Interaction instruction construction method and device, electronic equipment and storage medium

Abstract

The invention provides an interaction instruction construction method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining the historical interaction behavior information of a user, constructing an interaction behavior perception tree, determining a plurality of candidate interaction demands, and determining the candidate interaction demand most matched with a user portrait as a user interaction demand; performing knowledge base retrieval based on the user interaction demand, determining a corresponding service function contact, and determining an executable instruction set corresponding to the service function contact; sorting the instructions in the executable instruction set and issuing the instructions based on the user interaction context information; and if the feedback score of the user for the currently issued executable instruction is smaller than a preset score threshold value, issuing a new executable instruction until the user feedback score is greater than the preset score threshold value so as to obtain an instruction which is most matched with the user demand. According to the method, the execution of the instruction can be autonomously driven, a feedback loop is formed, flexible adjustment during operation is supported, and the self-adaptive capability in a variable environment is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of interactive instruction construction, and in particular to an interactive instruction construction method, device, electronic device and storage medium. Background Art

[0002] The large model instruction construction method is usually based on user input as a prompt, combined with contextual interaction semantic features and other task information, and after feature extraction, candidate instructions are screened out and further optimized to match the target task instructions.

[0003] However, the current instruction construction method relies on the user's active input and the contextual information required for instruction construction to generate output, and does not actively optimize and adjust the user's feedback. When the instruction cannot be executed normally or the instruction matching error leads to abnormal function execution (such as unresponsive function), there is a lack of optimization and adjustment means, and there is also a lack of diversified interactive tasks based on mobile terminals (such as active reminders, gesture interaction, page navigation, etc.). A cycle process of instruction-driven user UI (User Interface) interaction has not been formed, which affects the conversion efficiency of instruction interaction.

[0004] Therefore, how to achieve self-driven construction of interactive instructions and flexible adjustment during runtime is a problem that needs to be solved urgently. Summary of the invention

[0005] The present invention provides an interactive instruction construction method, device, electronic device and storage medium, which are used to realize self-driven construction of interactive instructions and flexible adjustment during operation.

[0006] The present invention provides an interactive instruction construction method, comprising: Obtain the user's historical interaction behavior information and build an interaction behavior perception tree; Based on the interaction behavior perception tree, multiple candidate interaction requirements are determined, and the candidate interaction requirement that best matches the user profile is determined as the user interaction requirement; Perform a knowledge base search based on the user interaction requirement, determine the business function touchpoint corresponding to the user interaction requirement, and determine the executable instruction set corresponding to the business function touchpoint; Based on the user interaction context information, the instructions in the executable instruction set are sorted, and the instructions are issued based on the sorted executable instruction set; If the user's feedback score for the currently issued executable instruction is less than the preset score threshold, a new executable instruction is issued until the user's feedback score for the new executable instruction is greater than the preset score threshold to obtain the executable instruction that best matches the user's actual needs.

[0007] According to an interactive instruction construction method provided by the present invention, the user's historical interactive behavior information includes the interactive instructions executed by the user on each navigation stack function page and the interactive results obtained by executing the interactive instructions; the acquiring of the user's historical interactive behavior information and construction of the interactive behavior perception tree includes: The application window is used as a first-level node, and a plurality of navigation stack function pages are inserted as child nodes of the first-level node to obtain a plurality of second-level nodes; For each second-level node, an interaction instruction executed by the user on the navigation stack function page corresponding to the second-level node and an interaction result obtained by executing the interaction instruction are inserted as child nodes of the second-level node to obtain at least one third-level node, so as to obtain an initial interaction behavior perception tree; The initial interaction behavior perception tree is pruned to obtain an interaction behavior perception tree.

[0008] According to an interaction instruction construction method provided by the present invention, the method of determining multiple candidate interaction requirements based on the interaction behavior perception tree, and determining the candidate interaction requirement that best matches the user profile as the user interaction requirement, includes: Traversing the interaction behavior perception tree, finding a path set whose completion degree is not 1; the path set whose completion degree is not 1 is used to represent multiple candidate interaction requirements; For each user portrait dimension, based on the user portrait dimension and the weight of the user portrait dimension under the corresponding candidate interaction requirement, calculate a weighted score of the user portrait dimension; The candidate interaction requirement corresponding to the user portrait dimension with the highest weighted score is determined as the user interaction requirement.

[0009] According to an interactive instruction construction method provided by the present invention, the step of determining an executable instruction set corresponding to the business function touchpoint includes: Perform word segmentation processing on the description texts in the business function instruction set, the business function contact instruction set, and the human-computer interaction instruction set, and obtain the business function instruction initial word segmentation result, the business function contact word segmentation result, and the human-computer interaction instruction word segmentation result respectively; Perform string fuzzy matching on the initial segmentation result of the business function instruction, the segmentation result of the business function contact point and the segmentation result of the human-computer interaction instruction, and remove irrelevant words in the initial segmentation result of the business function instruction based on the matching result to obtain the segmentation result of the business function instruction; Based on the business function instruction word segmentation results, construct multiple contact point-instruction item sets; Filter out the target contact-instruction item set that meets the threshold requirements of support, confidence and lift; Based on the target contact-instruction item set, an executable instruction set corresponding to the business function contact is determined.

[0010] According to an interactive instruction construction method provided by the present invention, the construction of multiple touchpoint-instruction item sets based on the business function instruction word segmentation result includes: Based on the instructions whose occurrence times in the business function instruction segmentation results are greater than the minimum support threshold, constructing an instruction candidate 1-item set, and determining the instruction candidate 1-item set whose support is greater than or equal to the minimum support threshold as the instruction frequent 1-item set; Based on the business function contacts whose number of occurrences in the business function instruction word segmentation results is greater than the minimum support threshold, a contact candidate 1-item set is constructed, and the contact candidate 1-item set whose support is greater than or equal to the minimum support threshold is determined as a contact frequent 1-item set; Multiple instruction frequent 1-item sets and multiple contact frequent 1-item sets are arranged and combined to generate multiple contact-instruction item sets.

[0011] According to an interactive instruction construction method provided by the present invention, the support, confidence and lift of the contact-instruction item set are calculated in the following way: For each contact-instruction item set, determining a target instruction frequent 1-item set and a target contact frequent 1-item set in the contact-instruction item set; Calculate the support of the target instruction frequent one-item set based on the number of business function instructions in the business function instruction segmentation results that include the target instruction frequent one-item set and the total number of business function instructions; Calculate the support of the target contact frequent one-item set based on the number of business function instructions in the business function instruction segmentation result that include the target contact frequent one-item set and the total number of business function instructions; Calculate the support of the contact-instruction item set based on the number of business function instructions that simultaneously include the target instruction frequent 1-item set and the target contact frequent 1-item set in the business function instruction segmentation result, and the total number of business function instructions; Calculating the confidence of the contact-instruction item set based on the support of the contact-instruction item set and the support of the instruction 1 item set; The lift of the contact-instruction itemset is calculated based on the support of the contact-instruction itemset, the support of the target instruction frequent 1-itemset and the support of the target contact frequent 1-itemset.

[0012] According to an interactive instruction construction method provided by the present invention, the user's feedback score for the currently issued executable instruction is calculated in the following way: Obtain the number of executions of commands adopted by users, the length of the command execution path, the command execution time, and the length of time that users interact with the functional touchpoints; A weighted sum calculation is performed based on the weight coefficients corresponding to the number of execution times of the instructions adopted by the user, the length of the instruction execution path, the instruction execution time, and the interaction time of the user at the functional contact point, so as to obtain the user's feedback score for the currently issued executable instruction.

[0013] The present invention also provides an interactive instruction construction device, comprising: The perception tree construction module is used to obtain the user's historical interaction behavior information and construct the interaction behavior perception tree; An interaction requirement determination module, configured to determine a plurality of candidate interaction requirements based on the interaction behavior perception tree, and determine the candidate interaction requirement that best matches the user profile as the user interaction requirement; An executable instruction determination module, used to search the knowledge base based on the user interaction requirement, determine the business function touchpoint corresponding to the user interaction requirement, and determine the executable instruction set corresponding to the business function touchpoint; An instruction sorting and issuing module, used for sorting the instructions in the executable instruction set based on the user interaction context information, and issuing instructions based on the executable instruction set after the instruction sorting; The user feedback module is used to issue new executable instructions if the user's feedback score for the currently issued executable instructions is less than a preset score threshold, until the user's feedback score for the new executable instructions is greater than the preset score threshold, so as to obtain the executable instructions that best match the user's actual needs.

[0014] The present invention also provides an electronic 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, any one of the above-mentioned interactive instruction construction methods is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the interactive instruction construction method described in any one of the above is implemented.

[0016] The interactive instruction construction method, device, electronic device and storage medium provided by the present invention construct an interactive behavior perception tree by acquiring the user's historical interactive behavior information; based on the interactive behavior perception tree, multiple candidate interactive requirements are determined, and the candidate interactive requirement that best matches the user portrait is determined as the user interactive requirement; based on the user interactive requirement, a knowledge base is searched to determine the business function touchpoints corresponding to the user interactive requirement, and the executable instruction set corresponding to the business function touchpoints is determined; based on the user interactive context information, the instructions in the executable instruction set are sorted, and the instructions are issued based on the executable instruction set after the instructions are sorted; if the user's feedback score for the currently issued executable instruction is less than the preset score threshold, a new executable instruction is issued until the user's feedback score for the new executable instruction is greater than the preset score threshold, so as to obtain the executable instruction that best matches the user's real needs. The present invention can autonomously drive the execution of interactive instructions, form a feedback loop, support flexible adjustment at runtime, and enhance the adaptive ability in a changing environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 It is a flowchart of the interactive instruction construction method provided by an embodiment of the present invention.

[0019] Figure 2 It is a schematic diagram of an instruction adaptation framework provided by an embodiment of the present invention.

[0020] Figure 3 It is a schematic diagram of the interactive behavior perception tree model structure provided by an embodiment of the present invention.

[0021] Figure 4 It is a structural diagram of an interactive instruction construction device provided in an embodiment of the present invention.

[0022] Figure 5 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In the description of the embodiments of the present invention, the terms "include", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0025] Figure 1 Schematic diagram of the flow of the interactive instruction construction method provided by the embodiment of the present invention. Figure 1 The embodiment of the present invention provides a method for constructing an interactive instruction, which may specifically include the following steps: Step 101: Obtain the user's historical interaction behavior information and construct an interaction behavior perception tree.

[0026] It should be noted that the execution subject of the interactive instruction construction method provided in the embodiment of the present invention may be an electronic device, a component in an electronic device, an integrated circuit or a chip. The electronic device may be a mobile electronic device or a non-mobile electronic device. For example, the mobile electronic device may be a mobile phone, a tablet computer, a laptop computer, a PDA, a wearable device, an ultra-mobile personal computer (UMPC), a netbook or a personal digital assistant (PDA), etc., and the non-mobile electronic device may be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present invention.

[0027] In some embodiments, the user's historical interactive behavior information may refer to the user's interactive behavior in the mobile terminal APP (Application), which may include the user's browsing touch information (including but not limited to pages, buttons, pop-up windows, etc.), the user's gesture operation (including but not limited to single click, double click, slide, etc.). By collecting the user's interactive behavior in the APP, it can be used to perceive the user's interactive needs.

[0028] In the APP, the display of any page can be composed of a view layout hierarchy. The view layout can be a tree-like data structure (including root nodes, element nodes, attribute nodes, text nodes, etc.), each node in the tree is a view, which can be text, pictures, or even container views. Each node can contain corresponding content and content ID (Identity document), content type, child nodes, and other information. H5 pages can be called DOM trees (Document Object Model), and mobile native rendering pages can be called DOM model trees, so that the model tree structure can accurately describe the interrelationships between the functions of each tag on the page.

[0029] In some embodiments, an interactive behavior perception tree can be constructed based on the DOM model tree structure. The interactive behavior perception model can include multiple child nodes and leaf nodes. Each node in the tree structure except the root node can represent a user's interactive action based on an instruction. Child nodes can be inserted into the position of the DOM model tree according to time, access order and functional contact. Leaf nodes can be represented as the terminal nodes where user needs are completed. The path from the root node to each leaf node can constitute an instruction set of the user's interactive intention. The internal components of the node Node may include a parent node, a child node set, and the current business function contact entity and instruction entity.

[0030] For example, from the time the user opens the App to the time the phone bill recharge is completed, each interactive instruction is a node in the tree structure, and the path of click jump - phone bill recharge - recharge successful contains multiple instructions, which can constitute an instruction set for the "complete recharge" scenario.

[0031] In some embodiments, a complete interaction path may include a start node and an end leaf node. Based on the characteristics of the operator's app, the business type may be divided into three major interaction scenarios: browsing and query, handling, and customer service consultation. Exiting the app may be used as a completion indicator for the browsing and query and customer service consultation interaction paths, and the success or failure of the business handling may be used as a completion indicator for the interaction path.

[0032] For example, if you exit the App after completing "Query Subscribed Services", it can be indicated that this path is completed and the package upgrade or downgrade is processed; if you exit the App after not completing "Query Subscribed Services", it can be considered that this interaction path is incomplete.

[0033] Step 102: Based on the interaction behavior perception tree, multiple candidate interaction requirements are determined, and the candidate interaction requirement that best matches the user profile is determined as the user interaction requirement.

[0034] In some embodiments, the interactive behavior perception tree model can be traversed starting from the root node to find a set of paths with a completion degree other than 1, and sorted according to the insertion time of the nodes. Each path with a completion degree other than 1 represents a user requirement that may need to continue interacting (i.e., a candidate interaction requirement).

[0035] For example, for the "broadband processing" functional scenario, only the broadband information is filled in, but it is not submitted for acceptance. The App is exited or switched to other functional scenarios for interaction until the App is exited. In this case, the completion degree of the path from the root node to the "broadband processing" node is not 1, and "broadband processing" can be regarded as an element in the set (that is, "broadband processing" is a candidate interaction requirement).

[0036] In some embodiments, while collecting the user's interactive behavior in the APP, the user's business information can be collected to build a user profile. The user's business information may include but is not limited to the balance of phone bills, package information, traffic information, etc.

[0037] In some embodiments, user portrait matching rules can be set based on user portraits and business functions. If the user portrait matching rules are not met, it can indicate that the user has relevant interaction needs and needs to be guided to complete. The relevant condition rules can be insufficient call charges, insufficient traffic, cards and coupons to be used, and rights to be collected. Under the reference and guidance of relevant business personnel, the priorities are set in sequence, and the candidate interaction needs with a completion degree not equal to 1 are combined with the user portrait matching rules to confirm and output a candidate interaction need that the user is most concerned about as the user interaction need.

[0038] Step 103: perform a knowledge base search based on the user interaction requirement, determine the business function touchpoint corresponding to the user interaction requirement, and determine the executable instruction set corresponding to the business function touchpoint.

[0039] In an embodiment of the present invention, the user interaction needs perceived based on the interaction behavior perception tree and the collected basic information can be used as input, and the business function contacts related to the user needs can be retrieved by searching the knowledge base according to the user's interaction needs, which can be used to build a matching model between human-computer interaction instructions and App business function contacts, thereby determining the executable instruction set corresponding to the business function contacts.

[0040] In the embodiment of the present invention, basic information such as page information, touch information set (Touch Collection), human-computer interaction instruction set based on mobile terminal App, business function instruction set (Business Instruction Collection) and the like may be collected.

[0041] In some embodiments, the mobile-based page can be natively implemented, or can be H5 or mini-program, which are all different forms of presentation based on functions. The page information can be a composite five-tuple consisting of <name, link, function description, page id, page category>.

[0042] In some embodiments, each page may include multiple touch points, which may be a general term for various services or options, etc. Different pages may include different touch point information, such as phone charge recharge, broadband application, etc., or different options for recharge amounts. The touch point information may include <touch point name, touch point id, touch point description, touch point function code / link, exposure time, exposure page, number of clicks, exposure frequency>, and a touch point set may be composed of multiple different touch points.

[0043] In some embodiments, an instruction is the smallest unit to complete a functional logic. Multiple instructions can build a huge functional system, which is regarded as an instruction set. The human-computer interaction instruction set based on mobile apps (including but not limited to smartphones, tablets, wearable devices) is the native system (including but not limited to iOS, Android, Harmony, etc.) instructions provided by the mobile system, which can support technical developers to complete and support the realization of business functions based on product requirements.

[0044] The instructions in the human-computer interaction instruction set may include but are not limited to single-click, double-click, drag, display pop-up window, display text, voice broadcast, play video, push, close page, close pop-up window, switch navigation stack, return / exit, zoom, rotate, fingerprint recognition, gesture sliding, redirection, etc. The names of instructions implemented based on different operating systems may be different, but the meaning of the instructions is the same. Each instruction may contain attribute fields such as <instruction definition (Chinese, English), instruction name, function explanation, instruction input parameter, instruction target, execution result>. Exemplarily, the "open" interaction instruction can be expressed as "open (open), openPage, explanation: open the specified page, input parameter 1, input parameter 2, instruction target: current controller, result is bool (0 / 1)".

[0045] In some embodiments, business function instructions can be encapsulated based on business function contacts and human-computer interaction instructions, and can be a data set sent to the mobile terminal for parsing and execution. Business function instructions can be understood within the App as a target task based on instruction execution, and a task executed once can be regarded as an interactive transaction.

[0046] Business function instructions can be actively triggered by the user, or they can be automatically triggered when the platform receives instructions on the App side. The attribute information of business function instructions can include <business function description, business function instruction category, business function instruction identifier, execution frequency, function contact identifier, function contact name, operation link (last business function instruction identifier), interactive action result, occurrence time, behavior scenario>. Exemplarily, the "display pop-up window" business function instruction can be expressed as "display a pop-up window on the home page, with the content: 'You have a phone bill voucher to be collected', instruction category: 'pop-up window', action result: 'Click to jump to the coupon page', occurrence time: '20241002 13:54:50', execution frequency: '12', order: '03'".

[0047] In some embodiments, the business function contacts can be developed by professional technicians outside the App and configured through the contact delivery platform, or they can be built into the App by developers inside the App. All functional contacts based on the App terminal constitute a contact set, which can be aggregated and organized based on the contact delivery platform or the operation management platform.

[0048] The human-computer interaction instruction set can be based on system capabilities and data capture system function API (Application Programming Interface) descriptions, and can be filtered, summarized and organized by business personnel or developers to form a human-computer interaction instruction set. The business function instruction set can be formed by statistically summarizing and organizing the function tracking points of the App terminal.

[0049] As an example, the collected contact information set, the human-computer interaction instruction set based on the mobile terminal App, and the set element keywords of the business function instruction set can be shown in Table 1: Table 1

[0050] Among them, each set code in the business function instruction set, such as bs02, bs02_1, can represent a group of interactive actions with a before-and-after relationship.

[0051] Step 104: sort the instructions in the executable instruction set based on the user interaction context information, and issue instructions based on the sorted executable instruction set.

[0052] In an embodiment of the present invention, the instructions in the executable instruction set can be sorted by analyzing the user interaction context information. The basis for sorting may include whether the instruction depends on the precondition, whether there is a binding relationship between the services involved in the instruction, etc. For example, the instruction "show nearby cinemas" depends on the precondition and requires the user to authorize the geographic location permission in advance, so the instruction "request geographic location permission" can be sorted before "show nearby cinemas".

[0053] In the embodiment of the present invention, based on the executable instruction set after the instruction sorting adjustment, the executable instructions can be sent to the mobile terminal for execution in the order of the instruction set.

[0054] Figure 2 Schematic diagram of the instruction adaptation framework provided by an embodiment of the present invention. Figure 2 In some embodiments, different from the existing "perception-decision-execution" classic control mode that relies on simple logical processing and preset control logic, the instruction adaptive framework provided by the embodiment of the present invention can form business function instruction task summary information through knowledge base retrieval, instruction protocol construction, filling instructions and other attribute information. The instruction task recognition is based on the perceived user interaction needs as the touch point as the input of the human-computer interaction instruction matching model, and combines the user interaction context information for semantic judgment and recognition, thereby eliminating redundancy and reducing the number of instructions in the matched optional instruction set. Under the premise of satisfying all relevant permission clause constraints, the order of instructions in the executable instruction set can be adjusted to form a single or multiple executable instructions with clear relationships. Under the guidance of this framework, the embodiment of the present invention can autonomously drive the execution of instructions and form a feedback loop, support flexible adjustment at runtime, and enhance its adaptability in a changing environment.

[0055] Step 105, if the user's feedback score for the currently issued executable instruction is less than the preset score threshold, a new executable instruction is issued until the user's feedback score for the new executable instruction is greater than the preset score threshold, so as to obtain the executable instruction that best matches the user's actual needs.

[0056] In an embodiment of the present invention, after the execution of the executable business function instruction is completed, the user interaction behavior data is analyzed to identify the user's feedback (positive feedback / negative feedback) on the issued instruction, and the instruction output is adjusted in real time according to the feedback result. Thus, the instruction can be adjusted in real time when the executed instruction does not meet the demand, thereby ensuring that the issued instruction can meet the user's demand.

[0057] In some embodiments, the feedback type (positive feedback / negative feedback) can be determined according to a feedback score calculation formula. If the user feedback score is less than 0.8, it can be considered negative feedback; if the user feedback score is greater than or equal to 0.8, it can be considered positive feedback.

[0058] In some embodiments, the instruction ranked first in the executable instruction set may be issued first. If the user provides negative feedback on the currently issued executable instruction, the instruction ranked second may be issued to the user from the matched executable instruction set, and the user's feedback on the issued new instruction may be identified again. If the user's feedback result on the issued new instruction is still negative feedback, it may be considered that the user demand perceived this time is inaccurate, and the interactive perception demand matching output may be autonomously driven to be re-executed, a new executable instruction set may be regenerated, and the instructions in the new executable instruction set may be issued.

[0059] The embodiment of the present invention performs instruction tuning based on user feedback scores, without the need for active user feedback (such as subjective evaluation methods such as scoring and questionnaires). By analyzing the user's interactive behavior data, the feedback score is automatically calculated, which can improve the objectivity and accuracy of the evaluation, and improve the interaction efficiency while ensuring the service quality. At the same time, the instruction output can be reversely adjusted according to the feedback results, forming a closed loop of instruction output-user feedback-instruction tuning, thereby effectively narrowing the gap with user needs.

[0060] In an embodiment of the present invention, an interactive behavior perception tree can be constructed based on the Dom tree model of the application page and the user's historical interactive behavior information, with the App window form as the root node of the tree, the business function driven by the user's interactive action as the child node, and a tree structure constructed according to the interactive record. Based on the user portrait, the incompleteness of the user's interactive path, and the relevant business rules, a matching rule algorithm is defined, so that the user's interactive needs can be perceived, and an executable instruction set is generated based on a matching model of a contact and a human-computer interaction instruction, and the executable instruction set is adjusted and output in combination with the user's interactive context to complete the perception and judgment of the interactive scene, and the instructions are transmitted to the mobile App through the established websocket connection to guide and drive the user to interact, and the user's interactive results and other information reversely drive the adjustment of the instructions until the user is guided to complete the desired demands in the App.

[0061] The embodiments of the present invention combine the human-computer interaction instructions and user interaction behaviors of the mobile terminal to construct and match the executable instruction set corresponding to each business function touch point, guide user interaction through the execution of instructions, and the feedback of user interaction can also reversely tune the construction of instructions, so as to timely discover anomalies and adjust the instruction output, avoid the unresponsiveness problem caused by instruction errors, better guide users to complete their demands in the App, and improve user perception.

[0062] The embodiment of the present invention constructs an interactive behavior perception tree by acquiring the user's historical interactive behavior information; based on the interactive behavior perception tree, multiple candidate interactive needs are determined, and the candidate interactive need that best matches the user portrait is determined as the user interactive need; based on the user interactive need, a knowledge base search is performed to determine the business function touchpoints corresponding to the user interactive need, and the executable instruction set corresponding to the business function touchpoints is determined; based on the user interactive context information, the instructions in the executable instruction set are sorted, and the instructions are issued based on the executable instruction set after the instructions are sorted; if the user's feedback score for the currently issued executable instruction is less than the preset score threshold, a new executable instruction is issued until the user's feedback score for the new executable instruction is greater than the preset score threshold, so as to obtain the executable instruction that best matches the user's real needs. The present invention can autonomously drive the execution of interactive instructions, form a feedback loop, support flexible adjustment at runtime, and enhance the adaptive ability in a changing environment.

[0063] In an optional embodiment, the user's historical interaction behavior information includes interaction instructions executed by the user on each navigation stack function page and interaction results obtained by executing the interaction instructions; the acquiring of the user's historical interaction behavior information and constructing the interaction behavior perception tree may specifically include: Step S11, taking the application window as a first-level node, and inserting a plurality of navigation stack function pages as child nodes of the first-level node to obtain a plurality of second-level nodes; Step S12: for each second-level node, inserting the interaction instruction executed by the user on the navigation stack function page corresponding to the second-level node and the interaction result obtained by executing the interaction instruction as a child node of the second-level node to obtain at least one third-level node, so as to obtain an initial interaction behavior perception tree; Step S13, pruning the initial interaction behavior perception tree to obtain an interaction behavior perception tree.

[0064] Figure 3 Schematic diagram of the interactive behavior perception tree model structure provided by an embodiment of the present invention. Figure 3 In an embodiment of the present invention, the App window can be used as a root node (first-level node), and each navigation stack function page can be inserted as a child node of the root node to obtain multiple second-level nodes. Based on the interaction instructions executed by the user on each navigation stack function page and the interaction results obtained by executing the interaction instructions, they can be inserted as child nodes of the second-level node in sequence to obtain at least one third-level node, thereby constructing an interaction behavior perception tree.

[0065] In some embodiments, if a click instruction is executed on the home navigation stack page and jumps to a certain A page, the A page after the jump and the corresponding instruction can be inserted into the second-level navigation page node as child nodes.

[0066] For example, if you click the recharge button on the home navigation stack page to enter the recharge page, you can traverse from the root node of the perception tree model to the "home" node of the secondary navigation stack, and insert the "recharge" node as a child node into the behavior perception tree. Similarly, each operation result of the command execution will be inserted into the tree model structure as a node. The interaction results generated by all the command executions from opening the App to exiting the App constitute an interaction path, gradually forming an interactive behavior perception tree model.

[0067] In some embodiments, the degree of completion of the path (from the root node to the leaf node) can be identified based on the scenario categories (three major interaction scenarios: browsing and query, processing, and customer service consultation) included in the user interaction path.

[0068] Completion refers to the ratio of the position of a certain functional touchpoint in the entire business path to the total number of touchpoints contained in the business path. The completion value is between 0 and 1. For browsing and query scenarios and customer service consultation scenarios, the completion is marked by content display and the user's active exit from the App. For processing scenarios, the completion can be marked by whether there is a completion node for the processing function in the interaction path. For example, the broadband processing service includes 6 touchpoints, and broadband appointment installation is the 4th touchpoint. The completion of the broadband appointment installation touchpoint is 0.66.

[0069] As instructions are continuously executed, the depth of the perception tree structure hierarchy and child nodes are constantly expanding, which seriously affects the traversal efficiency of the tree structure.

[0070] Therefore, the embodiment of the present invention can prune the interactive behavior perception tree structure according to the method that the node insertion time is more than a specified period from the current time (which can be set to a monthly period, such as 1 month) or the path completion degree is 1, that is, delete all nodes that do not meet the requirements in the third-level nodes, wherein the second-level nodes are set to retain N third-level node paths, and if the time condition is met but the conditions of N third-level nodes are not met, they will not be deleted, thereby ensuring that multiple paths for each third-level node are retained, which can provide a reference for the user's perception of interactive needs.

[0071] In an optional embodiment, determining multiple candidate interaction requirements based on the interaction behavior perception tree, and determining the candidate interaction requirement that best matches the user profile as the user interaction requirement may specifically include: Step S21, traversing the interaction behavior perception tree to find a path set whose completion degree is not 1; the path set whose completion degree is not 1 is used to represent multiple candidate interaction requirements; Step S22, for each user portrait dimension, based on the user portrait dimension and the weight of the user portrait dimension under the corresponding candidate interaction requirement, calculating a weighted score of the user portrait dimension; Step S23: determine the candidate interaction requirement corresponding to the user portrait dimension with the highest weighted score as the user interaction requirement.

[0072] In the embodiment of the present invention, the perception tree model can be traversed from the root node to find all the path sets with a completion degree not equal to 1. , sorted by the insertion time of the nodes, each path with a completion degree not equal to 1 represents a requirement that a user may need to continue to interact, thus obtaining multiple candidate interaction requirements.

[0073] In some embodiments, the user interaction requirements may be determined based on the following portrait matching rule algorithm: ; ; in, It is a matching rule algorithm based on user profiles and a set of paths with a completion degree not equal to 1. The priority matching logic of the algorithm can dynamically adjust the priority according to the actual interaction scenario and the user's personal portrait, so that the determined user interaction needs are more in line with the user's actual needs. The value can be [0, N]. The higher the value, the more it matches the rule, and the needs that best match this rule can be selected first.

[0074] in, Represents the weighted score for the i-th user portrait dimension; Represents a business rule set (i.e., a set of paths whose completion degree is not 1, which can represent multiple candidate interaction requirements), which can be insufficient call balance, insufficient traffic, or user level not meeting the requirements for receiving benefits, etc.; U represents the user's personal portrait, which can be personal points, age, package level, membership benefit level, or global communication level, and can be expanded according to specific business needs.

[0075] in, It is the dimension of user portrait, which can become a value between 0 and 1 after standardization.

[0076] in, It is defining business rules In the case of , the weight of user portrait dimension i can be a value between 0 and 1. The weight value can be determined according to the actual business proportion.

[0077] For example, suppose a China Unicom Gold Card user has 20 yuan in arrears, 99GB of data remaining, and 1 travel benefit to be used. 1 =0.8,U 2 = 0.3, U 3 = 0.5, at R 1 The weight w of the call charge dimension in the call charge service 1 = 0.4, at R 2 The weight w of the traffic dimension in the traffic charging service 2 = 0.2, at R 3 The global level dimension in the rights collection business 3 = 0.02, the rule weights of the three conditions can be calculated, which are S 1 =0.32, S 2 =0.06, S 3 =0.01. According to the calculation results, the rule matching degree of the user's arrears is the highest, so the arrears can be regarded as the user's interaction demand.

[0078] In an optional embodiment, the determining the executable instruction set corresponding to the business function touchpoint may specifically include: Step S31, performing word segmentation processing on the description texts in the business function instruction set, the business function contact instruction set and the human-computer interaction instruction set, and obtaining the business function instruction initial word segmentation result, the business function contact word segmentation result and the human-computer interaction instruction word segmentation result respectively; Step S32, performing string fuzzy matching on the initial segmentation result of the business function instruction, the segmentation result of the business function contact point and the segmentation result of the human-computer interaction instruction, and removing irrelevant words in the initial segmentation result of the business function instruction based on the matching result to obtain the segmentation result of the business function instruction; Step S33, constructing a plurality of contact-instruction item sets based on the business function instruction segmentation results; Step S34, screening out target contact-instruction item sets that meet the threshold requirements of support, confidence and lift; Step S35: determining an executable instruction set corresponding to the business function contact based on the target contact-instruction item set.

[0079] Business function instructions can be understood as target tasks, which are used to guide users to complete their target demands in the terminal. In an embodiment of the present invention, in order to guide users to complete their target demands in the terminal, a matching relationship between business function touchpoints and human-computer interaction instructions can be constructed; for each function touchpoint, the keyword information of the function touchpoint is extracted, and the business instruction set is segmented using a natural language processing word segmentation tool (Jieba), and the candidate instructions with high matching degree are determined as optional human-computer interaction instructions for business function instructions through association rule analysis.

[0080] In an embodiment of the present invention, the execution process based on the business function instruction can be represented as an interactive action. For example, if the business function instruction describes "click the bill button on the home page to jump to the phone bill", in this interactive action, the text description of the business function instruction can be split into "home page, click, bill, button, phone bill, jump" through the jieba word segmentation tool, so as to generate different candidate item sets according to the word segmentation results, and perform set screening according to association rules (support, confidence, and lift). When a specific threshold is met, the matching relationship table can be output based on the matching of the association relationship.

[0081] In an embodiment of the present invention, the Jieba word segmentation tool can be used to pre-process the description text of the collected business function instruction set (such as removing noise, deleting stop words, etc.), and then perform word segmentation operations on the pre-processed instruction description text.

[0082] In some embodiments, each description text in the business function instruction set can be segmented using the Jieba word segmentation tool to obtain a business function instruction phrase set containing multiple phrases, namely, a business function instruction word segmentation result set (BusinessInstructionWordCollection). Similarly, the description texts in the business function touch point set and the human-computer interaction instruction set can be segmented to obtain a business function touch point word segmentation result set (TouchWordCollection) and a human-computer interaction instruction word segmentation result set (InstructionWordCollectio), respectively.

[0083] In some embodiments, the initial word segmentation results of the business function instructions can be subjected to string fuzzy matching with the business function contact word segmentation results and the human-computer interaction instruction word segmentation results, respectively, to filter out the corresponding matching contact words and human-computer interaction instruction words in the initial word segmentation results of the business function instructions, and remove irrelevant words, thereby obtaining the final business function instruction word segmentation results.

[0084] In an optional embodiment, when constructing an item set, TF-IDF (Term Frequency-Inverse Document Frequency) can be used to calculate text similarity, determine keywords with similar semantics, and merge keyword descriptions with similar semantics into one item. For example, if "open the traffic recharge page" and "click on the traffic recharge page" are very similar in semantics, they can be treated as one item for processing, that is, "open" and "click" are both mapped to the human-computer interaction instruction "click", thereby reducing the number of item sets and improving the efficiency of the matching algorithm.

[0085] In some embodiments, the word segmentation result BusinessInstructionWordCollection obtained by word segmentation of the description text of the business function instruction set may be as shown in Table 2: Table 2

[0086] In an embodiment of the invention, for each contact-instruction frequent 2-item set, multiple contact-instruction frequent 2-item sets can be screened according to the calculated support, confidence and lift and the corresponding threshold requirements; assuming that a certain contact-instruction combination meets the threshold requirements of support, confidence and lift, it can be determined that there is a matching relationship between the contact-instruction 2-item set, and the contact-instruction 2-item set is determined to be the target contact-instruction item set, so that the executable instruction set corresponding to the business function contact can be determined based on the target contact-instruction item set.

[0087] The embodiment of the present invention combines human-computer interaction instructions with App business touchpoint functions based on an association rule algorithm and a natural language processing model (tf-idf). By constructing a matching algorithm model between human-computer instructions and business functions, it is possible to deeply analyze the complex relationship between instructions and touchpoints, and more accurately match and output a set of suitable human-computer interaction instruction sets. Based on the instruction set and according to the user's historical interaction behavior and interaction scenarios, the optimal interaction instruction tasks are automatically constructed and matched.

[0088] In an optional embodiment, the constructing of multiple contact-instruction item sets based on the business function instruction word segmentation result may specifically include: Step S331, constructing an instruction candidate 1-item set based on the instructions whose occurrence times in the business function instruction segmentation results are greater than the minimum support threshold, and determining the instruction candidate 1-item set whose support is greater than or equal to the minimum support threshold as the instruction frequent 1-item set; Step S332, constructing a contact candidate 1-item set based on the business function contacts whose number of appearances in the business function instruction word segmentation results is greater than the minimum support threshold, and determining the contact candidate 1-item set whose support is greater than or equal to the minimum support threshold as the contact frequent 1-item set; Step S333: Arrange and combine the multiple instruction frequent one-item sets and the multiple contact frequent one-item sets to generate multiple contact-instruction item sets.

[0089] In some embodiments, all business function instruction word segmentation result sets (BusinessInstructionWordCollection) can be scanned, and the number of occurrences of each different instruction and function contact therein can be counted. Based on the instructions whose number of occurrences is greater than the minimum support threshold, multiple instruction candidate 1-item sets can be constructed, and based on the contacts whose number of occurrences is greater than the minimum support threshold, multiple contact candidate 1-item sets can be constructed.

[0090] Exemplarily, if the number of occurrences of "click, pop-up, broadcast, read contacts, home page, call charges, recharge, global communication" is greater than the minimum support threshold, then multiple instruction candidate 1-item sets can be obtained: {"click"}, {"pop-up"}, {"broadcast"}, and multiple contact candidate 1-item sets: {"global communication"}, {"call charges"}, {"recharge"}, {"rights supermarket"}.

[0091] The minimum support threshold may be calculated and set based on actual experience or a value may be proposed, for example, it may be 0.1.

[0092] A frequent item set refers to a set of items in a data set that appear more frequently than a certain threshold. An association rule is an expression in the form of "X→Y", where X and Y are item sets. "X→Y" means that if X appears, then Y is also likely to appear, that is, "instruction A→contact point B" means that if "instruction A is executed, contact point B can be reached".

[0093] In some embodiments, the support of the candidate 1-item set can be calculated, and the candidate 1-item set with a support greater than or equal to a minimum support threshold is determined as a frequent 1-item set, thereby screening out the instruction frequent 1-item set and the contact frequent 1-item set. The support refers to the ratio of the number of business function instructions containing the item set to the total number of business function instructions.

[0094] In some embodiments, the screened multiple instruction frequent 1-item sets and multiple contact frequent 1-item sets can be arranged and combined to generate multiple contact-instruction 2-item sets, which need to meet the minimum support (minSupport) of all contact-instruction 2-item sets.

[0095] In an optional embodiment, the support, confidence and lift of the contact-instruction item set may be calculated by the following steps: Step S41, for each contact-instruction item set, determining a target instruction frequent 1-item set and a target contact frequent 1-item set in the contact-instruction item set; Step S42, calculating the support of the target instruction frequent one-item set based on the number of business function instructions in the business function instruction segmentation result that include the target instruction frequent one-item set and the total number of business function instructions; Step S43, calculating the support of the target contact frequent one-item set based on the number of business function instructions that contain the target contact frequent one-item set in the business function instruction segmentation result and the total number of business function instructions; Step S44, calculating the support of the contact-instruction item set based on the number of business function instructions that simultaneously include the target instruction frequent 1-item set and the target contact frequent 1-item set in the business function instruction segmentation result, and the total number of business function instructions; Step S45, calculating the confidence of the contact-instruction item set based on the support of the contact-instruction item set and the support of the instruction 1 item set; Step S46, calculating the lift of the contact-instruction itemset based on the support of the contact-instruction itemset, the support of the target instruction frequent one-item set and the support of the target contact frequent one-item set.

[0096] In some embodiments, the calculation formula of support may be as follows: ; A can represent one of the instruction candidate 1-item set or contact candidate 1-item set. For example, A can be any one of {"click"}, {"pop-up window"}, {"report"}, {"home page"}, {"phone charge recharge"}, {"rights supermarket"}; through the Support calculation, {"click"}, {"pop-up window"}, {"home page"}, {"phone charge recharge"}, {"global communication"} can be screened out to meet the minimum support threshold, and then they can be used as frequent 1-item sets, {"click"}, {"pop-up window"} as instruction frequent 1-item sets, {"home page"}, {"phone charge recharge"}, {"global communication"} as function contact frequent 1-item sets, thereby obtaining frequent 1-item sets for instructions and frequent 1-item sets for contacts.

[0097] In some embodiments, the support of a two-item set {A, B} is calculated, where A / B represents a command or a business touchpoint keyword set, such as A represents a frequent one-item set of a touchpoint, and B represents a frequent one-item set of a command, such as {"home", "click"}, {"home", "pop-up window"}, and the formula can be as follows: ; The metric can measure the occurrence frequency of the combination of instructions and contacts in the dataset, thus providing a basis for screening reliable matching relationships.

[0098] Exemplarily, assume that there are a total of 15 instructions, and {"click"}, {"pop-up window"} are selected as the instruction 1-item set, and {"home page"}, {"phone bill recharge"}, {"Global Connect"} are selected as the contact function 1-item set. The constructed 2-item sets include {"home page", "click"}, {"home page", "pop-up window"}, {"Global Connect", "click"}, etc. Then the support calculation results can be as follows: The support of the 1-item set {"Global Connect"} is Support("Global Connect") = 3 / 15, Support("click") = 4 / 15, Support("pop-up window") = 3 / 15, Support("home page") = 3 / 15, Support("home page", "click") = 1 / 15, Support("home page", "pop-up window") = 1 / 15, Support("Global Connect", "click") = 0, Support("Global Connect", "pop-up window") = 1 / 15.

[0099] Since the minimum support minSupport needs to be satisfied, according to experience, the minimum support threshold can be set to 0.26. Support("Global Connect", "click") = 0, and this item does not meet the requirement and can be directly filtered out.

[0100] In some embodiments, for a given dataset of instructions and contacts, the following method can be used to calculate the confidence of each contact-instruction 2-item set, that is, the probability that the elements in the matching item set appear in the historical dataset. The confidence calculation formula is: ; Among them, A can represent the contact item set, B can represent the instruction item set, and Confidence(A→B) can represent the probability that the instruction item set B appears when the contact item set A appears, that is, in the generated frequent 2-item set, the probability that the two elements appear in the same business instruction.

[0101] Exemplarily, Confidence("Global Connect"→"click") = 0, Confidence("Global Connect"→"pop-up window") = 1 / 3, Confidence("home page"→"click") = 1 / 3, Confidence("home page"→"pop-up window") = 1 / 3 can be calculated.

[0102] In some embodiments, the calculation formula of the lifting degree may be: ; The lift (A→B) can be used to determine whether the probability of the contact-instruction 2-item set {A, B} appearing in combination is higher than the probability of their independent appearance, thereby more accurately determining the truly meaningful matching relationship.

[0103] Based on the frequent set calculations generated by the above example: Confidence("Global Communication"→"Click") = 0, Confidence("Global Communication"→"Pop-up") = 1 / 3, Lift("Global Communication"→"Click") = 0, Lift("Global Communication"→"Pop-up") = 1 / 6.

[0104] In some examples, for each contact-instruction frequent 2-item set, screening can be performed based on the calculated support, confidence, and lift and the determined threshold. Assuming that the contact-instruction combination meets the threshold requirements of support, confidence, and lift, it can be determined that there is a matching relationship between its 2-item sets. Based on the above examples, the calculation results of the support, confidence, and lift of each contact-instruction frequent 2-item set can be shown in Table 3: Table 3

[0105] It can be seen that although the support and confidence of the functional instruction "pop-up window" corresponding to "Global Communication" meet the requirements, the "Homepage" touch point can select "pop-up window" as the optimal matching instruction.

[0106] In the embodiment of the present invention, the executable instruction set (Optional-Instruction Collection) corresponding to each functional contact can be screened out through the corresponding matching relationship table generated by the table.

[0107] In an optional embodiment, the user's feedback score for the currently issued executable instruction can be calculated by the following steps: Step S51, obtaining the number of execution times of the instructions adopted by the user, the length of the instruction execution path, the instruction execution time, and the interaction time of the user at the functional touch point; Step S52, a weighted sum calculation is performed based on the weight coefficients corresponding to the number of execution times of the instructions adopted by the user, the length of the instruction execution path, the instruction execution time and the interaction time of the user at the functional touch point, so as to obtain the user's feedback score for the currently issued executable instruction.

[0108] In an embodiment of the present invention, the dimensions of the feedback score calculation data may include: the number of execution times of instructions adopted by the user, the length of the instruction execution path, the instruction execution time, and the user's interaction time at the functional touch point. The calculation formula for the feedback score may be as follows: ; ; ; ; ; Where: N is the number of times the command is executed, and the value range is a positive integer, which depends on the user's operation in a specific scenario. N can represent the frequency with which the user executes a specific command. The fewer the number of executions, the higher the degree of completion of the command by the user. m It is the maximum number of instruction executions in the large data sample.

[0109] L is the execution path length, which ranges from positive integers. L can be quantified according to the specific interaction process, and can be used to quantify the hierarchical depth from the instruction start node to the instruction end node (for example, the hierarchical depth of the perception tree corresponding to an instruction is 4, and its path length is 4), reflecting the number of steps the user has gone through to complete an instruction. The smaller the path length, the lower the complexity of the instruction, and the easier it is for the user to complete the instruction task. L m It is the maximum execution path length in the large data sample.

[0110] T is the execution time, which is a non-negative real number. T is measured in seconds and represents the time required for the user to complete a specific operation. The shorter the execution time, the higher the execution efficiency of the instruction. m is the maximum execution time in the large data sample.

[0111] C is the duration of user interaction at a touchpoint, such as the time a user stays on a page, and its value range is a non-negative real number. C can reflect the user's concentration or possible hesitation at a specific interaction point. The shorter the interaction time, the higher the score of this dimension. m It is the maximum interaction time of users at a touchpoint in the big data sample.

[0112] w1, w2, w3, w4, and w5 can be the weight coefficients of each indicator respectively, and w1+w2+w3+w4+w5=1.

[0113] V is the user's feedback score on the instruction, and its value range can be [0, 1].

[0114] In summary, the embodiments of the present invention have the following advantages: (1) In combination with the characteristics of the mobile terminal, an innovative adaptive instruction generation and self-driven interactive instruction construction method is adopted. Compared with traditional matching methods that are often only applicable to specific scenarios or systems and have poor versatility, the multi-dimensional threshold determination method proposed in the present invention can flexibly adjust the thresholds of support, confidence and improvement according to different application scenarios and needs, so as to adapt to various human-computer interaction systems. In addition, through self-learning adjustment, the perception tree is used to construct matching candidate instructions, and the instruction order is adjusted under the premise of satisfying the authority constraints. The instructions to be executed are formed based on the optimal instructions, and the execution of instructions can be driven autonomously and a feedback loop can be formed. It supports flexible adjustment at runtime and adapts to changing environments. In addition, by providing a variety of interactive instructions to drive and guide users, the user interaction experience is improved. (2) Matching human-computer interaction instructions with app business functions to obtain execution instructions suitable for business functions. The embodiment of the present invention innovatively uses deep learning algorithms to realize the recognition and judgment of mobile terminal human-computer interaction instructions and business functions, and builds a perception tree through the user's historical interaction behavior, perceives the user's interaction scenario and intention, and identifies a set of function-instruction matching sets with executable and reference lines, which is convenient for reuse in multiple apps and improves business conversion efficiency.

[0115] (3) Adjust the command output in real time according to user feedback, and build a closed-loop command generation method. The embodiment of the present invention analyzes the user's feedback on the command based on the interactive behavior of the APP, and adjusts the command in real time according to the feedback results, forming a closed loop of command output-user feedback-command tuning, which can quickly respond to user needs. Compared with traditional solutions, it has better flexibility and real-time response capabilities, and improves the quality of interaction. Through the real-time feedback mechanism, the model can identify errors and continuously learn and improve, improve the decision-making ability of the model, and ensure that the decision meets user needs.

[0116] The embodiment of the present invention is described below by a specific example: (1) When the user opens the App, a websocket link is established to transmit instructions and execution results in real time.

[0117] (2) Build a perception tree based on the user's interaction records. Starting from the root node of the perception tree, traverse all paths, filter out paths with a current completion degree of 0, and match the user's interaction needs based on the user's personal portrait and business judgment rules. Retrieve and match the corresponding touch point functions of the user through the knowledge base. For example, when a user clicks a certain function touch point on the "Rights" page and jumps to the "Global Link Benefits" page, and the user has Global Link benefits that have not been received, the corresponding function touch point will be queried and matched from the knowledge base based on the user's received benefits. For example, "Receive Global Link Travel Benefits" will be output as a business function touch point.

[0118] (3) Use "receiving global travel benefits" as the input of the human-computer interaction instruction matching model to filter out the set of instructions to be executed that meet the confidence and improvement requirements.

[0119] (4) Improve the corresponding candidate instruction set parameter information based on the user's profile information and context information. If the first instruction in the sorting is "broadcast audio", the user has unclaimed benefits. At this time, the corresponding words are retrieved from the knowledge base, such as "You have 4 benefits to be claimed, please claim them in time", and encapsulated in json format and sent to the App.

[0120] (5) The App receives the "play audio" command JSON data, parses it, and calls the system player to play it.

[0121] (6) Calculate the user's feedback score on the instruction based on the user's interactive behavior in the APP. Assuming the feedback score is 0.6, reselect the second-ranked instruction from the matched instructions to be executed, such as "pop-up display", and continue to send it to the App through the socket for instruction execution, and update the perception tree.

[0122] (7) Based on the user interaction behavior after the second instruction is issued, the feedback score is recalculated. If it is positive feedback, the process ends. If it is negative feedback, it returns to the demand perception module to execute the interactive perception demand matching output step, and executes the subsequent steps to re-output the instruction.

[0123] The above steps are repeated repeatedly until the user completes his / her usage requirements.

[0124] The interactive instruction construction device provided by the present invention is described below. The interactive instruction construction device described below and the interactive instruction construction method described above can be referenced to each other.

[0125] Figure 4 Schematic diagram of the structure of the interactive instruction construction device provided by the embodiment of the present invention. Figure 4 The embodiment of the present invention provides an interactive instruction construction device, which may specifically include the following modules: A perception tree construction module 410 is used to obtain historical interaction behavior information of the user and construct an interaction behavior perception tree; An interaction requirement determination module 420 is used to determine a plurality of candidate interaction requirements based on the interaction behavior perception tree, and determine the candidate interaction requirement that best matches the user profile as the user interaction requirement; An executable instruction determination module 430 is used to search the knowledge base based on the user interaction requirement, determine the business function touchpoint corresponding to the user interaction requirement, and determine the executable instruction set corresponding to the business function touchpoint; An instruction sorting and issuing module 440 is used to sort the instructions in the executable instruction set based on the user interaction context information, and issue instructions based on the executable instruction set after the instructions are sorted; The user feedback module 450 is used to issue new executable instructions if the user's feedback score for the currently issued executable instructions is less than a preset score threshold, until the user's feedback score for the new executable instructions is greater than the preset score threshold, so as to obtain the executable instructions that best match the user's actual needs.

[0126] The embodiment of the present invention constructs an interactive behavior perception tree by acquiring the user's historical interactive behavior information; based on the interactive behavior perception tree, multiple candidate interactive needs are determined, and the candidate interactive need that best matches the user portrait is determined as the user interactive need; based on the user interactive need, a knowledge base search is performed to determine the business function touchpoints corresponding to the user interactive need, and the executable instruction set corresponding to the business function touchpoints is determined; based on the user interactive context information, the instructions in the executable instruction set are sorted, and the instructions are issued based on the executable instruction set after the instructions are sorted; if the user's feedback score for the currently issued executable instruction is less than the preset score threshold, a new executable instruction is issued until the user's feedback score for the new executable instruction is greater than the preset score threshold, so as to obtain the executable instruction that best matches the user's real needs. The present invention can autonomously drive the execution of interactive instructions, form a feedback loop, support flexible adjustment at runtime, and enhance the adaptive ability in a changing environment.

[0127] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5As shown, the electronic device may include: a processor (processor) 510, a communication interface (Communications Interface) 520, a memory (memory) 530 and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call the logic instructions in the memory 530 to execute the interaction instruction construction method, which includes: obtaining the user's historical interaction behavior information and constructing an interaction behavior perception tree; based on the interaction behavior perception tree, determining multiple candidate interaction needs, and determining the candidate interaction need that best matches the user portrait as the user interaction need; performing a knowledge base search based on the user interaction need, determining the business function touchpoints corresponding to the user interaction need, and determining the executable instruction set corresponding to the business function touchpoints; based on the user interaction context information, sorting the instructions in the executable instruction set, and issuing instructions based on the executable instruction set after the instructions are sorted; if the user's feedback score for the currently issued executable instruction is less than a preset score threshold, issuing a new executable instruction until the user's feedback score for the new executable instruction is greater than the preset score threshold, so as to obtain the executable instruction that best matches the user's real needs.

[0128] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0129] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the interaction instruction construction method provided by the above-mentioned methods, the method comprising: obtaining the user's historical interaction behavior information and constructing an interaction behavior perception tree; based on the interaction behavior perception tree, determining multiple candidate interaction needs, and determining the candidate interaction need that best matches the user portrait as the user interaction need; performing a knowledge base search based on the user interaction need, determining the business function touchpoints corresponding to the user interaction need, and determining the executable instruction set corresponding to the business function touchpoints; based on the user interaction context information, sorting the instructions in the executable instruction set, and issuing instructions based on the executable instruction set after the instructions are sorted; if the user's feedback score for the currently issued executable instruction is less than a preset score threshold, issuing a new executable instruction until the user's feedback score for the new executable instruction is greater than the preset score threshold, so as to obtain the executable instruction that best matches the user's real needs.

[0130] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0131] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing an interactive instruction, characterized in that: include: Obtain the user's historical interaction behavior information and build an interaction behavior perception tree; Based on the interaction behavior perception tree, multiple candidate interaction requirements are determined, and the candidate interaction requirement that best matches the user profile is determined as the user interaction requirement; Perform a knowledge base search based on the user interaction requirement, determine the business function touchpoint corresponding to the user interaction requirement, and determine the executable instruction set corresponding to the business function touchpoint; Based on the user interaction context information, the instructions in the executable instruction set are sorted, and the instructions are issued based on the sorted executable instruction set; If the user's feedback score for the currently issued executable instruction is less than the preset score threshold, a new executable instruction is issued until the user's feedback score for the new executable instruction is greater than the preset score threshold to obtain the executable instruction that best matches the user's actual needs.

2. The interactive instruction construction method according to claim 1, characterized in that: The user's historical interaction behavior information includes the interaction instructions executed by the user on each navigation stack function page and the interaction results obtained by executing the interaction instructions; The step of obtaining the user's historical interaction behavior information and constructing an interaction behavior perception tree includes: The application window is used as a first-level node, and a plurality of navigation stack function pages are inserted as child nodes of the first-level node to obtain a plurality of second-level nodes; For each second-level node, an interaction instruction executed by the user on the navigation stack function page corresponding to the second-level node and an interaction result obtained by executing the interaction instruction are inserted as child nodes of the second-level node to obtain at least one third-level node, so as to obtain an initial interaction behavior perception tree; The initial interaction behavior perception tree is pruned to obtain an interaction behavior perception tree.

3. The interactive instruction construction method according to claim 1, characterized in that: The determining, based on the interaction behavior perception tree, a plurality of candidate interaction requirements, and determining the candidate interaction requirement that best matches the user profile as the user interaction requirement, includes: Traversing the interaction behavior perception tree, finding a path set whose completion degree is not 1; the path set whose completion degree is not 1 is used to represent multiple candidate interaction requirements; For each user portrait dimension, based on the user portrait dimension and the weight of the user portrait dimension under the corresponding candidate interaction requirement, calculate a weighted score of the user portrait dimension; The candidate interaction requirement corresponding to the user portrait dimension with the highest weighted score is determined as the user interaction requirement.

4. The interactive instruction construction method according to claim 1, characterized in that: The determining of the executable instruction set corresponding to the business function touchpoint includes: Perform word segmentation processing on the description texts in the business function instruction set, the business function contact instruction set, and the human-computer interaction instruction set, and obtain the business function instruction initial word segmentation result, the business function contact word segmentation result, and the human-computer interaction instruction word segmentation result respectively; Perform string fuzzy matching on the initial segmentation result of the business function instruction, the segmentation result of the business function contact point and the segmentation result of the human-computer interaction instruction, and remove irrelevant words in the initial segmentation result of the business function instruction based on the matching result to obtain the segmentation result of the business function instruction; Based on the business function instruction word segmentation results, construct multiple contact point-instruction item sets; Filter out the target contact-instruction item set that meets the threshold requirements of support, confidence and lift; Based on the target contact-instruction item set, an executable instruction set corresponding to the business function contact is determined.

5. The interactive instruction construction method according to claim 4, characterized in that: The step of constructing a plurality of contact point-instruction item sets based on the word segmentation result of the business function instruction includes: Based on the instructions whose occurrence times in the business function instruction segmentation results are greater than the minimum support threshold, constructing an instruction candidate 1-item set, and determining the instruction candidate 1-item set whose support is greater than or equal to the minimum support threshold as the instruction frequent 1-item set; Based on the business function contacts whose number of occurrences in the business function instruction word segmentation results is greater than the minimum support threshold, a contact candidate 1-item set is constructed, and the contact candidate 1-item set whose support is greater than or equal to the minimum support threshold is determined as a contact frequent 1-item set; Multiple instruction frequent 1-item sets and multiple contact frequent 1-item sets are arranged and combined to generate multiple contact-instruction item sets.

6. The interactive instruction construction method according to claim 4, characterized in that: The support, confidence and lift of the contact-instruction item set are calculated in the following way: For each contact-instruction item set, determining a target instruction frequent 1-item set and a target contact frequent 1-item set in the contact-instruction item set; Calculate the support of the target instruction frequent one-item set based on the number of business function instructions in the business function instruction segmentation results that include the target instruction frequent one-item set and the total number of business function instructions; Calculate the support of the target contact frequent one-item set based on the number of business function instructions in the business function instruction segmentation result that include the target contact frequent one-item set and the total number of business function instructions; Calculate the support of the contact-instruction item set based on the number of business function instructions that simultaneously include the target instruction frequent 1-item set and the target contact frequent 1-item set in the business function instruction segmentation result, and the total number of business function instructions; Calculating the confidence of the contact-instruction item set based on the support of the contact-instruction item set and the support of the instruction 1 item set; The lift of the contact-instruction itemset is calculated based on the support of the contact-instruction itemset, the support of the target instruction frequent 1-itemset and the support of the target contact frequent 1-itemset.

7. The interactive instruction construction method according to claim 1, characterized in that: The user's feedback score for the currently issued executable instruction is calculated in the following way: Obtain the number of executions of commands adopted by users, the length of the command execution path, the command execution time, and the length of time that users interact with the functional touchpoints; A weighted sum calculation is performed based on the weight coefficients corresponding to the number of execution times of the instructions adopted by the user, the length of the instruction execution path, the instruction execution time, and the interaction time of the user at the functional contact point, so as to obtain the user's feedback score for the currently issued executable instruction.

8. An interactive instruction construction device, characterized in that: include: The perception tree construction module is used to obtain the user's historical interaction behavior information and construct the interaction behavior perception tree; An interaction requirement determination module, configured to determine a plurality of candidate interaction requirements based on the interaction behavior perception tree, and determine the candidate interaction requirement that best matches the user profile as the user interaction requirement; An executable instruction determination module, used to perform a knowledge base search based on the user interaction requirement, determine the business function touchpoint corresponding to the user interaction requirement, and determine the executable instruction set corresponding to the business function touchpoint; An instruction sorting and issuing module, used for sorting the instructions in the executable instruction set based on the user interaction context information, and issuing instructions based on the executable instruction set after the instruction sorting; The user feedback module is used to issue new executable instructions if the user's feedback score for the currently issued executable instructions is less than a preset score threshold, until the user's feedback score for the new executable instructions is greater than the preset score threshold, so as to obtain the executable instructions that best match the user's actual needs.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the interactive instruction construction method as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the interactive instruction construction method as described in any one of claims 1 to 7 is implemented.

Images