Information interaction method and device, computer equipment and storage medium

Abstract

The invention relates to an information interaction method and device, computer equipment and a storage medium. The method comprises the following steps: in response to acquired user input information, analyzing the user input information by using a language model, and acquiring a user intention and a target operation type, the target operation type comprising an attendance checking operation, a reimbursement operation and an ask-for-leave operation; determining a target operation instruction and a jump path of the target operation system based on the user intention and the target operation type; and skipping to the target operating system based on the skipping path, and executing the target operation on the target operating system based on the target operation instruction. By adopting the method, the system information interaction in various application scenes can be realized, and the efficiency and reliability of the system information interaction are improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an information interaction method, apparatus, computer equipment, and storage medium. Background Art

[0002] Most existing language models are designed for question-answering and retrieval. Requirement for them to perform specific actions, such as redirecting to another software system, is difficult. Firstly, language models are merely linguistic content and lack the ability to redirect. Secondly, the language model's non-standard content output can prevent action interceptors from accurately identifying the content.

[0003] In related technologies, the user's questions are input into a language model, and then the operating system is controlled to perform related operations based on the operation instructions generated by the language model. However, it is easy for the operation instructions generated based on the language model to not match the corresponding execution instructions of the operating system. Summary of the Invention

[0004] Based on this, it is necessary to provide an efficient and accurate information interaction method, device, computer equipment and storage medium to address the above technical problems.

[0005] In order to solve the above technical problems, in a first aspect, an information interaction method is provided, the method comprising:

[0006] In response to obtaining user input information, the user input information is parsed using a language model to obtain the user's intention and the target operation type, which includes attendance operation, reimbursement operation, and leave operation;

[0007] Determine the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type;

[0008] Jump to the target operating system based on the jump path, and execute the target operation on the target operating system based on the target operation instruction.

[0009] In one embodiment, parsing user input information using a language model to obtain user intent and target operation type includes:

[0010] Determining invalid information in the user input information, deleting the invalid information from the user input information, and obtaining the user input information to be processed, wherein the invalid information is information that is irrelevant to the user's intention;

[0011] Perform correlation feature analysis on the user input information to be processed to obtain user input correlation information;

[0012] Determine multiple user intention prompt information based on the user input information to be processed and the user input related information;

[0013] Retrieve the question that matches the user input information to be processed from the memory bank, input the question and user intention prompt information into the language model, obtain the answer information output by the language model, and use the answer information as the user intention;

[0014] Gets the target action type that matches the user intent.

[0015] In one embodiment, determining the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type further includes:

[0016] Get the user intent attribute type, which includes mutable user intent and immutable user intent;

[0017] In response to the user intention attribute being an immutable user intention, selecting an operation instruction that matches the user intention from an operation instruction library as a target operation instruction, and obtaining a jump path of a target operating system corresponding to the target operation type;

[0018] In response to the user intent attribute being a variable user intent, a target user intent matching the user intent and a jump intention having an operation link with the target user intent are determined from the intent knowledge graph, and operation instructions matching the target user intent and the jump intention are selected from the operation instruction library as target operation instructions and jump operation instructions respectively, and the jump operation type corresponding to the jump operation instruction is obtained, and the jump path of the target operating system and the jump path of the jump operating system corresponding to the target operation type and the jump operation type are obtained respectively, wherein the intent knowledge graph includes multiple user intentions, and different user intentions construct a network adjacency relationship through operation links, and the operation links represent the jump paths between different user intentions under different tasks.

[0019] In one embodiment, the target operation type is an attendance operation, and executing the target operation on the target operating system based on the target operation instruction includes:

[0020] In response to the target operation type being an attendance operation, acquiring a target operation instruction type, where the target operation instruction type includes a first attendance operation instruction and a second attendance operation instruction;

[0021] In response to the target operation instruction type being the first attendance operation instruction, obtaining a path to a browsing page of the first attendance operation instruction, and jumping to the browsing page of the first attendance operation instruction to perform the first attendance operation;

[0022] In response to the target operation instruction type being the second attendance operation instruction, the target attendance time in the user input information and the path of the browsing page of the second attendance operation instruction are obtained, and the browsing page of the second attendance operation instruction is jumped to, and the second attendance operation is performed based on the target attendance time.

[0023] In one embodiment, the target operating system type is a reimbursement operation, and executing the target operation on the target operating system based on the target operation instruction includes:

[0024] In response to the target operation type being a reimbursement operation, acquiring a target operation instruction type, where the target operation instruction type includes a first reimbursement operation instruction and a second reimbursement operation instruction;

[0025] In response to the target operation instruction type being the first reimbursement operation instruction, obtaining a path to a browsing page of the first reimbursement operation instruction, and jumping to the browsing page of the first reimbursement operation instruction to perform the first reimbursement operation;

[0026] In response to the target operation instruction type being the second reimbursement operation instruction, the reimbursement reason, reimbursement value, reimbursement identification number and the path of the browsing page of the second reimbursement operation instruction in the user input information are obtained, and the browsing page of the second reimbursement operation instruction is jumped to, and the second reimbursement operation is performed based on the reimbursement reason, reimbursement value and reimbursement identification number.

[0027] In one embodiment, the target operating system type is a leave operation, and executing the target operation on the target operating system based on the target operation instruction includes:

[0028] In response to the target operation type being a leave operation, acquiring a target operation instruction type, where the target operation instruction type includes a first leave operation instruction and a second leave operation instruction;

[0029] In response to the target operation instruction type being the first leave operation instruction, obtaining a path to a browsing page of the first leave operation instruction, and jumping to the browsing page of the first leave operation instruction to execute the first leave operation;

[0030] In response to the target operation instruction type being the second leave operation instruction, the first time, second time and the path of the browsing page of the second leave operation instruction in the user input information are obtained, and the browsing page of the second leave operation instruction is jumped to, and the second reimbursement operation is performed based on the first time and the second time.

[0031] In one embodiment, obtaining user input information includes:

[0032] Obtaining a type of multimodal information input by a user, where the multimodal information type includes at least one of text information, image information, and video information;

[0033] In response to the type of the multimodal information including text information, using the text information as user input information corresponding to the text information;

[0034] In response to the type of multimodal information including picture information, parsing the picture to obtain picture text information and picture graphic information, performing semantic analysis on the picture graphic information to obtain a picture graphic semantic representation, and integrating the picture text information and the picture graphic semantic representation to obtain user input information corresponding to the picture information;

[0035] In response to the fact that the type of multimodal information includes video information, voice recognition is performed on the video information to obtain video voice text information, and the video is sliced ​​to obtain multiple video picture groups. Semantic analysis is performed on the multiple video picture groups in the order in which each video picture group is played in the video to obtain a graphic semantic representation of the video picture. Text parsing is performed on each video picture to obtain video picture text information. The video voice text information, video picture text information and the graphic semantic representation of the video picture are integrated to obtain user input information corresponding to the video information.

[0036] In order to solve the above technical problems, in a second aspect, an information interaction device is provided, comprising:

[0037] An acquisition module is used to respond to user input information and parse the user input information using a language model to obtain user intent and target operation type, including attendance operation, reimbursement operation, and leave operation;

[0038] A determination module, configured to determine a target operation instruction and a jump path of a target operating system based on a user's intention and a target operation type;

[0039] The operation module is used to jump to the target operating system based on the jump path and perform a target operation on the target operating system based on the target operation instruction.

[0040] In order to solve the above technical problems, in the third aspect, a computer device is provided, including a memory, a processor and a computer program stored in the memory and runnable on the processor, and the processor implements the following steps when executing the computer program: the processor implements the steps of the above-mentioned first aspect method when executing the computer program.

[0041] In order to solve the above technical problems, in a fourth aspect, the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method of the above first aspect are implemented.

[0042] Different from related technologies, this application uses a language model to parse user input in response to user input, obtaining the user's intent and target operation type, which includes attendance, reimbursement, and leave. Based on the user's intent and target operation type, the application determines the target operation instruction and the target operating system's jump path. The application then jumps to the target operating system based on the jump path and executes the target operation on the target operating system based on the target operation instruction. This method enables operating system information exchange in various application scenarios, such as attendance, reimbursement, and leave, and improves the efficiency and reliability of information exchange between the operating system and the language model. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 1 is a flow chart of an information interaction method in one embodiment;

[0044] Figure 2 is a flow chart of an information interaction method in another embodiment;

[0045] Figure 3 is a flowchart of an information interaction method in yet another embodiment;

[0046] Figure 4 is a flowchart of an information interaction method in yet another embodiment;

[0047] Figure 5 is a structural block diagram of an information interaction device in one embodiment;

[0048] Figure 6 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0050] In one embodiment, Figure 1 As shown, the present application provides an information interaction method, which specifically includes the following steps:

[0051] Step 101: In response to obtaining user input information, the user input information is parsed using a language model to obtain user intention and target operation type, where the target operation type includes attendance operation, reimbursement operation, and leave operation.

[0052] Specifically, a language model is used to parse the user input information to obtain the user intention and the target operation type, including: determining the invalid information in the user input information, deleting the invalid information from the user input information to obtain the user input information to be processed, where the invalid information is information unrelated to the user intention; performing an associated feature analysis on the user input information to be processed to obtain the user input associated information; determining multiple user intention prompt messages based on the user input information to be processed and the user input associated information; retrieving a question matching the user input information to be processed from the memory bank, inputting the question and the user intention prompt messages into the language model to obtain the answer information output by the language model, and using the answer information as the user intention; and obtaining the target operation type matching the user intention.

[0053] The user input information is the text information corresponding to a certain business or a certain operation. After obtaining the user input information, word segmentation processing can be performed on the user input information. Specifically, the Jieba word segmentation technology can be used to perform word segmentation processing on the user input information. Suppose the user input information is "query the attendance on the morning of February 1st". The result of performing word segmentation processing on this text using the Jieba word segmentation technology is "query / February / 1st / morning / of / attendance / perform / query". The Jieba word segmentation technology can calculate the probability of each word segmentation result, that is, the rationality of each word segmentation result, and using the Jieba word segmentation technology to perform word segmentation processing on the user input can improve the accuracy of word segmentation.

[0054] Optionally, word segmentation processing can also be performed on the user input information through a language processing tool, and the language processing tool is not limited to Jieba, snowNLP, thulac, etc.

[0055] According to the word segmentation processing result, select the invalid information in the user input information, such as deleting the modal particle "of" and deleting words such as adjectives that have no impact on extracting the user intention from the word segmentation processing result to obtain the user input information to be processed.

[0056] Among them, the correlation feature analysis of the user input information to be processed includes obtaining a user input database and a user intention database, wherein the user input database stores multiple historical user input information, and the user intention database stores multiple historical user intentions. A corresponding relationship between the user input database and the user intention database is established. The recurrent neural network layer in the analysis model can be used to analyze the user input database and the user intention database, and the intrinsic correlation between multiple different historical user input information corresponding to the same historical user intention can be mined. The features of all historical user input information corresponding to the same historical user intention can be clustered together, and the activation function in the fully connected layer in the analysis model can be used to associate the features of historical user input information with dependencies under the same historical user intention, so as to obtain a historical user input correlation feature set corresponding to each historical user intention. The similarity between the user input features to be processed and the historical user input information in the user input database can be calculated, and the historical user input correlation feature set containing historical user input information with a similarity greater than a threshold is used as user input correlation information.

[0057] Keywords are extracted from the user input information to be processed and the user input-related information respectively to obtain the initial user intention corresponding to the user input information to be processed and the initial associated user intention corresponding to the user input-related information, and the initial user intention and the initial associated user intention are integrated to obtain user intention prompt information.

[0058] Search the memory bank corresponding to the language model for questions that match the user input information to be processed, wherein the memory bank stores historical questions and answers and historical conversation processes processed by the language model. Input the question and the user's question intention prompt information into the embedding model. The embedding model converts the question and the user's intention prompt information into vector representations to obtain the question vector and the user's intention prompt vector. Based on the question vector and the user's intention prompt vector, the most relevant content to the question is screened out. The screened out most relevant content to the question is combined with the question to obtain the query retrieval content, which is also the prompt word project. The query retrieval content is input into the language model. The language model generates an answer and outputs the answer. The answer is used as the user intention corresponding to the user input information to be processed. After obtaining the user intention, the target operation and target operation type corresponding to the user intention can be obtained from the operation instruction library. The operation instruction library can be the execution instruction of the operating system. For example, if the user intention is to query attendance, the target operation is to query attendance and the target operation type is attendance.

[0059] In this application, by focusing on the expression of different user input information corresponding to the same user intention, determining the user intention prompt words based on the user input related information, and using the user intention prompt words as the knowledge text of the language model dialogue, and using the answer output by the language model as the user intention, the accuracy of user intention acquisition can be improved.

[0060] In one embodiment, before obtaining user input information, the method includes obtaining the type of multimodal information input by the user, where the multimodal information type includes at least one of text information, image information, and video information. The present application supports converting the multimodal information input by the user into user input information, and determining the user intent based on the user input information and a language model.

[0061] In response to the type of the multimodal information including text information, the text information is used as user input information corresponding to the text information.

[0062] When the multimodal information includes text information, the multimodal information input by the user conforms to the format of the user input information, that is, the text format, and the text information is directly used as the user input information.

[0063] In response to the type of multimodal information including picture information, the picture is parsed to obtain picture text information and picture graphic information, semantic analysis is performed on the picture graphic information to obtain picture graphic semantic representation, and the picture text information and picture graphic semantic representation are integrated to obtain user input information corresponding to the picture information.

[0064] When the multimodal information input by the user includes image information, it is necessary to extract the text information from the image information. On the one hand, it is necessary to obtain the text information of the image information, and on the other hand, it is necessary to obtain the semantic representation of the graphic information of the image. In other words, it is necessary to analyze the graphics in the image to obtain the information expressed by the graphics. The graphics can be converted into natural language descriptions through image description generation technology to obtain the semantic representation of the image graphics. The image text information and the image graphic semantic representation are integrated to obtain the user input information corresponding to the image information.

[0065] In response to the fact that the type of multimodal information includes video information, voice recognition is performed on the video information to obtain video voice text information, and the video is sliced ​​to obtain multiple video picture groups. Semantic analysis is performed on the multiple video picture groups in the order in which each video picture group is played in the video to obtain a graphic semantic representation of the video picture. Text parsing is performed on each video picture to obtain video picture text information. The video voice text information, video picture text information and the graphic semantic representation of the video picture are integrated to obtain user input information corresponding to the video information.

[0066] Similarly, when the multimodal information input by the user includes video information, it is necessary to extract the text information in the video information. On the one hand, it is necessary to obtain the video voice text information, and on the other hand, it is necessary to obtain the graphic semantic information of each frame of the video and the text information of each frame of the video (video picture text information), integrate the video voice text information, the video picture text information and the graphic semantic representation of the video picture to obtain the user input information corresponding to the video information. For example, the video can be voice-recognized by voice-to-text technology to obtain video voice text information, and each frame of the video picture in multiple video picture groups can be converted into a natural language description by image description generation technology, and then the natural language description of each frame of the video picture is spliced ​​and integrated according to the playback order of each frame of the video picture in each video picture group to obtain a graphic semantic representation of the video picture.

[0067] Step 102 : determining a target operation instruction and a jump path of a target operating system based on the user intention and the target operation type.

[0068] This application also considers whether the user intention is a variable user intention, that is, whether the user intention is a step in a certain operation or business. If it is a variable user intention, the jump intention and jump path corresponding to the next or next steps of the operation can be automatically recommended.

[0069] Specifically, the user intention attribute type is obtained, and the user intention attribute type includes a variable user intention and an immutable user intention; in response to the user intention attribute being an immutable user intention, an operation instruction matching the user intention is selected from the operation instruction library as the target operation instruction, and the jump path of the target operating system corresponding to the target operation type is obtained.

[0070] In response to the user intent attribute being a variable user intent, a target user intent matching the user intent and a jump intent having an operation link with the target user intent are determined from the intent knowledge graph. Operation instructions matching the target user intent and the jump intent are selected from the operation instruction library as target operation instructions and jump operation instructions, respectively. The jump operation type corresponding to the jump operation instruction is obtained, and the jump path of the target operating system and the jump path of the jump operating system corresponding to the target operation type and the jump operation type are obtained, respectively. The intent knowledge graph includes multiple user intents, and different user intents are connected in a network through operation links. The operation links represent the jump paths between different user intents under different tasks.

[0071] After determining that the user intent type is variable, the target user intent that matches the user intent can be determined from the intent knowledge graph. The user intent with an operational link to the target user intent is used as the jump intention. The jump user intent here is the user intention corresponding to the next one or several operations of the target user intent under one or several operations. At the same time, the target user intent, the target operation instruction corresponding to the jump intention, and the jump operation instruction are obtained. In this way, by focusing on the behavioral jump path knowledge network data between different operation intentions to recommend jump user intentions and jump paths, it is possible to more comprehensively predict and recommend operation instructions and operating system jump paths.

[0072] Step 103 : Jump to the target operating system based on the jump path, and execute the target operation on the target operating system based on the target operation instruction.

[0073] In one embodiment, the target operation type is an attendance operation, and executing the target operation on the target operating system based on the target operation instruction includes: in response to the target operation type being an attendance operation, obtaining the target operation instruction type, the target operation instruction type including a first attendance operation instruction and a second attendance operation instruction; in response to the target operation instruction type being the first attendance operation instruction, obtaining the path of the browsing page of the first attendance operation instruction, and jumping to the browsing page of the first attendance operation instruction to execute the first attendance operation; in response to the target operation instruction type being the second attendance operation instruction, obtaining the target attendance time in the user input information and the path of the browsing page of the second attendance operation instruction, jumping to the browsing page of the second attendance operation instruction, and executing the second attendance operation based on the target attendance time.

[0074] The first attendance operation instruction may be an attendance query operation, specifically, whether a clock-in is punched in at a certain time, the clock-in time at a certain time, attendance status within a period of time, etc. In the case of the attendance query operation, a browsing page of the attendance operation system for the attendance query operation is obtained, and the specific attendance query operation is directly performed on the browsing page of the attendance operation system.

[0075] The second attendance operation instruction can be to parse the punch-in time. The corresponding user intention is to forget to apply for punch-in. In the case of the punch-in time parsing operation, the user-input forgotten punch-in time (target attendance time) can be extracted from the user input information. After the target attendance time is successfully extracted, the specific punch-in time parsing operation is performed on the browsing page of the attendance operating system.

[0076] The target operating system type is a reimbursement operation, and executing the target operation on the target operating system based on the target operation instruction includes: in response to the target operation type being a reimbursement operation, obtaining the target operation instruction type, the target operation instruction type including a first reimbursement operation instruction and a second reimbursement operation instruction; in response to the target operation instruction type being the first reimbursement operation instruction, obtaining the path of the browsing page of the first reimbursement operation instruction, and jumping to the browsing page of the first reimbursement operation instruction to execute the first reimbursement operation; in response to the target operation instruction type being the second reimbursement operation instruction, obtaining the reimbursement reason, reimbursement value, reimbursement identification number in the user input information and the path of the browsing page of the second reimbursement operation instruction, and jumping to the browsing page of the second reimbursement operation instruction, and executing the second reimbursement operation based on the reimbursement reason, reimbursement value, and reimbursement identification number.

[0077] The first reimbursement operation instruction may be a reimbursement query operation, specifically, viewing reimbursement records at a certain time. In the case of the reimbursement query operation, a browsing page of the reimbursement operation system for executing the reimbursement query operation is obtained, and the specific reimbursement query operation is directly executed on the browsing page of the reimbursement system.

[0078] The second reimbursement operation instruction can be used for invoice reimbursement. In this case, the user's reimbursement reason, reimbursement amount, and reimbursement identification number can be extracted from the user's input information. The user's reimbursement reason can be the type of reimbursement, such as meals, travel, accommodation, or purchases; the reimbursement amount is the amount to be reimbursed; and the reimbursement identification number can be the invoice tax number. Once the reimbursement reason, reimbursement amount, and reimbursement identification number in the user's input information are successfully recognized, the user can be directly redirected to the browsing page for the second reimbursement operation instruction, and the second reimbursement operation can be executed based on the reimbursement reason, reimbursement amount, and reimbursement identification number.

[0079] In one embodiment, the target operating system type is a leave operation, and executing the target operation on the target operating system based on the target operation instruction includes: in response to the target operation type being a leave operation, obtaining the target operation instruction type, the target operation instruction type including a first leave operation instruction and a second leave operation instruction; in response to the target operation instruction type being the first leave operation instruction, obtaining the path of the browsing page of the first leave operation instruction, and jumping to the browsing page of the first leave operation instruction to execute the first leave operation; in response to the target operation instruction type being the second leave operation instruction, obtaining the first time, the second time and the path of the browsing page of the second leave operation instruction in the user input information, and jumping to the browsing page of the second leave operation instruction, and executing the invoice reimbursement operation based on the first time and the second time.

[0080] The first leave operation instruction may be a leave time query operation, specifically, checking the remaining leave and querying the leave record. In the case of the leave query operation, a browsing page of the leave operation system for executing the leave query operation is obtained, and the specific leave time query operation is directly executed on the browsing page of the leave system.

[0081] The second leave request instruction can be used to request a leave. During the request, the first time and second time in the user input information can be extracted from the user input information. Here, the first time can be the leave start time, and the second time can be the leave end time. Once the first and second times in the user input information are successfully recognized, the user can be redirected to the browsing page for the second leave request instruction, and the leave request can be executed based on the first and second times.

[0082] The information interaction method of the present application can be implemented based on an information interaction system, which includes a management system, a language model, and multiple operating systems (attendance system, leave system, reimbursement system). Multiple assistants (including guidance assistants, attendance assistants, leave assistants, and reimbursement assistants) can be set on the language model to establish an association between the management system and multiple assistants.

[0083] Among them, the guidance assistant receives user input information transmitted by the management system, calls the language model to parse the user input information, extracts the user intention from the user input information, searches for the target operation type corresponding to the user intention from the preset intention and instruction information table, and performs the corresponding operation according to the target operation type (for example, entering the attendance system, entering the leave system, etc.).

[0084] Assuming that the user intention analyzed by the guidance assistant is to query attendance, the guidance assistant generates a query attendance instruction and calls the attendance assistant according to the query attendance instruction. After receiving the query attendance instruction, the attendance assistant will jump the client's browsing page to the page corresponding to the query attendance.

[0085] Assuming the guidance assistant interprets the user's intent as requesting leave, it generates a leave request instruction and calls the leave assistant based on the leave instruction. Upon receiving the leave instruction, the leave assistant determines whether it can obtain the leave start and end times. If so, it directly outputs the leave instruction and calls the leave request interface on the management system to submit the leave. If the leave start and end times are not recognized, an alarm is issued to the user.

[0086] Assuming that the user intention analyzed by the guidance assistant is invoice reimbursement, the guidance assistant will generate an invoice reimbursement instruction and call the reimbursement assistant according to the reimbursement instruction. After the reimbursement assistant receives the invoice reimbursement instruction, it will jump the client's browsing page to the page corresponding to invoice reimbursement.

[0087] The protocol definition of the assistant is set in this application, for example:

[0088] 1. red irect: / / agent=$name red irect means to jump to the assistant on another model. 2. agent=$name, where $name is the name of the assistant on a specific model.

[0089] 3. promote: / / key=$name promote means setting the prompt word of the model, and searching the corresponding prompt word from the storage object according to the name; call: / / ap i=$path¶m1=$param1¶m2=$param2……paramN=$paramN call means calling the interface.

[0090] 4. path is the specific path of the called interface, param1, param2...paramN are specific parameters.

[0091] 5. jump: / / url=xxx¶m1=$param1…paramN=$paramN jump means jumping to a specific web page. Param1, param2…paramN are specific parameters, which are added according to the specific scenario.

[0092] In one embodiment, see Figure 2 as well as Figure 3 In this application, a session interceptor is set up. The output of all assistants on the model passes through the interceptor. The interceptor performs different operations according to the specific protocol. When a change in user intention is detected, the assistant switching operation can be executed. For example, assuming that assistant A needs to switch to assistant B, assistant A needs to exit the conversation first. After switching to the entry assistant, the output interceptor determines that the output of the model assistant complies with the protocol and switches to assistant B to perform the operation.

[0093] Specifically, an intent escape assistant is added after each specific operation assistant output. The intent escape assistant is used to detect the user's dialogue and determine whether the user's intention has changed.

[0094] The prompt can be set as: Based on the user input and context, please determine whether the conversation is related to operation X. If not, change the output to redirect: / / agent=ma in.

[0095] Operation X corresponds to requesting leave, taking an exam, or reimbursing a fee. These are set to the corresponding values ​​in the corresponding assistants. agent = main, indicating that the model being switched is the entry model.

[0096] During the conversation with a specific operation assistant, a forced exit mechanism can be set to increase a certain degree of fault tolerance. The specific mechanism is that when no agreement is detected in multiple consecutive rounds of conversations, the current switching entry assistant will be directly exited for conversation.

[0097] The entry assistant determines the target action type. Each target action type has a corresponding prompt. This prompt is used to determine whether the user requires attendance, leave, or reimbursement during the conversation. If attendance is required, the agent outputs redirect: / / agent=attendance; if leave is required, the agent outputs redirect: / / agent=leave; and if reimbursement is required, the agent outputs redirect: / / agent=reimbursement.

[0098] The Attendance Assistant determines the attendance operation type corresponding to the attendance operation. Each attendance operation type has a corresponding prompt word information. For example, in the conversation with the user, please determine whether the user wants to query attendance or apply for a forgotten punch-in. If the user wants to query attendance, directly input jump: / / url=xxx. If the user wants to apply for a forgotten punch-in, directly input promote: / / to resolve the forgotten punch-in time.

[0099] It also includes operation prompt information. For example, when the intention is to apply for forgetting to punch in, you can remind the user that the time when the user forgot to punch in needs to be extracted from the user's conversation and recorded as A. After the time is successfully extracted, call: / / api=$path&time=A is directly output to execute the analysis of the forgotten punch-in time.

[0100] The Leave Request Assistant determines the corresponding leave type for a leave request. Each leave type has a corresponding prompt word indicating the user's intent. For example, during a conversation with the user, determine whether the user wants to check remaining leave, check leave history, or request leave. To check remaining leave, enter jump: / / url=xxx; to check leave history, enter jump: / / url=xxx; to request leave, enter promote: / / key=parsed leave time.

[0101] It also includes operation prompt information. For example, when the intention is to ask for leave, it can remind you that the user's leave start time A and end time B need to be extracted from the user's conversation. After successful parsing, call: / / ap i=$path&start=A&end=B is directly output to execute the parsing of the leave time.

[0102] The Reimbursement Assistant determines the appropriate reimbursement action type. Each reimbursement action type has a corresponding prompt. For example, during a conversation with a user, determine whether the user wants to view reimbursement records or submit an invoice. To view reimbursement records, simply enter jump: / / url=xxx. To submit an invoice, simply enter promote: / / key=upload invoice.

[0103] It also includes operation prompt information. For example, when the intention is to reimburse an invoice, it can remind the user that the reason for reimbursement needs to be extracted from the user's conversation, recorded as A, and the user is prompted to upload the invoice, and the amount in the invoice is identified as B, and the invoice tax number is recorded as C. After the A, B, and C parameters are successfully extracted, call: / / api=path&value=B&id=C is directly output to execute the invoice upload operation.

[0104] In a specific embodiment, Figure 4 As shown, the process of the information interaction method can be as follows:

[0105] S1: The management system obtains user input information and transmits the user input information to the guidance assistant.

[0106] S2: The guidance assistant receives user input information, calls the language model to parse the user input information, and obtains user intent data and the operation type corresponding to the user intent.

[0107] S3: Determine whether the operation type corresponding to the user intent obtained by the analysis is the first type (here the first type includes entering the attendance system, entering the leave system, and entering the reimbursement system);

[0108] S4: If the operation type obtained by parsing is the first type, the guidance assistant searches for a corresponding target instruction according to the operation type, and enters a corresponding system according to the instruction.

[0109] S5: If the service type obtained by parsing is not the first type, it is considered that the guidance assistant cannot independently complete the service of this type, and it is determined whether it needs to be transferred to other assistants for processing.

[0110] S6: If necessary, obtain the types of other assistants (attendance assistant, leave assistant, reimbursement assistant), and use the operation assistant corresponding to the type of other assistants that matches the operation type as the target operation assistant.

[0111] S7: The guidance assistant generates an operation type instruction, the operation type instruction carries the operation data, and the guidance assistant sends the operation type instruction to the target operation assistant.

[0112] S8: In response to the target operation assistant (at least one of the attendance assistant, leave assistant, and reimbursement assistant) receiving the operation type instruction, the target operation assistant parses the operation type instruction and obtains the target operation data. The target operation assistant calls the corresponding interface of the management system based on the target operation data, jumps to the target interface, and executes the target operation.

[0113] It should be understood that although Figures 1-4 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figures 1-4 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0114] In one embodiment, Figure 5 As shown, an information interaction device is provided, including: an acquisition module 20, a determination module 21 and an operation module 22, wherein:

[0115] The acquisition module 20 is configured to, in response to obtaining user input information, parse the user input information using a language model to obtain the user's intention and the target operation type, where the target operation type includes attendance operation, reimbursement operation, and leave operation;

[0116] A determination module 21 is configured to determine a target operation instruction and a jump path of a target operating system based on a user intention and a target operation type;

[0117] The operation module 22 is configured to jump to the target operating system based on the jump path and execute a target operation on the target operating system based on the target operation instruction.

[0118] In one embodiment, another implementation of the information interaction method that can be implemented by the above device includes the following specific steps:

[0119] Use language models to parse user input information to obtain user intent and target operation types, including:

[0120] Determining invalid information in the user input information, deleting the invalid information from the user input information, and obtaining the user input information to be processed, wherein the invalid information is information that is irrelevant to the user's intention;

[0121] Perform correlation feature analysis on the user input information to be processed to obtain user input correlation information;

[0122] Determine multiple user intention prompt information based on the user input information to be processed and the user input related information;

[0123] Retrieve the question that matches the user input information to be processed from the memory bank, input the question and user intention prompt information into the language model, obtain the answer information output by the language model, and use the answer information as the user intention;

[0124] Gets the target action type that matches the user intent.

[0125] In one embodiment, determining the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type further includes:

[0126] Get the user intent attribute type, which includes mutable user intent and immutable user intent;

[0127] In response to the user intention attribute being an immutable user intention, selecting an operation instruction that matches the user intention from an operation instruction library as a target operation instruction, and obtaining a jump path of a target operating system corresponding to the target operation type;

[0128] In response to the user intent attribute being a variable user intent, a target user intent matching the user intent and a jump intention having an operation link with the target user intent are determined from the intent knowledge graph, and operation instructions matching the target user intent and the jump intention are selected from the operation instruction library as target operation instructions and jump operation instructions respectively, and the jump operation type corresponding to the jump operation instruction is obtained, and the jump path of the target operating system and the jump path of the jump operating system corresponding to the target operation type and the jump operation type are obtained respectively, wherein the intent knowledge graph includes multiple user intentions, and different user intentions construct a network adjacency relationship through operation links, and the operation links represent the jump paths between different user intentions under different tasks.

[0129] In one embodiment, the target operation type is an attendance operation, and executing the target operation on the target operating system based on the target operation instruction includes: in response to the target operation type being an attendance operation, obtaining the target operation instruction type, the target operation instruction type including a first attendance operation instruction and a second attendance operation instruction; in response to the target operation instruction type being the first attendance operation instruction, obtaining the path of the browsing page of the first attendance operation instruction, and jumping to the browsing page of the first attendance operation instruction to execute the first attendance operation; in response to the target operation instruction type being the second attendance operation instruction, obtaining the target attendance time in the user input information and the path of the browsing page of the second attendance operation instruction, jumping to the browsing page of the second attendance operation instruction, and executing the second attendance operation based on the target attendance time.

[0130] In one embodiment, the target operating system type is a reimbursement operation, and executing the target operation on the target operating system based on the target operation instruction includes: in response to the target operation type being a reimbursement operation, obtaining the target operation instruction type, the target operation instruction type including a first reimbursement operation instruction and a second reimbursement operation instruction; in response to the target operation instruction type being the first reimbursement operation instruction, obtaining the path of the browsing page of the first reimbursement operation instruction, and jumping to the browsing page of the first reimbursement operation instruction to execute the first reimbursement operation; in response to the target operation instruction type being the second reimbursement operation instruction, obtaining the reimbursement reason, reimbursement value, reimbursement identification number in the user input information and the path of the browsing page of the second reimbursement operation instruction, and jumping to the browsing page of the second reimbursement operation instruction, and executing the second reimbursement operation based on the reimbursement reason, reimbursement value, and reimbursement identification number.

[0131] In one embodiment, the target operating system type is a leave operation, and executing the target operation on the target operating system based on the target operation instruction includes: in response to the target operation type being the leave operation, obtaining the target operation instruction type, the target operation instruction type including a first leave operation instruction and a second leave operation instruction;

[0132] In response to the target operation instruction type being the first leave operation instruction, obtaining a path to a browsing page of the first leave operation instruction, and jumping to the browsing page of the first leave operation instruction to execute the first leave operation;

[0133] In response to the target operation instruction type being the second leave operation instruction, the first time, second time and the path of the browsing page of the second leave operation instruction in the user input information are obtained, and the browsing page of the second leave operation instruction is jumped to, and the second reimbursement operation is performed based on the first time and the second time.

[0134] In one embodiment, obtaining user input information includes:

[0135] Obtaining a type of multimodal information input by a user, where the multimodal information type includes at least one of text information, image information, and video information;

[0136] In response to the type of the multimodal information including text information, using the text information as user input information corresponding to the text information;

[0137] In response to the type of multimodal information including picture information, parsing the picture to obtain picture text information and picture graphic information, performing semantic analysis on the picture graphic information to obtain a picture graphic semantic representation, and integrating the picture text information and the picture graphic semantic representation to obtain user input information corresponding to the picture information;

[0138] In response to the fact that the type of multimodal information includes video information, voice recognition is performed on the video information to obtain video voice text information, and the video is sliced ​​to obtain multiple video picture groups. Semantic analysis is performed on the multiple video picture groups in the order in which each video picture group is played in the video to obtain a graphic semantic representation of the video picture. Text parsing is performed on each video picture to obtain video picture text information. The video voice text information, video picture text information and the graphic semantic representation of the video picture are integrated to obtain user input information corresponding to the video information.

[0139] For specific definitions of the information interaction device, please refer to the definitions of the information interaction method above and will not be repeated here. Each module in the above-mentioned information interaction device can be implemented in whole or in part through software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each of the above modules.

[0140] In one embodiment, the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the information interaction methods provided by the above methods.

[0141] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. 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 computer program in the non-volatile storage medium. The database of the computer device is used to store data used in the information interaction method. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, an information interaction method is implemented.

[0142] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0143] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are performed:

[0144] In response to obtaining user input information, the user input information is parsed using a language model to obtain the user's intention and the target operation type, which includes attendance operation, reimbursement operation, and leave operation;

[0145] Determine the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type;

[0146] Jump to the target operating system based on the jump path, and execute the target operation on the target operating system based on the target operation instruction.

[0147] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0148] Use language models to parse user input information to obtain user intent and target operation types, including:

[0149] Determining invalid information in the user input information, deleting the invalid information from the user input information, and obtaining the user input information to be processed, wherein the invalid information is information that is irrelevant to the user's intention;

[0150] Perform correlation feature analysis on the user input information to be processed to obtain user input correlation information;

[0151] Determine multiple user intention prompt information based on the user input information to be processed and the user input related information;

[0152] Retrieve the question that matches the user input information to be processed from the memory bank, input the question and user intention prompt information into the language model, obtain the answer information output by the language model, and use the answer information as the user intention;

[0153] Gets the target action type that matches the user intent.

[0154] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0155] Determining the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type also includes:

[0156] Get the user intent attribute type, which includes mutable user intent and immutable user intent;

[0157] In response to the user intention attribute being an immutable user intention, selecting an operation instruction that matches the user intention from an operation instruction library as a target operation instruction, and obtaining a jump path of a target operating system corresponding to the target operation type;

[0158] In response to the user intent attribute being a variable user intent, a target user intent matching the user intent and a jump intention having an operation link with the target user intent are determined from the intent knowledge graph, and operation instructions matching the target user intent and the jump intention are selected from the operation instruction library as target operation instructions and jump operation instructions respectively, and the jump operation type corresponding to the jump operation instruction is obtained, and the jump path of the target operating system and the jump path of the jump operating system corresponding to the target operation type and the jump operation type are obtained respectively, wherein the intent knowledge graph includes multiple user intentions, and different user intentions construct a network adjacency relationship through operation links, and the operation links represent the jump paths between different user intentions under different tasks.

[0159] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0160] The target operation type is attendance operation. The target operations performed on the target operating system based on the target operation instructions include:

[0161] In response to the target operation type being an attendance operation, acquiring a target operation instruction type, where the target operation instruction type includes a first attendance operation instruction and a second attendance operation instruction;

[0162] In response to the target operation instruction type being the first attendance operation instruction, obtaining a path to a browsing page of the first attendance operation instruction, and jumping to the browsing page of the first attendance operation instruction to perform the first attendance operation;

[0163] In response to the target operation instruction type being the second attendance operation instruction, the target attendance time in the user input information and the path of the browsing page of the second attendance operation instruction are obtained, and the browsing page of the second attendance operation instruction is jumped to, and the second attendance operation is performed based on the target attendance time.

[0164] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0165] The target operating system type is a reimbursement operation. The target operation performed on the target operating system based on the target operation instruction includes:

[0166] In response to the target operation type being a reimbursement operation, acquiring a target operation instruction type, where the target operation instruction type includes a first reimbursement operation instruction and a second reimbursement operation instruction;

[0167] In response to the target operation instruction type being the first reimbursement operation instruction, obtaining a path to a browsing page of the first reimbursement operation instruction, and jumping to the browsing page of the first reimbursement operation instruction to perform the first reimbursement operation;

[0168] In response to the target operation instruction type being the second reimbursement operation instruction, the reimbursement reason, reimbursement value, reimbursement identification number and the path of the browsing page of the second reimbursement operation instruction in the user input information are obtained, and the browsing page of the second reimbursement operation instruction is jumped to, and the second reimbursement operation is performed based on the reimbursement reason, reimbursement value and reimbursement identification number.

[0169] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0170] The target operating system type is a leave operation. The target operations performed on the target operating system based on the target operation instruction include:

[0171] In response to the target operation type being a leave operation, acquiring a target operation instruction type, where the target operation instruction type includes a first leave operation instruction and a second leave operation instruction;

[0172] In response to the target operation instruction type being the first leave operation instruction, obtaining a path to a browsing page of the first leave operation instruction, and jumping to the browsing page of the first leave operation instruction to execute the first leave operation;

[0173] In response to the target operation instruction type being the second leave operation instruction, the first time, second time and the path of the browsing page of the second leave operation instruction in the user input information are obtained, and the browsing page of the second leave operation instruction is jumped to, and the second reimbursement operation is performed based on the first time and the second time.

[0174] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0175] Before obtaining user input information, include:

[0176] Obtaining a type of multimodal information input by a user, where the multimodal information type includes at least one of text information, image information, and video information;

[0177] In response to the type of the multimodal information including text information, using the text information as user input information corresponding to the text information;

[0178] In response to the type of multimodal information including picture information, parsing the picture to obtain picture text information and picture graphic information, performing semantic analysis on the picture graphic information to obtain a picture graphic semantic representation, and integrating the picture text information and the picture graphic semantic representation to obtain user input information corresponding to the picture information;

[0179] In response to the fact that the type of multimodal information includes video information, voice recognition is performed on the video information to obtain video voice text information, and the video is sliced ​​to obtain multiple video picture groups. Semantic analysis is performed on the multiple video picture groups in the order in which each video picture group is played in the video to obtain a graphic semantic representation of the video picture. Text parsing is performed on each video picture to obtain video picture text information. The video voice text information, video picture text information and the graphic semantic representation of the video picture are integrated to obtain user input information corresponding to the video information.

[0180] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0181] In response to obtaining user input information, the user input information is parsed using a language model to obtain the user's intention and the target operation type, which includes attendance operation, reimbursement operation, and leave operation;

[0182] Determine the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type;

[0183] Jump to the target operating system based on the jump path, and execute the target operation on the target operating system based on the target operation instruction.

[0184] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0185] Use language models to parse user input information to obtain user intent and target operation types, including:

[0186] Determining invalid information in the user input information, deleting the invalid information from the user input information, and obtaining the user input information to be processed, wherein the invalid information is information that is irrelevant to the user's intention;

[0187] Perform correlation feature analysis on the user input information to be processed to obtain user input correlation information;

[0188] Determine multiple user intention prompt information based on the user input information to be processed and the user input related information;

[0189] Retrieve the question that matches the user input information to be processed from the memory bank, input the question and user intention prompt information into the language model, obtain the answer information output by the language model, and use the answer information as the user intention;

[0190] Gets the target action type that matches the user intent.

[0191] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0192] Determining the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type also includes:

[0193] Get the user intent attribute type, which includes mutable user intent and immutable user intent;

[0194] In response to the user intention attribute being an immutable user intention, selecting an operation instruction that matches the user intention from an operation instruction library as a target operation instruction, and obtaining a jump path of a target operating system corresponding to the target operation type;

[0195] In response to the user intent attribute being a variable user intent, a target user intent matching the user intent and a jump intention having an operation link with the target user intent are determined from the intent knowledge graph, and operation instructions matching the target user intent and the jump intention are selected from the operation instruction library as target operation instructions and jump operation instructions respectively, and the jump operation type corresponding to the jump operation instruction is obtained, and the jump path of the target operating system and the jump path of the jump operating system corresponding to the target operation type and the jump operation type are obtained respectively, wherein the intent knowledge graph includes multiple user intentions, and different user intentions construct a network adjacency relationship through operation links, and the operation links represent the jump paths between different user intentions under different tasks.

[0196] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0197] The target operation type is attendance operation. The target operations performed on the target operating system based on the target operation instructions include:

[0198] In response to the target operation type being an attendance operation, acquiring a target operation instruction type, where the target operation instruction type includes a first attendance operation instruction and a second attendance operation instruction;

[0199] In response to the target operation instruction type being the first attendance operation instruction, obtaining a path to a browsing page of the first attendance operation instruction, and jumping to the browsing page of the first attendance operation instruction to perform the first attendance operation;

[0200] In response to the target operation instruction type being the second attendance operation instruction, the target attendance time in the user input information and the path of the browsing page of the second attendance operation instruction are obtained, and the browsing page of the second attendance operation instruction is jumped to, and the second attendance operation is performed based on the target attendance time.

[0201] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0202] The target operating system type is a reimbursement operation. The target operation performed on the target operating system based on the target operation instruction includes:

[0203] In response to the target operation type being a reimbursement operation, acquiring a target operation instruction type, where the target operation instruction type includes a first reimbursement operation instruction and a second reimbursement operation instruction;

[0204] In response to the target operation instruction type being the first reimbursement operation instruction, obtaining a path to a browsing page of the first reimbursement operation instruction, and jumping to the browsing page of the first reimbursement operation instruction to perform the first reimbursement operation;

[0205] In response to the target operation instruction type being the second reimbursement operation instruction, the reimbursement reason, reimbursement value, reimbursement identification number and the path of the browsing page of the second reimbursement operation instruction in the user input information are obtained, and the browsing page of the second reimbursement operation instruction is jumped to, and the second reimbursement operation is performed based on the reimbursement reason, reimbursement value and reimbursement identification number.

[0206] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0207] The target operating system type is a leave operation. The target operations performed on the target operating system based on the target operation instruction include:

[0208] In response to the target operation type being a leave operation, acquiring a target operation instruction type, where the target operation instruction type includes a first leave operation instruction and a second leave operation instruction;

[0209] In response to the target operation instruction type being the first leave operation instruction, obtaining a path to a browsing page of the first leave operation instruction, and jumping to the browsing page of the first leave operation instruction to execute the first leave operation;

[0210] In response to the target operation instruction type being the second leave operation instruction, the first time, second time and the path of the browsing page of the second leave operation instruction in the user input information are obtained, and the browsing page of the second leave operation instruction is jumped to, and the second reimbursement operation is performed based on the first time and the second time.

[0211] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0212] Before obtaining user input information, include:

[0213] Obtaining a type of multimodal information input by a user, where the multimodal information type includes at least one of text information, image information, and video information;

[0214] In response to the type of the multimodal information including text information, using the text information as user input information corresponding to the text information;

[0215] In response to the type of multimodal information including picture information, parsing the picture to obtain picture text information and picture graphic information, performing semantic analysis on the picture graphic information to obtain a picture graphic semantic representation, and integrating the picture text information and the picture graphic semantic representation to obtain user input information corresponding to the picture information;

[0216] In response to the fact that the type of multimodal information includes video information, voice recognition is performed on the video information to obtain video voice text information, and the video is sliced ​​to obtain multiple video picture groups. Semantic analysis is performed on the multiple video picture groups in the order in which each video picture group is played in the video to obtain a graphic semantic representation of the video picture. Text parsing is performed on each video picture to obtain video picture text information. The video voice text information, video picture text information and the graphic semantic representation of the video picture are integrated to obtain user input information corresponding to the video information.

[0217] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0218] The technical features of the above embodiments can 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.

[0219] The above embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. An information interaction method, characterized in that: The method comprises: In response to obtaining user input information, the user input information is parsed using a language model to obtain user intent and target operation type, wherein the target operation type includes attendance operation, reimbursement operation, and leave operation; Determining a target operation instruction and a jump path of a target operating system based on the user intention and the target operation type; Jump to the target operating system based on the jump path, and execute the target operation on the target operating system based on the target operation instruction.

2. The method according to claim 1, characterized in that The method of using a language model to parse user input information and obtain user intent and target operation type includes: Determining invalid information in the user input information, deleting the invalid information from the user input information, and obtaining user input information to be processed, wherein the invalid information is information that is irrelevant to the user's intention; Perform correlation feature analysis on the user input information to be processed to obtain user input correlation information; Determining a plurality of user intention prompt information based on the to-be-processed user input information and the user input associated information; Retrieving a question matching the user input information to be processed from a memory database, inputting the question and the user intention prompt information into a language model, obtaining answer information output by the language model, and using the answer information as the user intention; Gets the target action type that matches the user intent.

3. The method according to claim 1, characterized in that The step of determining the target operation instruction and the jump path of the target operating system based on the user intention and the target operation type further includes: Obtaining a user intent attribute type, where the user intent attribute type includes a mutable user intent and an immutable user intent; In response to the user intention attribute being an immutable user intention, selecting an operation instruction that matches the user intention from an operation instruction library as a target operation instruction, and obtaining a jump path of a target operating system corresponding to the target operation type; In response to the user intention attribute being a variable user intention, a target user intention matching the user intention and a jump intention having an operation link with the target user intention are determined from the intention knowledge graph, and operation instructions matching the target user intention and the jump intention are selected from the operation instruction library as target operation instructions and jump operation instructions respectively, and the jump operation type corresponding to the jump operation instruction is obtained, and the jump path of the target operating system and the jump path of the jump operating system corresponding to the target operation type and the jump operation type are obtained respectively, wherein the intention knowledge graph includes multiple user intentions, and different user intentions construct a network adjacency relationship through operation links, and the operation links represent the jump paths between different user intentions under different tasks.

4. The method according to claim 1, wherein The target operation type is an attendance operation, and executing the target operation on the target operating system based on the target operation instruction includes: In response to the target operation type being an attendance operation, acquiring a target operation instruction type, the target operation instruction type including a first attendance operation instruction and a second attendance operation instruction; In response to the target operation instruction type being the first attendance operation instruction, obtaining a path to a browsing page of the first attendance operation instruction, and jumping to the browsing page of the first attendance operation instruction to perform the first attendance operation; In response to the target operation instruction type being the second attendance operation instruction, the target attendance time in the user input information and the path of the browsing page of the second attendance operation instruction are obtained, and the browsing page of the second attendance operation instruction is jumped to, and the second attendance operation is performed based on the target attendance time.

5. The method according to claim 1, wherein The target operating system type is a reimbursement operation, and executing the target operation on the target operating system based on the target operation instruction includes: In response to the target operation type being a reimbursement operation, acquiring a target operation instruction type, the target operation instruction type including a first reimbursement operation instruction and a second reimbursement operation instruction; In response to the target operation instruction type being a first reimbursement operation instruction, obtaining a path to a browsing page of the first reimbursement operation instruction, and jumping to the browsing page of the first reimbursement operation instruction to perform the first reimbursement operation; In response to the target operation instruction type being the second reimbursement operation instruction, the reimbursement reason, reimbursement value, reimbursement identification number and the path of the browsing page of the second reimbursement operation instruction in the user input information are obtained, and the browsing page of the second reimbursement operation instruction is jumped to, and the second reimbursement operation is performed based on the reimbursement reason, the reimbursement value and the reimbursement identification number.

6. The method according to claim 1, characterized in that The target operating system type is a leave operation, and executing the target operation on the target operating system based on the target operation instruction includes: In response to the target operation type being a leave operation, acquiring a target operation instruction type, the target operation instruction type including a first leave operation instruction and a second leave operation instruction; In response to the target operation instruction type being the first leave operation instruction, obtaining a path to a browsing page of the first leave operation instruction, and jumping to the browsing page of the first leave operation instruction to perform the first leave operation; In response to the target operation instruction type being the second leave operation instruction, the first time, the second time and the path of the browsing page of the second leave operation instruction in the user input information are obtained, and the browsing page of the second leave operation instruction is jumped to, and the second reimbursement operation is performed based on the first time and the second time.

7. The method according to claim 1, characterized in that The step of obtaining the user input information includes: Acquire a type of multimodal information input by a user, where the multimodal information type includes at least one of text information, image information, and video information; In response to the type of the multimodal information including text information, using the text information as user input information corresponding to the text information; In response to the type of the multimodal information including picture information, parsing the picture to obtain picture text information and picture graphic information, performing semantic analysis on the picture graphic information to obtain a picture graphic semantic representation, and integrating the picture text information and the picture graphic semantic representation to obtain user input information corresponding to the picture information; In response to the fact that the type of the multimodal information includes video information, voice recognition is performed on the video information to obtain video voice text information, and the video is sliced ​​to obtain multiple video picture groups. Semantic analysis is performed on the multiple video picture groups in sequence according to the playback order of each video picture group in the video to obtain a graphic semantic representation of the video picture. Text parsing is performed on each video picture to obtain video picture text information. The video voice text information, the video picture text information and the graphic semantic representation of the video picture are integrated to obtain user input information corresponding to the video information.

8. An information interaction device, characterized in that: The device comprises: An acquisition module is configured to, in response to obtaining user input information, parse the user input information using a language model to obtain user intent and target operation type, wherein the target operation type includes attendance operation, reimbursement operation, and leave operation; A determination module, configured to determine a target operation instruction and a jump path of a target operating system based on the user intention and the target operation type; An operation module is configured to jump to a target operating system based on the jump path, and execute a target operation on the target operating system based on the target operation instruction.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images