Instruction response method and device, equipment and medium

Abstract

The invention discloses an instruction response method and device, equipment and a medium. The instruction response method comprises the steps that text content carried in an instruction input by a user is acquired; extracting a target keyword corresponding to the text content; the target keyword is input into a classification model for matching, a first classification label matched with the target keyword is obtained, the classification model comprises a tree created based on a page of a target system, and nodes in the tree are classification labels created based on page information of the page; and pushing a first page link corresponding to the first classification tag to a user side as a response result of the user input instruction. Therefore, the response speed can be improved, and the user experience is improved.

Description

Technical Field

[0001] The present application relates to the technical field of instruction response, and in particular to an instruction response method, device, equipment and medium. Background Art

[0002] With the development of computer technology, various industries usually use digital business systems to implement their business. At present, business systems tend to be diversified, and product functions are numerous. Users need to perform multiple steps to find the functions they want, which affects the user experience. Therefore, how to improve the response speed and thus improve the user experience is an urgent problem to be solved. Summary of the invention

[0003] In view of this, the purpose of this application is to provide a command response method, device, equipment and medium, which can improve the response speed and thus improve the user experience. The specific scheme is as follows:

[0004] In a first aspect, the present application discloses a command response method, comprising:

[0005] Get the text content carried in the user input command;

[0006] Extracting target keywords corresponding to the text content;

[0007] Inputting the target keyword into a classification model for matching, and obtaining a first classification label matching the target keyword, wherein the classification model includes a tree created based on a page of the target system, and a node in the tree is a classification label created based on page information of the page;

[0008] The first page link corresponding to the first classification tag is pushed to the user terminal as a response result of the user input instruction.

[0009] Optionally, extracting target keywords corresponding to the text content includes:

[0010] The text content is segmented using a natural language understanding model to obtain a segmentation result, and target keywords in the segmentation result are extracted based on attributes of the segmentation result.

[0011] Optionally, the attributes include nouns, verbs and adjectives, and the target keywords include screening conditions, operation objects and operation actions.

[0012] Optionally, extracting target keywords corresponding to the text content includes:

[0013] Accurately matching the text content with a preset database of the target system, the preset database containing page information of pages in the target system;

[0014] If there is no precise match to the page information, fuzzy matching is performed, and when multiple fuzzy matching results are obtained, target keywords corresponding to the text content are extracted based on the multiple fuzzy matching results.

[0015] Optionally, inputting the target keyword into a classification model for matching includes:

[0016] Input the target keyword into the classification model, and match the target keyword using the operation process tree and / or system information structure tree corresponding to the target system;

[0017] Among them, the operation process tree is a tree created based on the operation process, the nodes in the operation process tree are the classification labels corresponding to the nodes in the operation process, and the system information structure tree is a tree created based on the page hierarchy, and the nodes in the system information structure tree are the classification labels of the pages corresponding to the nodes.

[0018] Optionally, after pushing the first page link corresponding to the first classification tag to the user terminal as a response result of the user input instruction, the method further includes:

[0019] When the page corresponding to the first classification label ends displaying, a second page link corresponding to the second classification label is pushed to the user terminal;

[0020] Among them, the second classification label is the next node corresponding to the first classification label in the operation process tree, and / or the related node of the first classification label in the system information structure tree, and the related node is the classification label of the jump page corresponding to the page element in the page corresponding to the first classification label.

[0021] Optionally, the step of pushing the first page link corresponding to the first classification tag as a response result to the user terminal includes:

[0022] Determining a third page link based on user information corresponding to the user input instruction; wherein the user information includes one or more of user role information, user authority information, and user behavior preference information;

[0023] The first page link and the third page link corresponding to the first classification tag are pushed to the user end as response results.

[0024] In a second aspect, the present application discloses a command response device, comprising:

[0025] The acquisition module is used to obtain the text content carried in the user input instruction;

[0026] An extraction module, used to extract target keywords corresponding to the text content;

[0027] A matching module, used for inputting the target keyword into a classification model for matching, and obtaining a first classification label matching the target keyword, wherein the classification model includes a tree created based on a page of the target system, and a node in the tree is a classification label created based on page information of the page;

[0028] A push module is used to push the first page link corresponding to the first classification tag to the user terminal as a response result of the user input instruction.

[0029] In a third aspect, the present application discloses an electronic device, including a memory and a processor, wherein:

[0030] The memory is used to store the computer program;

[0031] The processor is used to execute the computer program to implement the aforementioned instruction response method.

[0032] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program, wherein the computer program implements the aforementioned instruction response method when executed by a processor.

[0033] It can be seen from the above scheme that the present application provides an instruction response method, including: obtaining text content carried in a user input instruction; extracting target keywords corresponding to the text content; inputting the target keywords into a classification model for matching, and obtaining a first classification label that matches the target keyword, wherein the classification model includes a tree created based on a page of a target system, and the nodes in the tree are classification labels created based on the page information of the page; and pushing a first page link corresponding to the first classification label to the user end as a response result of the user input instruction.

[0034] It can be seen that the beneficial effects of the present application are: extracting the target keywords corresponding to the text content carried in the user input instructions, using the tree created based on the page of the target system in the classification model to match the keywords, obtaining the matching classification labels, and pushing the corresponding page links to the user end. In this way, the user does not need to search for pages step by step, which improves the response speed and thus improves the user experience.

[0035] Correspondingly, the instruction response device, equipment and readable storage medium provided by the present application also have the above-mentioned technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0037] Figure 1 A flow chart of a command response method provided in an embodiment of the present application;

[0038] Figure 2 A schematic diagram of an instruction-clear type provided for an embodiment of the present application;

[0039] Figure 3 Another instruction-explicit schematic diagram provided for an embodiment of the present application;

[0040] Figure 4 A schematic diagram of a user terminal display with clear instructions provided in an embodiment of the present application;

[0041] Figure 5 Another schematic diagram of a user terminal display with clear instructions provided in an embodiment of the present application;

[0042] Figure 6 A schematic diagram of a semi-explicit instruction provided in an embodiment of the present application;

[0043] Figure 7 A schematic diagram of a semi-explicit user terminal display of instructions provided in an embodiment of the present application;

[0044] Figure 8 A schematic diagram of an unclear instruction type provided in an embodiment of the present application;

[0045] Fig. 9 A schematic diagram of a user terminal display with unclear instructions provided in an embodiment of the present application;

[0046] Fig.10 A command response flow chart provided for an embodiment of the present application;

[0047] Fig.11 A schematic diagram of guessing what you want provided in an embodiment of the present application;

[0048] Fig.12 A schematic diagram of prompting a user to input text provided in an embodiment of the present application;

[0049] Fig.13 A guess what you want - change schematic diagram provided in an embodiment of the present application;

[0050] Fig.14 A schematic diagram of an approximate instruction provided in an embodiment of the present application;

[0051] Fig.15 A schematic diagram of searching for information within an application provided in an embodiment of the present application;

[0052] Fig.16 A schematic diagram of filling in information in an application provided in an embodiment of the present application;

[0053] Fig.17 A schematic diagram of a wiki process provided for an embodiment of the present application;

[0054] Fig.18 Another wiki process diagram provided for an embodiment of the present application;

[0055] Fig.19 A schematic diagram of the structure of a command response device provided in an embodiment of the present application;

[0056] Fig. 20 A structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0058] See also Figure 1 As shown, the embodiment of the present application discloses a command response method, including:

[0059] Step S11: Obtain the text content carried in the user input instruction.

[0060] The user input instruction can be understood as an instruction sent by the user end and obtained by the server end, carrying the user input text content. For example, the user input text content is "I want to view the statistics of the store time of all supervisors today". The instruction can be encapsulated based on the communication protocol between the server end and the user end. The instruction can be a unique command corresponding to the last level independent and non-repetitive page in the system module.

[0061] Step S12: extracting target keywords corresponding to the text content.

[0062] In an optional implementation, the present embodiment can segment the text content, match the segmentation result with a preset word library, and obtain the target keyword. The preset word library contains keywords related to the business.

[0063] Furthermore, in an optional implementation, the text content may be segmented using a natural language understanding model to obtain a segmentation result, and target keywords in the segmentation result may be extracted based on the attributes of the segmentation result. The natural language understanding model may be a deep learning model trained based on a text training set, capable of segmenting words and extracting target keywords based on attributes.

[0064] Among them, the attributes may include nouns, verbs and adjectives, and the target keywords may include screening conditions, operation objects and operation actions. That is, the operation objects can be extracted based on nouns, the operation actions can be extracted based on verbs, and the screening conditions can be extracted based on adjectives. For example, nouns can be used to match system module names, stores, personnel and other entity names, including to-do lists, task names, etc., verbs can be used to match operation actions within the system, including viewing, creating, processing, etc., and adjectives such as dates are used to match the search scope, such as "today's". A preset vocabulary can be created, and the preset vocabulary can be matched according to the part of speech. The preset vocabulary is classified according to the part of speech.

[0065] In another optional implementation, the text content can be precisely matched with a preset database of the target system, and the preset database contains page information of pages in the target system; if the page information is not precisely matched, fuzzy matching is performed, and when multiple fuzzy matching results are obtained, the target keywords corresponding to the text content are extracted based on the multiple fuzzy matching results. Among them, the fuzzy matching result can be understood as the page information matched from the preset database. The page information can be understood as information containing the semantics corresponding to the page, such as the page name. Precise matching can be understood as the complete consistency of the text content and the page information, and fuzzy matching can be understood as the partial consistency of the text content and the page information. For example, fuzzy matching rules can be used for matching, and the fuzzy matching rules are AD**CD, and the * character can match any character, and non-* characters need to be consistent.

[0066] In this embodiment, if the page information is accurately matched, the page link of the accurately matched page information is pushed to the user end as the response result of the user input instruction. If the fuzzy matching result of the fuzzy matching is one, the page link corresponding to the fuzzy matching result is pushed to the user end as the response result of the user input instruction.

[0067] Step S13: input the target keyword into the classification model for matching, and obtain a first classification label matching the target keyword, wherein the classification model includes a tree created based on the page of the target system, and the nodes in the tree are classification labels created based on the page information of the page.

[0068] In this embodiment, the page name can be preset to obtain a classification label. The page name can be segmented. For example, "Create a to-do list" is a complete page name, and the classification labels are "Create" and "To-do list". The similarity between the target keyword and the classification label in the tree is calculated to obtain a first classification label that matches the target keyword, wherein the similarity is a semantic similarity. In an optional implementation, the classification label with the highest similarity can be determined as the first classification label that matches the keyword. In another optional implementation, the classification label that matches the target keyword can be a classification label with a similarity higher than a preset threshold. If the number of classification labels with a similarity higher than the preset threshold is greater than a preset number, a preset number of classification labels with the highest similarity can be taken as the first classification label. The preset number can be set according to actual needs, for example, 1, 2, 3, etc.

[0069] In an optional implementation, the target keyword can be input into a classification model, and the target keyword can be matched using an operation process tree and / or a system information structure tree corresponding to the target system; wherein the operation process tree is a tree created based on the operation process, and the nodes in the operation process tree are classification labels corresponding to the nodes in the operation process; and the system information structure tree is a tree created based on a page hierarchy, and the nodes in the system information structure tree are classification labels of the pages corresponding to the nodes.

[0070] It is understandable that the present embodiment can create an operation process tree and a system information structure tree corresponding to the target system, and can create a corresponding operation process tree and / or system information structure tree for each functional module of the target system. The operation process tree is created based on the nodes in the operation process. The system information structure tree is a tree created based on the page hierarchy, for example, the classification label corresponding to the first-level page is the root node, and the second-level page under the first-level page is the child node of the root node. The system information structure tree can be created for each page of the functional module according to the page hierarchy.

[0071] Furthermore, in an optional implementation, an index can be constructed for the operation process tree and the system information structure tree, thereby speeding up the search. For example, an index can be created for each node in the tree, containing key information of the node, such as node type, attributes, etc., so that the relevant tree or node can be quickly located directly through the index during matching. In another optional implementation, hierarchical matching can be adopted: for a tree with a hierarchical structure, a hierarchical matching strategy can be adopted, first matching the upper nodes, and then deciding whether to match the lower nodes based on the matching results of the upper nodes, so as to reduce unnecessary matching operations.

[0072] Step S14: Pushing the first page link corresponding to the first classification tag to the user terminal as a response result of the user input instruction.

[0073] The response result can be pushed to the user end in the form of a tab sheet for display, which is convenient for the user to click. The user can enter the corresponding page by clicking the tab sheet.

[0074] In an optional implementation, a third page link can be determined based on user information corresponding to the user input instruction; wherein the user information includes one or more of user role information, user authority information, and user behavior preference information; and the first page link corresponding to the first classification label and the third page link are pushed to the user end as response results.

[0075] That is, this embodiment can push multiple page links to the user based on the user information for selection. The user behavior preference information can be frequency information, for example, the user has opened the same page multiple times.

[0076] Furthermore, after pushing the first page link corresponding to the first classification tag to the user end as a response result of the user input instruction, it also includes: when the page corresponding to the first classification tag finishes displaying, pushing the second page link corresponding to the second classification tag to the user end; wherein, the second classification tag is the next node corresponding to the first classification tag in the operation process tree, and / or, the related node of the first classification tag in the system information structure tree, and the related node is the classification tag of the jump page corresponding to the page element in the page corresponding to the first classification tag.

[0077] It is understandable that if the first classification tag is a classification tag in the operation process tree, after the first page link corresponding to the first classification tag is pushed to the user end as the response result of the user input instruction, when the page corresponding to the first classification tag ends displaying, the second page link corresponding to the second classification tag is pushed to the user end; wherein, the second classification tag is the next node corresponding to the first classification tag in the operation process tree. That is, when the user exits the page corresponding to the first classification tag, the page link corresponding to the next node can be pushed to the user end based on the operation process tree. It is understandable that when the user finishes the operation of a process node, it is likely to operate the next node. By actively pushing the page link of the next node, the user operation steps can be reduced, thereby improving the user experience. Taking a business as an example, after a new task is created, the user wants to view it, then the next process node of the new task is to view the task. If the first classification tag is a classification tag in the system information structure tree, the target element in the page corresponding to the first classification tag can be determined, and the target element can be a button for page jump. The classification tag of the adjustment page is determined based on the target element to obtain the second classification tag, and the second page link corresponding to the second classification tag is pushed to the user end.

[0078] It can be seen that the embodiment of the present application extracts the target keywords corresponding to the text content carried in the user input instruction, uses the tree created based on the page of the target system in the classification model to match the keywords, obtains the matching classification label, and pushes the corresponding page link to the user end. In this way, the user does not need to search for the page step by step, which improves the response speed and thus improves the user experience.

[0079] Below, taking a certain business system as an example, the instruction response solution provided by this application is further explained:

[0080] The user inputs text, and the NLU model is used to extract the business-related filtering conditions, operation actions and operation objects in the user-input sentence; the role of the NLU model includes but is not limited to parsing text, word segmentation output, synonym processing, etc.; the NLU model can segment the text and match different objects according to the attributes of each word segmentation result (such as noun matching system module name / store, personnel and other entity names, including to-do lists, task names, etc., verb matching system operation actions, including viewing, creating, processing, etc., and others such as date are used to match the search scope, that is, adjectives corresponding to the filtering conditions. When the NLU model extracts keywords, it can use the knowledge graph of store-related content, specifically a preset word library composed of relevant words in the knowledge graph. The knowledge graph is a form of information organization. In this embodiment, a corpus can be constructed based on business-related data. The corpus is constructed in the form of a knowledge graph, which includes the aforementioned preset database and the aforementioned preset word library as well as corporate nouns, store names, personnel names, organizational structures, role positions and other entity information.

[0081] Based on the business-related screening conditions, operation actions and objects, they are input into the classification model to determine whether there is a rule match. Specifically, the process tree of functional operations in the system and the system information structure tree can be matched according to the definition rules of the expert system. If there is a rule match, the tab sheet is output according to the corresponding definition of the rule.

[0082] The classification model stores the expert system rule set (i.e., the operation process tree and the system information structure tree). Through similarity calculation (the matching similarity output here can be in the form of a list, which supports sorting from high to low), the most similar classification label is output, the business API corresponding to the classification label is obtained, and the corresponding URL is returned as a response result through the business API and displayed on the page.

[0083] In some embodiments, the training corpus of the NLU text model is a knowledge base such as its own business documents, subdivided industry corpora, etc., and also includes a database storing user behavior and user preferences, etc.; the classification model stores a structure tree of each business module. If a new business module needs to be added to this method for application, the steps also include: adding the new business module to the corresponding position based on the system information structure tree (whose structure includes primary classification labels and secondary classification labels) and adding a new label and its corresponding URL address. A corresponding operation process tree can also be added.

[0084] The user confirms whether the tab sheet meets the user's expectations. If so, the user clicks on the target option, calls the corresponding business API, and the corresponding target page can be displayed on the front end.

[0085] Furthermore, in some embodiments, after the call of the business module is completed, a tab sheet corresponding to the next step of the process node where the current call is located can be provided and pushed to the user for selection. It should be pointed out that the relevant process operations of the business module are also structured, and the same as the information structure of the system, the next node of the operation process of the tree structure is the tab sheet corresponding to the next step of the process node where the current call is located.

[0086] In some embodiments, the content generation of the tab sheet includes but is not limited to the output content according to the rule definition and the classification model, and can also be combined with the user role / authority / subscription preference to output a further refined option list card for the user to confirm. That is, the command recognition and execution can be performed concurrently, and common concurrent processing includes user authority verification or user identification and command execution.

[0087] If the option sheet does not meet the user's expectations, you can use guided inquiries and other methods until you provide similar options that meet the user's expectations (the generation logic of similar options: according to the above rules, the matching options from high to low similarity are matched, and the option with the highest matching degree is displayed to the user first. If the option sheet does not meet the user's expectations, other options will be displayed next time.)

[0088] Furthermore, this call can be recorded in the background and used as a material for generating high-frequency recommended options. In this way, the corresponding module can be intelligently pushed to the homepage (i.e. displayed to the user on the front end) according to the frequency of the module corresponding to the user input command.

[0089] In an optional implementation, a precise search may be performed first (using input to filter whether there are pages that fully match the user input), and then a fuzzy search may be performed (using user input to partially match, whether there are pages in the page configuration that match the user input, partial match means: I want to read the task == reading task, which is continuous, and can be understood as a partial match, the task I want to read != reading task, there is no partial match. If a fuzzy match is found, it is returned directly. If multiple fuzzy matches are found, the result set of the fuzzy match is used for NLU reasoning, and the classification model is used for matching.

[0090] This embodiment provides a one-stop, simple and fast workbench module for users by providing a conversational interactive interface, which allows users to search for information across applications, and realize functions or view data results that were originally required to be searched and clicked layer by layer. Users can directly input their own needs, and the system responds quickly, and can quickly find the functions they want in the system and accurately respond to user needs. After the user's question is uploaded to the model, it is answered, triggering "regenerate" and "edit question" as options for the next process, which are placed at the bottom of the conversation page for users to choose.

[0091] In this embodiment, the text input by the user can be divided into three categories: clear instruction type, semi-clear instruction type, and unclear instruction type. Clear instruction type is an instruction that can be accurately matched, semi-clear instruction type is an instruction with a fuzzy matching result of one, and unclear instruction type is an instruction that needs to be matched with a classification model.

[0092] See also Figure 2 As shown, Figure 2 A schematic diagram of a clear instruction type provided in an embodiment of the present application. The text entered by the user is "I want to view the statistical ranking of the time spent in the store by all supervisors today". The object is the user (i.e. supervisor) using the service, and the time spent in the store by the user after arriving at the store is counted. Figure 3 As shown, Figure 3 Another instruction-specific schematic diagram provided in an embodiment of the present application. The user inputs text: "I want to check today's store inspection tasks." Figure 4 As shown, Figure 4 A schematic diagram of a user terminal display with clear instructions provided in an embodiment of the present application. Figure 5 As shown, Figure 5 Another user-side display schematic diagram of a clear instruction type provided in an embodiment of the present application. This embodiment performs semantic analysis on the text input by the user, corresponds to a clear App-side business page, outputs the relevant business interface, and provides page options related to the page information structure or process for the user to choose from.

[0093] See also Figure 6 As shown, Figure 6 A schematic diagram of a semi-explicit instruction provided in an embodiment of the present application. The text entered by the user is "I want to view today's passenger flow report". The text entered by the user is semantically parsed, corresponding to the corresponding APP-side module, and the business interface under the output module is used as an option. Combined with the user role and authority, relevant page options are provided to the user as a choice. See Figure 7 As shown, Figure 7 A schematic diagram of a semi-explicit instruction user terminal display provided in an embodiment of the present application.

[0094] See also Figure 8 As shown, Figure 8 A schematic diagram of an undefined instruction type provided in an embodiment of the present application. The text input by the user is "How is the store inspection going?" The user input text is semantically parsed, at least one noun is entity recognized, and the relevant content of the noun in the knowledge graph is output as an option. Combined with the user's subscription and behavior preferences, relevant page options are provided to the user as a choice. Fig. 9 As shown, Fig. 9A schematic diagram of a user terminal display with unclear instructions provided in an embodiment of the present application. From top to bottom: 1. The highest level module page of the recognition word, 2. Store inspection related functions frequently used by users, 3. The functions most frequently used by the same role in the system, 4. New functions in the system, 5. Change a batch of questions.

[0095] Furthermore, this embodiment can search for information across applications and complete some internal operations of the application. After completion, user confirmation is required. For example, a user wants to view the store inspection record of store A today. The specific page is the store inspection record. There are filtering options on this page. The parameters in the options can select [Today] and [Store A]. You can use the workbench to input: I want to view the store inspection record of store A today. On the opened store inspection record page, the filtering options have helped the user select [Today] and [Store A].

[0096] In addition, when facing general questions (i.e. questions for which no information can be retrieved in the system), the user is prompted whether to use the general model to upload and answer. For generalized general question answers, the large language model can generate more extensive language knowledge by pre-training on large-scale text data without annotations through massive data; the NLU model focuses on a specific task or a set of tasks, and the goal is to accurately complete the task. It uses task-oriented supervised learning, such as classification, sequence labeling, etc., which is more advantageous in terms of accuracy and pertinence in specific fields; the NLU model can achieve high accuracy when trained for specific tasks, and can be fine-tuned using knowledge documents of SaaS in the retail field; compared with the general model, the self-built corpus for vertical SaaS systems in the retail management field is more accurate; the knowledge graphs such as stores, personnel, and system fields are constantly updated to improve the accuracy and controllability; OpenNLU divides NLU tasks into two categories: extraction and classification. The extraction task goal is to extract some fragments from a given text, such as entity recognition, event extraction, keyword extraction, etc. The classification task goal is to classify a given text according to a given category system, such as topic classification, sentiment classification, intent classification, etc.

[0097] For further information, see Fig.10 As shown, Fig.10A flow chart of command response disclosed in the embodiment of the present application. Start entering the page, and display hello [Card C] and guess what you want [Card B] on the user side, and guess what you want [Card B] is sorted according to the frequency of user use. In response to the user clicking the change button, guess what you want [Card B] is displayed on the user side, sorted according to the frequency of user use, which is different from the content displayed before the change, for example, options with lower frequency can be displayed. In response to the user clicking the business option, a personalized configuration check is performed. If the page-enterprise corresponding configuration is not opened, a pop-up window prompts the configuration, that is, if the page is closed in the enterprise's system configuration, the user cannot reach the page. If the personalized configuration check passes, a permission check is performed. If the user does not have access rights to the page, a pop-up window prompts that there is no permission. If the user permission verification is passed, a version update verification is performed. The user has not updated the latest iteration of the module command, and a pop-up window prompts the user to update. That is, if the app version used by the user is not the latest, the jump page will not be supported. If the version update verification is passed, enter the business page and click the upper left corner or swipe left to return. In response to the user clicking "None of the above", the input text card [Card C] is displayed on the user side. The database is first accurately searched and matched. If the accurate match is successful, the corresponding business API [Card A] is displayed on the user side. If the accurate match is not successful, fuzzy matching is performed to match whether the option pool in the total option pool (that is, the database) is unique. If so, the corresponding business API [Card A] is displayed on the user side. If it is not unique, based on the user's choice of whether to enter the text search, if the user chooses to enter the text search, a text answer is given based on the knowledge base built on the wiki document, and the bottom options are triggered, including regenerating and editing the question. If the user does not choose to enter the text search, the NLU full data set reasoning is performed, that is, the classification model is used for matching, sorted by semantic similarity, and the guess you want [Card B] is obtained. Among them, when the text entered by the user fails to match the module command in the existing database, that is, when a single specific business card cannot be matched, the card pops up: "No relevant business was retrieved. Do you want to continue searching using the system's own knowledge base?" If the user selects [Cancel], the original NLU full data set reasoning process will be entered, and the cards you guessed you wanted will be displayed in descending order of semantic similarity; if the user selects [Confirm], the question will be uploaded to the knowledge base built by the expert system document, and the knowledge base will generate a text answer. The knowledge base built by wiki documents can be understood as a database built by wiki documents related to the business, containing business-related knowledge, and text answers are generated based on this knowledge base.

[0098] See also Fig.11 As shown, Fig.11 A schematic diagram of guessing what you want is provided in the embodiment of the present application. Fig.12 As shown, Fig.12A schematic diagram of prompting a user to input text provided in an embodiment of the present application. Fig.13 As shown, Fig.13 A Guess What You Want - Change Schematic Diagram provided for an embodiment of the present application. This embodiment can combine user roles and permission settings, as well as common operations in the recent period, to provide common options that users may need. When it does not meet user expectations, a guided inquiry method is provided to optimize the experience. For example, prompt the user to enter in a text input box.

[0099] In addition, this embodiment can also display content corresponding to instructions similar to the user input instructions on the user terminal, see Fig.14 As shown, Fig.14 A schematic diagram of an approximate instruction provided in an embodiment of the present application.

[0100] Furthermore, the content included in the user input instruction may be search information, and in response to the user input instruction, corresponding information is displayed on the user terminal. Fig.15 As shown, Fig.15 A schematic diagram of searching for information within an application provided in an embodiment of the present application.

[0101] Furthermore, in this embodiment, in response to a user filling operation, the application program can fill in information. Fig.16 As shown, Fig.16 A schematic diagram of filling in information in an application provided in an embodiment of the present application.

[0102] See also Fig.17 As shown, Fig.17 A schematic diagram of a wiki process provided by an embodiment of the present application. The user enters a question, and the answer to the question is entered based on the wiki knowledge base and displayed on the front end. Fig.18 As shown, Fig.18 Another wiki flow diagram provided for an embodiment of the present application. In response to a user clicking a preset button, the preset button is to edit the question source, and the user input question is displayed in the question input box for editing. This embodiment is based on a knowledge base built on wiki documents related to the system business, and a text answer is generated by the knowledge base.

[0103] This embodiment can provide users with a one-stop, simple and fast workbench module, search for information across applications, and realize functions or view data results that originally required searching and clicking and jumping layer by layer. Provide users with the functions they want to use accurately and efficiently, and complete the operation step by step according to the prompts to ensure that users can accurately obtain the desired business interface. For the ever-expanding system content such as store names, personnel names, and field names, an open corpus can be established, and the knowledge graph can be continuously updated as part of the system's built-in knowledge base to provide regular updated data support for system user queries.

[0104] For further information, see Fig.19 As shown, this embodiment discloses a command response device, including:

[0105] The acquisition module 11 is used to acquire the text content carried in the user input instruction;

[0106] An extraction module 12, used to extract target keywords corresponding to the text content;

[0107] A matching module 13 is used to input the target keyword into a classification model for matching, and obtain a first classification label that matches the target keyword, wherein the classification model includes a tree created based on a page of the target system, and a node in the tree is a classification label created based on page information of the page;

[0108] The push module 14 is used to push the first page link corresponding to the first classification tag to the user terminal as a response result of the user input instruction.

[0109] In an optional implementation, the extraction module 12 is specifically configured to segment the text content using a natural language understanding model to obtain segmentation results, and extract target keywords from the segmentation results based on attributes of the segmentation results.

[0110] The attributes may include nouns, verbs, and adjectives, and the target keywords may include screening conditions, operation objects, and operation actions.

[0111] In another optional embodiment, the extraction module 12 is specifically used to accurately match the text content with a preset database of the target system, and the preset database contains page information of the pages in the target system; if the page information is not accurately matched, fuzzy matching is performed, and when multiple fuzzy matching results are obtained, the target keywords corresponding to the text content are extracted based on the multiple fuzzy matching results.

[0112] In an optional embodiment, the matching module 13 is specifically used to input the target keyword into the classification model, and match the target keyword using the operation process tree and / or system information structure tree corresponding to the target system; wherein the operation process tree is a tree created based on the operation process, and the nodes in the operation process tree are the classification labels corresponding to the nodes in the operation process; the system information structure tree is a tree created based on the page hierarchy, and the nodes in the system information structure tree are the classification labels of the pages corresponding to the nodes.

[0113] In an optional embodiment, the push module 14 is also used to, if the first classification label is a classification label in the operation process tree, after the first page link corresponding to the first classification label is pushed to the user end as a response result of the user input instruction, when the page corresponding to the first classification label finishes displaying, push the second page link corresponding to the second classification label to the user end; wherein, the second classification label is the next node corresponding to the first classification label in the operation process tree, and / or, the related node of the first classification label in the system information structure tree, and the related node is the classification label of the jump page corresponding to the page element in the page corresponding to the first classification label.

[0114] In an optional embodiment, the push module 14 is specifically used to determine the third page link based on the user information corresponding to the user input instruction; wherein the user information includes one or more of user role information, user authority information, and user behavior preference information; and the first page link corresponding to the first classification label and the third page link are pushed to the user end as response results.

[0115] It can be seen that the embodiment of the present application extracts the target keywords corresponding to the text content carried in the user input instruction, uses the tree created based on the page of the target system in the classification model to match the keywords, obtains the matching classification label, and pushes the corresponding page link to the user end. In this way, the user does not need to search for the page step by step, which improves the response speed and thus improves the user experience.

[0116] See also Fig. 20 As shown, an embodiment of the present application discloses an electronic device 20, including a processor 21 and a memory 22; wherein the memory 22 is used to store a computer program; the processor 21 is used to execute the computer program, and the instruction response method disclosed in the above embodiment.

[0117] For the specific process of the above-mentioned instruction response method, reference may be made to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0118] Furthermore, the memory 22 as a carrier for storing resources may be a read-only memory, a random access memory, a magnetic disk or an optical disk, etc., and the storage method may be temporary storage or permanent storage.

[0119] In addition, the electronic device 20 also includes a power supply 23, a communication interface 24, an input / output interface 25 and a communication bus 26; wherein the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and an external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input / output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0120] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the instruction response method disclosed in the aforementioned embodiment.

[0121] For the specific process of the above-mentioned instruction response method, reference may be made to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0122] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0123] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0124] The above is a detailed introduction to a command response method, device, equipment and medium provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A command response method, characterized in that: include: Get the text content carried in the user input command; Extracting target keywords corresponding to the text content; Inputting the target keyword into a classification model for matching, and obtaining a first classification label matching the target keyword, wherein the classification model includes a tree created based on a page of the target system, and nodes in the tree are classification labels created based on page information of the page; The first page link corresponding to the first classification tag is pushed to the user terminal as a response result of the user input instruction.

2. The instruction response method according to claim 1, characterized in that: The extracting target keywords corresponding to the text content includes: The text content is segmented using a natural language understanding model to obtain a segmentation result, and target keywords in the segmentation result are extracted based on attributes of the segmentation result.

3. The instruction response method according to claim 2, characterized in that: The attributes include nouns, verbs and adjectives, and the target keywords include screening conditions, operation objects and operation actions.

4. The instruction response method according to claim 1, characterized in that: The extracting target keywords corresponding to the text content includes: Accurately matching the text content with a preset database of the target system, the preset database containing page information of pages in the target system; If there is no precise match to the page information, fuzzy matching is performed, and when multiple fuzzy matching results are obtained, target keywords corresponding to the text content are extracted based on the multiple fuzzy matching results.

5. The instruction response method according to claim 1, characterized in that: The step of inputting the target keyword into a classification model for matching includes: Input the target keyword into the classification model, and match the target keyword using the operation process tree and / or system information structure tree corresponding to the target system; Among them, the operation process tree is a tree created based on the operation process, the nodes in the operation process tree are the classification labels corresponding to the nodes in the operation process, and the system information structure tree is a tree created based on the page hierarchy, and the nodes in the system information structure tree are the classification labels of the pages corresponding to the nodes.

6. The instruction response method according to claim 5, characterized in that: After pushing the first page link corresponding to the first classification tag as a response result of the user input instruction to the user terminal, the method further includes: When the page corresponding to the first classification label ends displaying, a second page link corresponding to the second classification label is pushed to the user terminal; Among them, the second classification label is the next node corresponding to the first classification label in the operation process tree, and / or the related node of the first classification label in the system information structure tree, and the related node is the classification label of the jump page corresponding to the page element in the page corresponding to the first classification label.

7. The instruction response method according to any one of claims 1 to 6, characterized in that: The step of pushing the first page link corresponding to the first classification tag as a response result to the user terminal includes: Determining a third page link based on user information corresponding to the user input instruction; wherein the user information includes one or more of user role information, user authority information, and user behavior preference information; The first page link and the third page link corresponding to the first classification tag are pushed to the user end as response results.

8. A command response device, characterized in that: include: The acquisition module is used to obtain the text content carried in the user input instruction; An extraction module, used to extract target keywords corresponding to the text content; A matching module, used for inputting the target keyword into a classification model for matching, and obtaining a first classification label matching the target keyword, wherein the classification model includes a tree created based on a page of the target system, and a node in the tree is a classification label created based on page information of the page; A push module is used to push the first page link corresponding to the first classification tag to the user terminal as a response result of the user input instruction.

9. An electronic device, characterized in that: comprising a memory and a processor, wherein: The memory is used to store the computer program; The processor is used to execute the computer program to implement the instruction response method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the instruction response method according to any one of claims 1 to 7 is implemented.

Images