back to step S200.

[0064] Optionally, the method includes: determining to generate a language database corresponding to the reply content according to the conversation service selected by the user.

[0065] In the embodiment of the present application, different conversation services are deployed with corresponding language databases, which include relevant information in different conversation services. The language databases corresponding to the financial management services and their bank virtual customer service products contain financial management information, such as whether the XX financial management product is currently on sale, sales rules, and yields. The language databases corresponding to the financial management services and their bank virtual customer service products contain loan information, such as loan interest rates, whether untrustworthy persons can obtain loans, and so on.

[0066] Furthermore, in an embodiment of the present application, it may also include: different banks have different roles of virtual customer service staff, the virtual customer service role corresponding to the wealth management business is male, and the virtual customer service role corresponding to the loan business is female. By setting the role, voice replies with different tones can be made to the conversation, and the reply scripts can also be distinguished. For example, in the process of replying to user conversations, different roles reply with different expressions, symbols, etc., thereby providing users with a better conversation interaction experience.

[0067] Optionally, sending the conversation text to the obfuscation model to generate corresponding reply content includes:

[0068] Performing semantic analysis on the conversation text, classifying the conversation text into tasks according to the semantic analysis result, and determining the task type corresponding to the conversation text;

[0069] According to the task type and preset priority rules, the conversation text is input into the corresponding confusion model to generate the corresponding reply content.

[0070] In the embodiment of the present application, the semantic analysis here is more detailed than the semantic recognition before sensitive information processing. It will identify in detail the topics and specific needs of the questions that need to be answered in the user's conversation text, and further classify the conversation text according to the identified topics and needs to determine the corresponding task type, which includes language translation, sentiment analysis, business problem solving, etc.

[0071] In the process of developing banking software, a model capability database is established for the AIGC model (referred to as the big model) in the confusion model to record the task types and historical performance that each big model is good at handling. Set priorities for different types of tasks to ensure that key tasks can be assigned to the most appropriate AIGC model first. The priorities of different task types are, for example, business question answering> language translation> sentiment analysis. According to the historical performance and current status of the big model, a weight is assigned to each big model, and a big model with a high weight is more likely to be selected. At the same time, when selecting a big model to process conversation text, it is also necessary to dynamically select the most appropriate model based on the current system load, model status, and task requirements. It is also necessary to conduct A / B tests regularly to compare the effects of different big models on the same task in order to optimize the model allocation strategy; design a fault-tolerant mechanism so that when the preferred big model is unavailable, it can automatically switch to the alternative big model to ensure service continuity; monitor the performance indicators of each big model, that is, the model status, such as response time, error rate, etc. in real time, so as to adjust the allocation strategy in a timely manner.

[0072] For example, the current confusion model includes three large models, A, B, and C. Among them, A's historical performance shows that it is suitable for processing business problem answers, B's historical performance shows that it is suitable for processing language translation, and C's historical performance shows that it is suitable for processing sentiment analysis. The current system load is good, and the status of the large models is A>B>C. The task type of the current conversation text is sentiment analysis. In order to process the conversation text as quickly as possible, the A model is used to process it. Similarly, when the system load status is extremely poor, the current conversation text is assigned to the large model with a better model status. In addition, when it is determined that the current conversation text contains multiple task types, the current conversation text can be semantically segmented according to the task type. For example, "What is the Chinese for aa Financial Products? What is the yield of this product?" The segmented easy segment "What is the Chinese for aa Financial Products?" can be input into the B model. After the B model translates, according to the translation result, the semantic segment "What is the yield of this product" is input into the A model in combination with the context, and finally fed back to the user.

[0073] Optionally, inputting the conversation text into the corresponding confusion model to generate the corresponding reply content includes:

[0074] Inputting the conversation text into the corresponding obfuscation model to generate pre-reply content;

[0075] The reply content is generated using the language model according to the conversation service selected by the user and the pre-reply content.

[0076] In the embodiment of the present application, a language model is also included, such as a large language model (LLM), which is pre-trained by using preset rules and data from a product database and a marketing database of a bank's corresponding business. When the user conversation text is obfuscated by the model, a pre-reply content is generated. The system determines the scenario of the language model reply based on the conversation business corresponding to the current conversation, and generates the reply content in combination with the pre-reply content.

[0077] Optionally, the determining whether the reply content is compliant includes:

[0078] According to the conversation service selected by the user in the banking service interface, it is determined whether the reply content exceeds the service scope of the conversation service.

[0079] In this embodiment, each business of the bank has its own corresponding virtual customer service, and the business scope of each virtual customer service is different. According to the user's selection of virtual customer service for loans, financial management and other businesses on the bank's business interface, it is judged whether the reply content exceeds the business scope of the corresponding virtual customer service. If it exceeds the scope, it is not compliant, otherwise it is compliant.

[0080] Optionally, obtaining the user conversation text includes:

[0081] Obtain the user's voice conversation data;

[0082] Perform speech recognition on the speech conversation data to obtain a conversation text.

[0083] In this embodiment, the user's input to the virtual customer service can be text conversation data or voice conversation data. If it is voice conversation data, the system needs to perform voice recognition on the voice conversation data first to obtain the conversation text. During the voice recognition process, dialects can also be recognized to enhance practicality.

[0084] The solution of the present application will be described below in conjunction with specific embodiments.

[0085] In Example 1, the user clicks on the virtual customer service icon on the bank loan business interface. Figure 2 As shown in the figure, the virtual customer service will give the user a fixed feedback in the conversation interface: "Hello, I am a bank loan assistant. Thank you for your attention to our bank's loan product series. Do you have any loan needs?" A conversation is established, and the content of the virtual customer service's reply will be marked as generated by AI.

[0086] At this point, the system automatically encrypts the conversation. Users can input through voice or text from below. This time, enter "Ask your account manager to contact me at 186AAAA1254" in the input box at the bottom. The system obtains the conversation text, performs semantic recognition, and finds that the conversation text contains personal privacy. The personal privacy part "186AAAA1254" in the conversation text is desensitized and replaced to obtain the replaced conversation text "Ask your account manager to contact me at 186****1254" and displays it in the conversation. Further semantic recognition is used to determine whether the replaced conversation text contains sensitive information. The user conversation does not contain sensitive information. The replaced conversation text is then subjected to detailed semantic analysis. It is found that the conversation text does not contain specific questions or content that requires customer service to answer. The conventional security interception words are used to reply "The assistant noticed that you mentioned information related to the contact information. For security reasons, the assistant has desensitized and intercepted the message. It is recommended that you do not send such information."

[0087] In Example 2, the user clicks on the virtual customer service logo in the bank fund financial management service interface, and the virtual customer service feedbacks the user with fixed words in the conversation interface to establish a conversation. The user asks for relevant information about the AA fund in the conversation. During the conversation, Figure 3 As shown, the user inputs "You had a salesperson tell me the profit was 5.8 before", and the virtual customer service identifies the conversation text, first determines that the conversation text does not involve personal privacy information, and then the system determines that the conversation text does not involve sensitive information. The conversation text does not involve sensitive information, and then proceeds to the next step, performs semantic analysis on the conversation text, determines that the task type of the conversation text is business problem answering, obtains the large models (A, B, C) that can be applied to the current confusion model, determines the historical performance and current status of all large models on business problem answering, among which A has the best historical performance in business problem answering, followed by B, and finally C. The current status is A load 60%, B load 45%, and C load 30%. The preset rule of the system is that when the large model load does not exceed 90%, the large model with the best historical performance is given priority.

[0088] Combined with the context semantics, the text of this conversation is input into A, and A outputs the pre-response content "Regarding the rate of return of the aa fund, it may be that the previous salesperson mentioned the highest expected rate of return during the introduction. In fact, the performance benchmark of the "aa fund" is 3.5% annualized. I need to clarify to avoid any misunderstanding. First of all, the aa fund... When bank sales staff introduce fund products, they may be affected by various factors such as market environment, sales strategy, personal understanding, etc., resulting in deviations or misunderstandings when conveying information. Therefore, when listening to the sales staff's introduction, investors should think rationally and make comprehensive considerations based on their own risk tolerance and investment goals. In summary, the performance benchmark of the "aa fund" is 3.5% annualized.", the pre-response content of the large model in the confusion model has too much content and is not suitable as the customer service conversation reply content. The customer service reply to the user needs to be concise.

[0089] Therefore, the system inputs the pre-reply content into the large language model corresponding to the pre-trained fund wealth management business, and obtains the reply content according to the preset rules. For example, the system only retains the first paragraph of the pre-reply content, and then removes the repeated content of the conversation text in the first paragraph, and the obtained text is combined with the product or marketing database to generate the reply content, or, only the summary content in the pre-reply content is retained, and other detailed explanations, repeated summaries and repeated content of the conversation text are deleted, and the obtained text is combined with the product or marketing database to generate the reply content. The final reply content is "It may be that the previous sales mentioned the highest expected rate of return during the introduction. In fact, the performance benchmark of "aa fund" is 3.5% annualized. I need to clarify to avoid misunderstanding. What do you think of this level of return?"

[0090] Finally, whether the reply content is compliant. Compliance in this embodiment refers to whether the reply content exceeds the corresponding business scope. Through context analysis and understanding, the reply content does not exceed the scope of fund management business, and the reply content is displayed in the conversation.

[0091] Figure 3 The second conversation text sent by the user shown in the figure has the same processing logic as the first conversation text, and the specific process will not be elaborated in detail. In addition, multiple conversation texts in the same conversation may be processed using different large models. During the processing, except for the first conversation text of the conversation, other conversation texts need to be analyzed in combination with the contextual semantic context, and the large model has the function of combining contextual semantic recognition. Therefore, in the actual application process, when the latter user conversation text and the previous user conversation text are applied with the same large model, the latter user conversation text and the previous user conversation text are conversation texts sent by two consecutive users, for example Figure 3For the two conversation texts displayed in , the system can directly input the latter user conversation text into the big model. If the two user conversation texts need to use different big models according to the rules, before the latter user conversation text is input into the big model, all conversation texts sent by the user before the conversation text in the conversation are input into the big model in the order of sending, so as to facilitate the big model to perform contextual semantic recognition.

[0092] The present application also provides an interactive system for bank users. Figure 4 As shown, including:

[0093] The acquisition module 401 is used to obtain the user conversation text;

[0094] A generation module 402 is used to identify whether the user's conversation text contains sensitive information; if not, the conversation text is sent to the obfuscation model to generate corresponding reply content;

[0095] The compliance judgment module 403 is used to judge whether the reply content is compliant, and if so, to feedback the corresponding reply content on the banking business interface.

[0096] The system corresponds to the above method and can execute all processing logic of the above interaction method through corresponding modules. Therefore, it will not be described in detail.

[0097] Figure 5 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application.

[0098] See also Figure 5 , an embodiment of the present application provides an electronic device 500 including a memory 501 and a processor 502.

[0099] The processor 502 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0100] The memory 501 may include various types of storage units, such as system memory, read-only memory (ROM) and permanent storage devices. Among them, ROM can store static data or instructions required by the processor 502 or other modules of the computer. The permanent storage device can be a readable and writable storage device. The permanent storage device can be a non-volatile storage device that will not lose the stored instructions and data even if the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device can be a removable storage device (such as a floppy disk, optical drive). The system memory can be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory can store some or all instructions and data required by the processor at run time.

[0101] In addition, the memory 501 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (e.g., DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and magnetic disks and / or optical disks may also be used. In some embodiments, the memory 501 may include a removable storage device that can be read and / or written, such as a laser disc (CD), a read-only digital versatile disk (e.g., DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disk, an ultra-density optical disk, a flash memory card (e.g., SD card, min SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and instantaneous electronic signals transmitted wirelessly or wired.

[0102] The memory 501 stores executable codes, and when the executable codes are processed by the processor 502 , the processor 502 can execute part or all of the methods described above.

[0103] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.

[0104] Alternatively, the present application can also be implemented as a computer-readable storage medium (or non-transitory machine-readable storage medium or machine-readable storage medium) on which executable code (or computer program or computer instruction code) is stored. When the executable code (or computer program or computer instruction code) is executed by a processor of an electronic device (or server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present application.

[0105] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An interactive method for bank users, characterized in that: include: Get user session text; Identify whether the user's conversation text contains sensitive information; If not, the conversation text is sent to the confusion model to generate corresponding reply content; Determine whether the reply content is compliant, and if so, feedback the corresponding reply content on the banking business interface.

2. The interactive method according to claim 1, characterized in that: The obtaining of the user conversation text includes: A session service selected by a user on a banking service interface is received, and a session in the session service is encrypted.

3. The interactive method according to claim 1, characterized in that: Before the identifying whether the user's conversation text contains sensitive information, the following step is also included: Determining whether the conversation text contains personal privacy information; If personal privacy information is included, the personal privacy content in the conversation text is replaced to obtain the conversation text after the privacy content is replaced.

4. The interactive method according to claim 2, characterized in that: include: According to the conversation service selected by the user, a language database corresponding to the reply content is generated.

5. The interactive method according to claim 4, characterized in that: The sending of the conversation text to the confusion model to generate corresponding reply content includes: Performing semantic analysis on the conversation text, classifying the conversation text into tasks according to the semantic analysis result, and determining the task type corresponding to the conversation text; According to the task type and preset priority rules, the conversation text is input into the corresponding confusion model to generate the corresponding reply content.

6. The interactive method according to claim 5, characterized in that: The step of inputting the conversation text into the corresponding confusion model to generate the corresponding reply content includes: Inputting the conversation text into the corresponding obfuscation model to generate pre-reply content; The reply content is generated using the language model according to the conversation service selected by the user and the pre-reply content.

7. The interactive method according to claim 2, characterized in that: The determination of whether the reply content is compliant includes: According to the conversation service selected by the user in the banking service interface, it is determined whether the reply content exceeds the service scope of the conversation service.

8. The interactive method according to claim 1, characterized in that: The obtaining of the user conversation text comprises: Obtain the user's voice conversation data; Perform speech recognition on the speech conversation data to obtain a conversation text.

9. A user conversation interaction system for banking products, characterized in that include: The acquisition module is used to obtain the user session text; A generating module, used for identifying whether the conversation text of the user contains sensitive information; If not, the conversation text is sent to the confusion model to generate corresponding reply content; The compliance judgment module is used to judge whether the reply content is compliant, and if so, to feedback the corresponding reply content on the banking business interface.

10. A computer-readable storage medium having executable codes stored thereon, which, when executed by a processor of an electronic device, causes the processor to execute the interaction method according to claims 1 to 8.

Images