Conversation interactive service closed-loop receipt integration method

Abstract

The invention relates to a dialogue interactive service closed-loop receipt integration method, belongs to the technical field of artificial intelligence, and solves the problems that an existing receipt integration interaction process cannot be flexibly adjusted and a service closed loop cannot be realized. The method comprises the steps of constructing a receipt integral interaction process and loading the receipt integral interaction process into a memory; guiding a user to upload a receipt picture through a dialogue node in the process, and obtaining and storing receipt element information; whether the receipt element information is complete or not is recognized through the judgment node, and whether the receipt element information is correct or not is recognized according to user feedback information; the receipt element information is fed back to the user through the reply node, and user feedback information is received; when it is recognized that the receipt element information is correct according to the user feedback information, the receipt element information is sent to the integral system through the query node; and when the receipt element information is identified to be incorrect for multiple times according to the user feedback information, feeding back to the user through a reply node and entering a manual auditing system. And an interactive service closed loop of receipt points is realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a closed-loop ticket points method for interactive dialogue services. Background Art

[0002] Receipt points, as a common marketing tool, can effectively attract consumers' attention and promote sales growth.

[0003] However, the traditional receipt points system requires customers to go to the receipt points counter and show the receipt to the staff, who will then enter the receipt points. This method brings great inconvenience to customers and also increases costs for merchants.

[0004] With the advancement of technology, in order to solve the problems of traditional receipt points, the market has provided some solutions, such as submitting shopping receipts to the receipt points system, the system automatically submits the receipt information, submits the receipt information to the points system, and cooperates with manual review. However, this method still has problems such as low accuracy of information extraction, time-consuming manual review, poor customer experience, high overall cost and inability to achieve a closed service loop. Summary of the invention

[0005] In view of the above analysis, an embodiment of the present invention aims to provide a dialog interactive service closed-loop ticket points method to solve the problem that the existing ticket points interaction process cannot be flexibly adjusted and cannot achieve a service closed loop.

[0006] An embodiment of the present invention provides a closed-loop ticket points method for interactive dialogue services, comprising the following steps:

[0007] Build a ticket points interaction process, save each node and its configuration information and load it into memory; each node includes: dialogue node, judgment node, reply node and query node;

[0008] When it is identified that the user's intention is related to the receipt points, the receipt points interaction process is started; the user is guided to upload the receipt picture through the dialogue node, and the receipt element information is obtained and saved; the receipt element information is identified through the judgment node whether it is complete, and the receipt element information is identified based on the user's feedback information whether it is correct; the receipt element information is fed back to the user through the reply node, and the user feedback information is received;

[0009] When the receipt element information is identified as correct based on user feedback information, the receipt element information is sent to the points system through the query node, ending the receipt points interaction process; when the receipt element information is identified as incorrect multiple times based on user feedback information, the information is fed back to the manual review system to the user through the reply node, ending the receipt points interaction process.

[0010] Based on the further improvement of the above method, the same number of slot entities as the receipt element information are set in the dialogue node, and the types of the slot entities are all set to system entities. The system entities are obtained by sending the receipt image uploaded by the user to the optical character recognition model to recognize the receipt text information, and then sending the receipt text information to the entity recognition model.

[0011] Based on the further improvement of the above method, the entity recognition model includes: a two-dimensional feature extraction module, a feature fusion module, a decoder module and an output module in sequence; the two-dimensional feature extraction module includes two parallel branches, the first branch includes a text embedding layer and an encoder layer connected in sequence; the second branch includes a Bert embedding layer, a width embedding layer and a splicing layer connected in sequence; the feature fusion module fuses the feature vectors output by the two branches in the two-dimensional feature extraction module through a linear layer, passes them to the decoder module to generate an output sequence, and finally obtains the identified entity through the output module.

[0012] Based on the further improvement of the above method, the dialogue node also includes: guiding words, timeout period, number of timeouts and counter-question strategy; after entering the dialogue node, identify whether the slot entity set by the dialogue node has a value. If there is no value, the set first guiding words are sent to the user to remind the user to upload the receipt picture. When the set timeout period is reached and the receipt picture is still not obtained, the set first guiding words are sent to the user again according to the set counter-question strategy, and the actual number of timeouts is recorded. When the actual number of timeouts exceeds the set timeout number, exit the dialogue node and enter the manual review system.

[0013] Based on the further improvement of the above method, after entering the dialogue node, it is identified whether the slot entity set by the dialogue node has a value. If it has a value, the set second guiding words will be sent to the user to remind the user to upload a clear receipt picture. When the receipt picture is obtained within the timeout period, for each receipt element, it is identified whether the newly acquired information is consistent with the value in the slot entity. If they are consistent, the value in the slot entity is completed according to the newly acquired information. If they are inconsistent, the newly acquired receipt element information will overwrite the value of the corresponding slot entity.

[0014] Based on the further improvement of the above method, when the judgment node is used to identify whether the receipt element information is complete, two branches are set in the judgment node, and one of them is the default branch. A condition group consisting of multiple conditions is set for the non-default branch; each condition is achieved by introducing a slot entity and setting the slot entity to have a value.

[0015] Based on the further improvement of the above method, when the judgment node identifies whether the receipt element information is correct based on the user feedback information, two branches are set in the judgment node, and one of them is the default branch, and a condition group consisting of 1 condition is set for the non-default branch; the condition is set according to the button variable and button return value set in the reply node.

[0016] Based on the further improvement of the above method, the reply node includes: reply type, reply content type and reply content; by setting the reply type to a template, the reply content type to text, setting a reply template in the reply content and introducing a slot entity as a variable, the receipt element information is fed back to the user through the reply node, and the obtained slot entity value is replaced with the variable in the reply template to obtain the receipt element information and send it to the user.

[0017] Based on the further improvement of the above method, the reply node also includes: whether to enable the button setting, when the button setting is enabled, set the button variable, select the button type and set the button wording and button return value; receiving user feedback information through the reply node is to receive the button return value corresponding to the button selected by the user, and set it in the button variable.

[0018] Based on the further improvement of the above method, the query node includes: setting the request type, request method, input parameters and their values, and output parameters; sending the receipt element information to the points system through the query node is to use the slot entity and its value as input parameters and its value, and call the points system set in the request method according to the set request type.

[0019] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0020] 1. By flexibly building automated interaction processes, the needs for receipt points can be quickly met; through the collaborative work of dialogue nodes, judgment nodes, and reply nodes in the automated interaction process, multiple rounds of dialogue can be achieved to cope with complex business scenarios and improve the user dialogue experience.

[0021] 2. By guiding users to upload receipt images multiple times, we ensure that the collected receipt information is accurate, greatly improving the overall accuracy of receipt recognition. We dynamically execute the process based on the customer's responses and operations, complete the points process in different situations in real time, and realize an interactive service closed-loop receipt points process, reducing the time and workload of manual processing of receipt points, reducing manpower input costs, and improving the efficiency of points processing.

[0022] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components;

[0024] Figure 1 This is a flow chart of a closed-loop ticket points method for interactive dialogue services in an embodiment of the present invention;

[0025] Figure 2 Schematic diagram of the ticket points interaction process in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0027] A specific embodiment of the present invention discloses a closed-loop ticket points method for interactive dialogue services, such as Figure 1 As shown, the following steps are included:

[0028] S1. Construct a ticket points interaction process, save each node and its configuration information and load it into memory; each node includes: dialogue node, judgment node, reply node and query node;

[0029] S2. When it is identified that the user's intention is related to the receipt points, the receipt points interaction process is started; the user is guided to upload the receipt picture through the dialogue node, and the receipt element information is obtained and saved; the receipt element information is identified through the judgment node whether it is complete, and the receipt element information is identified based on the user's feedback information whether it is correct; the receipt element information is fed back to the user through the reply node, and the user feedback information is received;

[0030] S3. When the receipt element information is identified as correct based on user feedback information, the receipt element information is sent to the points system through the query node, ending the receipt points interaction process; when the receipt element information is identified as incorrect multiple times based on user feedback information, the information is fed back to the manual review system to the user through the reply node, ending the receipt points interaction process.

[0031] During implementation, the demand for receipt points is quickly met by flexibly building an automated interaction process. Multi-round conversations are achieved through the collaborative work of dialogue nodes, judgment nodes, and reply nodes in the automated interaction process to cope with complex business scenarios, enhance user conversation experience, and improve the efficiency and accuracy of point processing.

[0032] Specifically, the ticket points interaction process in step S1 is constructed through a visual interface, including: start node, dialogue node, collection node, query node, judgment node, reply node, assignment node and end node. Each node has its own unique node ID, and is associated with the node ID of the previous node through the parent node ID, and is associated with the node ID of the next node through the child node ID, and different nodes are identified by node type.

[0033] Combine the following Figure 2 The schematic diagram of the ticket points interaction process is as follows, which specifically explains the configuration and implementation methods of each node.

[0034] (1) Start Node

[0035] The start node is only used as a process start node, and there is only one; in this embodiment, when it is recognized that the user's intention is related to the receipt points, the receipt points interaction process is started, and the child nodes are entered from the start node.

[0036] It should be noted that users communicate with the interactive dialogue system in text or voice through mini programs or official accounts. If it is voice, the user's voice is first converted into text through the automatic speech recognition model, and then the natural language processing model is used to identify the text intent. If the text intent is related to ticket points, the ticket points interaction process is started.

[0037] (2) Dialogue Node

[0038] The dialogue node is used to conduct multiple rounds of dialogue with the user through guided speech to collect the required information. In the dialogue node, the required information is obtained by constructing slot entities. The types of slot entities include: enumeration entities, regular entities, intention entities and system entities; among them, enumeration entities are predefined enumeration values, and when used, the entity information that is the same as the predefined enumeration value is obtained according to the information uploaded by the user; regular entities are predefined regular expressions, and when used, the entity information that conforms to the regular expression is extracted from the information uploaded by the user; intention entities are predefined similarity thresholds, and when used, the intent is identified from the information uploaded by the user, and the intent with the maximum similarity and greater than the similarity threshold is obtained as entity information; system entities are predefined entity recognition models, and when used, the entity recognition model is used to extract the corresponding entity information from the information uploaded by the user.

[0039] Furthermore, the dialogue node also includes: guiding speech, timeout period, number of timeouts and counter-question strategy; wherein, guiding speech is a fixed speech, used to remind users of the information they need to provide; timeout period is used to set the time to wait for users to upload information; timeout number is used to set the number of times to remind users to upload information; counter-question strategy is used in conjunction with timeout period, if the user does not reply after the timeout, guiding speech is actively sent to the user according to the counter-question strategy, and the user is guided to provide information again, so as to facilitate the collection of the value of the slot entity. Preferably, the counter-question strategy includes: custom speech and recent speech, when custom speech is selected, the configuration of counter-question speech is added, and the counter-question speech is sent to the user; when the recent speech is selected, the speech sent last time is automatically sent to the user.

[0040] In this embodiment, the information of each node involved in the ticket points interaction process and the parent-child association information between nodes are stored in the database. When the ticket points service is started, the relevant records of the ticket points interaction process are read from the database and loaded into the memory. Preferably, a redis server is used to store the data loaded into the memory.

[0041] It should be noted that, in this embodiment, the receipt element information to be collected includes: brand name, mall name, order number, transaction time and transaction amount. In this case, the same number of slot entities as the receipt element information are set in the dialogue node, and the types of the slot entities are all set to system entities, that is, the receipt text information is recognized by sending the receipt image uploaded by the user to the optical character recognition model, and then the receipt text information is sent to the entity recognition model to obtain the value of each slot entity.

[0042] Specifically, after entering the dialog node, it is identified whether the slot entity set in the dialog node has a value. If not, the first guidance words set are sent to the user to remind the user to upload the receipt picture. This situation corresponds to Figure 2 In the dialog node 1 after the start node, there is no value in the five slot entities, and the first guiding words are sent, such as: "Please upload a picture of the receipt."

[0043] Furthermore, when the set timeout period is reached and the receipt image is still not obtained, the set first guiding words will be sent to the user again according to the set counter-question strategy (that is, the most recent words are configured in the counter-question strategy), and the actual number of timeouts will be recorded. When the actual number of timeouts exceeds the set timeout number, the dialogue node will be exited and transferred to the manual review system.

[0044] After entering the dialogue node, identify whether the slot entity set in the dialogue node has a value. If it has a value, send the set second guidance words to the user to remind the user to upload a clear receipt picture. This situation corresponds to Figure 2The middle dialogue node 2 has executed the first dialogue node 1 and collected some or all of the receipt element information. However, because it is incomplete or the user believes it is incorrect, the second guiding words are sent, such as: "Please upload a clear picture of the receipt."

[0045] Furthermore, when the receipt image is obtained within the timeout period, the newly obtained information of each receipt element is identified to see whether it is consistent with the value in the slot entity. If it is consistent, the value in the slot entity is completed according to the newly obtained information. If it is inconsistent, the newly obtained receipt element information overwrites the value of the corresponding slot entity. In other words, this embodiment performs a consistency check on the collected data to ensure that the information of the same receipt element remains consistent in different collections, thereby improving the accuracy and reliability of the data.

[0046] It should be noted that the entity recognition model in this embodiment includes, in sequence: a two-dimensional feature extraction module, a feature fusion module, a decoder module and an output module; wherein the two-dimensional feature extraction module includes two parallel branches, the first branch includes a text embedding layer and an encoder layer connected in sequence; the second branch includes a Bert embedding layer, a width embedding layer and a splicing layer connected in sequence; the feature fusion module fuses the feature vectors output by the two branches in the two-dimensional feature extraction module through a linear layer, and passes them to the decoder module to generate an output sequence, and finally obtains the identified entity through the output module.

[0047] Specifically, the first branch of the two-dimensional feature extraction module focuses on capturing contextual information in the text and understanding the overall structure and semantics of the text; the text embedding layer includes a trainable word embedding matrix, which maps each word in the text to a vector space of a fixed size, and then adds the position encoding and passes it to the encoder layer; the encoder layer includes multiple encoders connected in sequence, each encoder includes a multi-head self-attention mechanism layer and a feedforward neural network layer connected in sequence, and each layer is followed by a residual connection and a layer normalization, and the encoder layer is used to convert the input sequence into a feature vector containing rich semantic information to obtain the first feature vector.

[0048] The second branch of the two-dimensional feature extraction module focuses on identifying the width and boundaries of entities; the Bert embedding layer uses the Bert model to obtain the token embedding vector and the special tag CLS vector that adds the whole sentence information; the width embedding layer includes a trainable width embedding matrix, which contains a fixed-size embedding vector for each width (from 1 to the maximum width). The width embedding vector is obtained using the width embedding matrix, that is, the width of the entity is k+1, which means that the entity contains k+1 tokens, then the width embedding vector of the entity is an embedding vector with a width of k+1 obtained by indexing in the width embedding matrix; the token embedding vector, the width embedding vector and the special tag CLS vector are concatenated through the concatenation layer to obtain the second feature vector.

[0049] The feature fusion module performs weighted fusion on the first feature vector and the second feature vector through a fully connected layer to obtain a fused feature vector which is passed to the decoder module.

[0050] The decoder module includes multiple decoders connected in sequence, each decoder includes a masked multi-head self-attention sublayer, an encoder-decoder attention layer, and a feedforward neural network layer connected in sequence; each layer is followed by a residual connection and a layer normalization.

[0051] The output module is mapped to the dimension of the number of labels through a linear layer, and then the entity is annotated through a softmax function or other classification functions (such as a CRF layer) to output the recognized entity.

[0052] (3) Judgment Node

[0053] The judgment node includes multiple branches, each branch includes at least one condition group, and each condition group includes at least one condition; the judgment symbols of the condition include: greater than, less than, equal to, not equal to, included, not included, with value, and without value. Among the multiple branches, there must be one default branch, which is used to enter the node pointed to by the default branch when the conditions of all branches are not met.

[0054] exist Figure 2 Judgment nodes 1 and 3 are used to identify whether the receipt element information is complete. Two branches are set in these two judgment nodes. A condition group consisting of multiple conditions is set for the non-default branch. Each condition introduces a slot entity, which corresponds to the brand name, mall name, order number, transaction time and transaction amount respectively. The judgment symbol is set to have a value, which means that all five slot entities must have a value. Correspondingly, the default branch indicates that the values ​​in the five slot entities are incomplete.

[0055] exist Figure 2Judgment nodes 2 and 4 are used to identify whether the receipt element information is correct based on user feedback information. Two branches are also set, and a condition group consisting of 1 condition is set for the non-default branch; the condition is set according to the button return value set in the reply node. Figure 2 The parent nodes of judgment node 2 and judgment node 4 are both reply nodes. In the reply node, the user feedback information is received by setting the button variable and the button return value, and stored in the memory. The condition in the judgment node corresponds to an expected user feedback situation by referencing the button variable and setting the judgment symbol to equal, such as the user feedback that the collected receipt element information is correct.

[0056] (4) Reply Node

[0057] The reply node includes: reply type, reply content type and reply content; reply types include: display and template; display means directly displaying the reply content according to the reply content type; template means obtaining dynamic values ​​to replace the variables in the reply template set in the reply content to obtain dynamic reply content, and displaying the dynamic reply content according to the reply content type. Reply content types include: text, picture, audio, video, file and link; reply content is configured according to the selected reply content type.

[0058] Figure 2 The settings of reply node 1 and reply node 2 are the same. Both set the reply type to template, the reply content type to text, set the reply template in the reply content and introduce the slot entity as a variable. During runtime, the values ​​of the five slot entities are obtained from the memory to replace the variables in the reply template and displayed to the user in the form of text, allowing the user to confirm whether the collected receipt element information is correct.

[0059] Furthermore, the reply node also includes: whether to enable button settings, when enabling button settings, setting button variables, selecting button types and setting button words and button return values; enabling button settings is used to display button words on the front end according to the button type to interact with users. Figure 2 The child nodes of reply node 1 and reply node 2 are both judgment nodes, and the judgment nodes need to jump to different branches based on the user's feedback on the reply content displayed by the reply node. Therefore, the button setting is enabled in reply node 1 and reply node 2, the button variable is set, the button type is selected, and the button wording and button return value are set. After the user selects the button for feedback, the user feedback information is received and stored in the memory.

[0060] For example, set the button variable to btnparam, select the button type as a radio button, add the button text "correct", and the corresponding button return value is "1"; the button text is "incorrect", and the corresponding button return value is "0". Generate two radio buttons on the front-end interface. After the user selects, set the button return value of the selected button to the button variable btnparam and store it in the memory. Subsequently, set btnparam equal to 1 in the branch condition of the judgment node, which means that the receipt element information collected from the user feedback is correct.

[0061] It should be noted that Figure 2 In the reply node 3, feedback is given to the user through the manual review system. At this time, there is no need to receive any feedback information from the user. In this case, there is no need to enable the button setting in the reply node 3. Set the reply type to display and the reply content type to text or picture. Display the text content entered in the reply content or the uploaded picture in the front-end interface, and use text content to inform the user to enter the manual review system, or use pictures to inform the user of the subsequent operation process.

[0062] (5) Query Node

[0063] The query node includes: setting the request type, request method, input parameters and their values, and output parameters; the request types include: GET and POST. Figure 2 Both query node 1 and query node 2 in the example send the receipt element information to the points system, that is, by using the slot entity and its value as input parameters and their values, the points system set in the request method is called according to the set request type.

[0064] (6) End Node

[0065] The end node is used to mark the end of the process operation, and there is only one.

[0066] In step S3, the process ends when two conditions are met: the user feedback that the receipt element information is correct, the receipt element information is sent to the points system, and the user completes the points record normally; and, if the correct and complete receipt element information is not collected for two consecutive times, it is transferred to manual review and further completion of the points record is required.

[0067] (7) Collection nodes and assignment nodes

[0068] Preferably, this embodiment further includes a collection node and an assignment node when constructing an interaction process.

[0069] Among them, the assignment node is used to process variables, add new variables and assign values, or modify variable values ​​after referencing existing variables, thereby increasing the flexibility of the process.

[0070] The collection node is used to collect information filled in by users through the set form, including: designing the form, selecting the form, prompting words, timeout time and number of timeouts. Use "design form" to create a form for collecting information in the form of a form, select the designed form through "select form", and remind users to enter information in the form through "prompt words"; when entering the collection node, the front-end interface displays the form link. If the user fails to complete the filling within the set timeout time, the prompt words will be output on the front-end interface. If the set timeout times are exceeded, the collection node will be exited; otherwise, enter the next node.

[0071] Compared with the prior art, the closed-loop receipt points method for interactive conversational services provided in this embodiment can quickly meet the needs of receipt points by flexibly constructing an automated interaction process; multiple rounds of conversations can be achieved through the collaborative work of conversation nodes, judgment nodes, and reply nodes in the automated interaction process to cope with complex business scenarios and improve the user conversation experience; the accuracy of overall receipt recognition can be greatly improved by guiding users to upload receipt images multiple times to ensure that the collected receipt information is accurate; the process is dynamically executed according to the customer's response and operation, and the points process in different situations is completed in real time, realizing an interactive service closed-loop receipt points process, reducing the time and workload of manual processing of receipt points, reducing manpower input costs, and improving the efficiency of points processing.

[0072] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0073] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A closed-loop ticket scoring method for interactive dialogue services, characterized in that: The following steps are involved: Build a ticket points interaction process, save each node and its configuration information, and load it into memory; The nodes include: dialogue node, judgment node, reply node and query node; When it is identified that the user's intention is related to the receipt points, the receipt points interaction process is started; the user is guided to upload the receipt picture through the dialogue node, and the receipt element information is obtained and saved; the receipt element information is identified through the judgment node whether it is complete, and the receipt element information is identified based on the user's feedback information whether it is correct; the receipt element information is fed back to the user through the reply node, and the user feedback information is received; When the receipt element information is identified as correct based on user feedback information, the receipt element information is sent to the points system through the query node, ending the receipt points interaction process; when the receipt element information is identified as incorrect multiple times based on user feedback information, the information is fed back to the manual review system to the user through the reply node, ending the receipt points interaction process.

2. The closed-loop ticket points method for interactive dialogue services according to claim 1, characterized in that: The same number of slot entities as the receipt element information are set in the dialogue node, and the types of the slot entities are all set to system entities. The system entities are obtained by sending the receipt image uploaded by the user to the optical character recognition model to recognize the receipt text information, and then sending the receipt text information to the entity recognition model.

3. The closed-loop ticket points method for interactive dialogue services according to claim 2, characterized in that: The entity recognition model comprises in sequence: a two-dimensional feature extraction module, a feature fusion module, a decoder module and an output module; wherein the two-dimensional feature extraction module comprises two parallel branches, the first branch comprises a text embedding layer and an encoder layer connected in sequence; the second branch comprises a Bert embedding layer, a width embedding layer and a splicing layer connected in sequence; the feature fusion module fuses the feature vectors output by the two branches in the two-dimensional feature extraction module through a linear layer, and transmits them to the decoder module to generate an output sequence, and finally obtains the recognized entity through the output module.

4. The closed-loop ticket points method for interactive dialogue services according to claim 2, characterized in that: The dialogue node also includes: guiding words, timeout period, number of timeouts and counter-question strategy; after entering the dialogue node, it is identified whether the slot entity set by the dialogue node has a value. If there is no value, the set first guiding words are sent to the user to remind the user to upload the receipt picture. When the set timeout period is reached and the receipt picture is still not obtained, the set first guiding words are sent to the user again according to the set counter-question strategy, and the actual number of timeouts is recorded. When the actual number of timeouts exceeds the set timeout number, the dialogue node is exited and the manual review system is entered.

5. The closed-loop ticket points method for interactive dialogue services according to claim 4, characterized in that: After entering the dialogue node, identify whether the slot entity set for the dialogue node has a value. If so, send the set second guiding words to the user to remind the user to upload a clear receipt picture. When the receipt picture is obtained within the timeout period, identify for each receipt element whether the newly acquired information is consistent with the value in the slot entity. If they are consistent, complete the value in the slot entity according to the newly acquired information. If they are inconsistent, overwrite the corresponding slot entity value with the newly acquired receipt element information.

6. The closed-loop ticket points method for interactive dialogue services according to claim 2, characterized in that: When the judgment node is used to identify whether the receipt element information is complete, two branches are set in the judgment node, and one of them is a default branch. A condition group consisting of multiple conditions is set for the non-default branch; each condition is achieved by introducing a slot entity and setting the slot entity to have a value.

7. The closed-loop ticket points method for interactive dialogue services according to claim 2, characterized in that: When the judgment node identifies whether the receipt element information is correct based on user feedback information, two branches are set in the judgment node, and one of them is a default branch. A condition group consisting of 1 condition is set for the non-default branch; the condition is set based on the button variable and button return value set in the reply node.

8. The closed-loop ticket points method for interactive dialogue services according to claim 2, characterized in that: The reply node includes: reply type, reply content type and reply content; by setting the reply type to a template, the reply content type to text, setting a reply template in the reply content and introducing a slot entity as a variable, the receipt element information is fed back to the user through the reply node, and the acquired slot entity value is replaced with the variable in the reply template to obtain the receipt element information and send it to the user.

9. The closed-loop ticket points method for interactive dialogue services according to claim 2, characterized in that: The reply node also includes: whether to enable button setting, when enabling button setting, setting button variables, selecting button type and setting button words and button return value; receiving user feedback information through the reply node is to receive the button return value corresponding to the button selected by the user, and set it in the button variable.

10. The closed-loop ticket points method for interactive dialogue services according to claim 2, characterized in that: The query node includes: setting the request type, request method, input parameters and their values, and output parameters; sending the receipt element information to the points system through the query node is to use the slot entity and its value as the input parameter and its value, and call the points system set in the request method according to the set request type.

Images