Machine learning tools for navigating conversation flows
Navigating the dialogue flow through intelligent robot programs, analyzing user intentions and performing enterprise data retrieval and batch processing, solving the problem of insufficient interaction of existing AI systems and improving user experience and data processing efficiency in enterprise applications.
Patent Information
- Application Number
- CN201980009117.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2019-04-18
- Filing Date
- 2019-10-04
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2040-01-22
Smart Images

Figure CN111630506B_ABST
Abstract
Description
[0001] Cross-reference to related applications
[0002] This application claims priority to U.S. Provisional Patent Application Serial No. 62 / 748,929, filed on October 22, 2018, the disclosure of which is incorporated herein by reference. Technical Field
[0003] Embodiments of the present disclosure generally relate to a machine learning tool for navigating a conversation flow. Background Art
[0004] The implementation of artificial intelligence ("AI") has produced tangible benefits, ranging from supervised learning-based predictions to facial recognition. However, challenges surrounding AI remain, such as implementation-specific issues that often vary between different systems. For example, human-computer interaction with AI systems still leaves much to be desired. Additionally, it can be difficult to implement AI solutions that produce valuable results (e.g., results that have specific value to a business, customer, or individual). Therefore, AI tools that can deliver expected, consistent, and valuable results would be beneficial to users of such systems. Summary of the invention
[0005] Embodiments of the present disclosure are generally directed to systems and methods for navigating a conversation flow using a trained intelligent robot program, which substantially improves upon the prior art. When a chat session is initiated between a user and a trained intelligent robot program, one or more utterances may be received from the user. The trained intelligent robot program may be used to process the utterances to resolve intents from a plurality of predefined intents, wherein the intelligent robot program is trained to resolve the predefined intents from the user utterances based on training data associated with each predefined intent. The intelligent robot program may be used to navigate a predefined conversation flow associated with the resolved intent, wherein the intelligent robot program uses context variables associated with the user or the chat session to guide the user in the conversation flow. Enterprise data retrieved by the intelligent robot program may be provided to the user using a retrieval request generated based on one or more of the navigation or context variables of the conversation flow.
[0006] Features and advantages of the embodiments are set forth in the following description, or will be obvious from the description, or may be learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0007] Other embodiments, details, advantages and modifications will become apparent from the following detailed description of the preferred embodiments taken in conjunction with the accompanying drawings.
[0008] Figure 1 A system for navigating a conversation flow using a trained intelligent bot is shown according to an example embodiment.
[0009] Figure 2 A block diagram of a computing device operatively coupled to a system is shown according to an example embodiment.
[0010] Figure 3 A diagram illustrating an implementation dialog flow according to an example embodiment is shown.
[0011] Figure 4 A system for implementing a dialog flow using a trained intelligent robot program and business logic according to an example embodiment is shown.
[0012] Figure 5A System components for implementing a dialog flow using a cloud service client are shown according to an example embodiment.
[0013] Figure 5B System components for implementing a dialog flow using a web logic container are shown according to an example embodiment.
[0014] Figures 6A-6C An example conversation flow is shown according to an example embodiment.
[0015] Figure 7 A flow chart of navigating a conversation flow using a trained intelligent robot program is shown according to an example embodiment.
[0016] Figure 8 An integrated supplier, inventory, and logistics system including improved supply operations as disclosed herein is shown according to an example embodiment. DETAILED DESCRIPTION
[0017] Embodiments use trained intelligent robot programs to navigate conversation flows. In some embodiments, a chat session can be initiated between a user and a trained AI component (e.g., a trained intelligent robot program). For example, a user can interact with a software application such as an enterprise software application with a given configuration. The software application configuration may include data displayed by the software application, a typical representation of the data that the user is viewing, analyzing, or otherwise selecting, a user identification or role, and any other appropriate configuration information. In some embodiments, such software application configuration information may be referred to as a context.
[0018] When chatting with a trained intelligent robot program, the user may have an intention. For example, when using an enterprise application, the user may want to retrieve certain enterprise data, process or analyze certain data, run a batch software process or some other function, a combination of these intentions, or any other appropriate intention. In some embodiments, the intelligent robot program can be trained based on the text or user speech input by the user to parse the intention from multiple intentions. For example, the intelligent robot program can be trained with example text for each of the multiple intentions, and thus the example can be used to parse the intention from the user speech. The software application context that can be passed to the intelligent robot program can also be used to parse the intention.
[0019] In some embodiments, the intelligent robot program can navigate a predefined dialog flow associated with the parsed intent. For example, the dialog flow associated with the intent can be defined in a markup language data document (or other data description language document). In some embodiments, the data document can define variables, states, and transitions of the dialog flow. The intelligent robot program can exchange messages with the user based on the predefined dialog flow. In some embodiments, the state transitions followed by the intelligent robot program to navigate the dialog flow are based on responses provided by the user to the messages of the intelligent robot program.
[0020] In some embodiments, the navigation of the dialog flow can trigger the execution of functions. For example, a retrieval request (e.g., a query) can be generated to retrieve enterprise data, wherein the parameters for the retrieval request are based on the response and / or software application context provided by the user when navigating the dialog flow. In another example, a software process (e.g., a batch process) can be triggered, such as a process that is adjusted based on user intervention / editing or a process that includes user approval, wherein the dialog flow can involve user intervention, editing, and / or approval related to the software process. In some embodiments, one or more of these functions can be included in the navigation of the dialog flow.
[0021] Reference will now be made in detail to embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings. In the detailed description that follows, many specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be apparent to one of ordinary skill in the art that the present disclosure may be practiced without these specific details. In other cases, well-known methods, processes, components, and circuits are not described in detail to avoid unnecessarily making the various aspects of the embodiments unclear. Wherever possible, the same reference numerals will be used for the same elements.
[0022] Figure 1A system for navigating a conversation flow using a trained intelligent robot program according to an example embodiment is shown. System 100 includes a service component 102, a network logic container 104, and a robot program service 106. In some embodiments, the service component 102 can provide a client that communicates with the network logic container 104 and / or the robot program service 106 to navigate the conversation flow. For example, a chat session can be initiated with a user via the service component 102, and the network logic container 104 and the robot program service 106 can provide functions for the chat session.
[0023] In an embodiment, the robot program service 106 can provide a trained intelligent robot program for a chat session. In an embodiment, the network logic container 104 can host a chat server for a session and / or implement a custom business logic component for navigating a dialog flow. The service component 102, the network logic container 104, and the robot program service 106 can be used in combination to implement a chat session with a user, wherein the trained intelligent robot program navigates the dialog flow with the user. Any other suitable framework, component, or combination of components can be similarly implemented.
[0024] Figure 2 is a block diagram of a computer server / system 200 according to an embodiment. All or part of the system 200 may be used to implement Figure 1 Any components shown. Figure 2 As shown, the system 200 may include a bus device 212 and / or other communication mechanism(s) configured to communicate information between various components (e.g., the processor 222 and the memory 214) of the system 200. In addition, the communication device 220 may enable connectivity between the processor 222 and other devices by encoding data to be sent from the processor 222 to another device over a network (not shown), and decoding data received from another system over a network for the processor 222.
[0025] For example, the communication device 220 may include a network interface card configured to provide wireless network communications. Various wireless communication technologies may be used, including infrared, radio, Wi-Fi and / or cellular communications. Alternatively, the communication device 220 may be configured to provide a wired network connection(s), such as an Ethernet connection.
[0026] The processor 222 may include one or more general or special purpose processors to perform the computation and control functions of the system 200. The processor 222 may include a single integrated circuit, such as a microprocessing device, or may include multiple integrated circuit devices and / or circuit boards working in concert to perform the functions of the processor 222. In addition, the processor 222 may execute computer programs stored in the memory 214, such as the operating system 215, the AI module 216, and the additional function module 218.
[0027] The system 200 may include a memory 214 for storing information and instructions for execution by the processor 222. The memory 214 may contain various components for retrieving, presenting, modifying, and storing data. For example, the memory 214 may store software modules that provide functions when executed by the processor 222. The modules may include an operating system 215 that provides operating system functions for the system 200. The modules may include an operating system 215, an AI module 216 configured to perform dialog flow navigation and all other functions disclosed herein, and an additional function module 218. The operating system 215 provides operating system functions for the system 200. In some cases, the AI module 216 may be implemented as a configuration in the memory. In some embodiments, when the system 200 performs the functions of the AI module 216, it implements an unconventional special-purpose computer system that performs the functions disclosed herein.
[0028] The non-transitory memory 214 may include various computer-readable media accessible by the processor 222. For example, the memory 214 may include any combination of random access memory ("RAM"), dynamic RAM ("DRAM"), static RAM ("SRAM"), read-only memory ("ROM"), flash memory, cache, and / or any other type of non-transitory computer-readable media. The processor 222 is further coupled to a display 224, such as a liquid crystal display ("LCD"), via the bus 212. A keyboard 226 and a cursor control device 228 (e.g., a computer mouse) are further coupled to the communication device 220 to enable a user to interact with the system 200.
[0029] In some embodiments, the system 200 may be part of a larger system. Therefore, the system 200 may include one or more additional functional modules 218 to include additional functionalities. The additional functional modules 218 may include, for example, Cloud infrastructure, Cloud Platform, Various modules of the cloud application. AI module 216, additional function module 218 and any other suitable components of system 200 may include Retail Advanced Science Cloud Services, Intelligent robot program, Various modules for digital assistants and other appropriate frameworks / services.
[0030] Database 217 is coupled to bus 212 to provide centralized storage for modules 216 and 218 and to store data such as AI module 216 or other data sources. Database 217 can store data in an integrated collection of logically related records or files. Database 217 can be an operational database, an analytical database, a data warehouse, a distributed database, an end-user database, an external database, a navigational database, an in-memory database, a document-oriented database, a real-time database, a relational database, an object-oriented database, a non-relational database, a NoSQL database, Distributed File System ("HFDS") or any other database known in the art.
[0031] Although shown as a single system, the functionality of system 200 may be implemented as a distributed system. For example, memory 214 and processor 222 may be distributed across multiple different computers that collectively represent system 200. In one embodiment, system 200 may be part of a device (e.g., a smartphone, tablet, computer, etc.). In embodiments, system 200 may be separate from the device and the disclosed functionality may be provided remotely to the device. Additionally, one or more components of system 200 may not be included. For example, for functionality as a user or customer device, system 200 may be a smartphone or other wireless device that includes a processor, memory, and display, but does not include a processor, memory, and display. Figure 2 One or more of the other components shown, and including Figure 2 Additional components not shown, such as antennas, transceivers, or any other appropriate wireless device components. In addition, when implemented to perform the functions disclosed herein, system 200 is a special-purpose computer specifically adapted to provide AI chat session functionality.
[0032] Figure 3 A diagram illustrating a dialog flow according to an example embodiment is shown. For example, Figure 3 Dialog flow 310 is depicted to include intent 302, utterance 304, entity 306, and business logic 308. For example, a combination of intelligent bots and / or other system components may use intent 302, utterance 304, entity 306, and business logic 308 to navigate dialog flow 310.
[0033] In some embodiments, utterances 304 may be received from a user during a chat session with an intelligent robot program. Using utterances 304, intent 302 may be identified from multiple intents by a trained intelligent robot program. For example, an intent may be a permutation and combination of a typical user request and statement or utterance 304. In some embodiments, an intelligent robot program may be trained to identify a given intent using example text indicating a given intent. When utterances 304 are similar to the examples of intent 302 used to train the intelligent robot program, the trained intelligent robot program may identify intent 302 from multiple intents. A rich collection of example utterances enables the intelligent robot program to understand the user's needs. For example, an intent may be identified from the utterance "Show me optimized results for location Boston" and distinguished from other predefined intents.
[0034] Intents enable an intelligent bot to understand what the user wants it to do. For example, an intent may categorize typical user requests by tasks and actions that the bot performs. An example intent using an embodiment with an enterprise application for retail optimization may be an optimization profile intent, which is associated with a direct request, reviewing an optimization profile.
[0035] In some embodiments, intents for recognition can be created based on a compilation of utterances for a particular action, e.g., a robust dataset of text (e.g., one to two dozen utterances). The compilation of utterances can also vary so that a trained intelligence can interpret ambiguous user input. For example, a rich collection of utterances enables an intelligent robot program to understand a user's needs from user messages with different structures, tones, grammar, etc.
[0036] Intents can also be viewed as descriptions of use cases, as intents describe various actions that a robot helps its users accomplish. Not only do intents describe what a robot can do, but intents are also used by the robot to recognize user input, as each intent has a set of utterances associated with it. The following represents an example intent:
[0037] Intents can be created for typical pricing users.
[0038] Management by exception. For example, a user can ask to see the position of departments with profit growth below 5%.
[0039] Approval processes. For example, rejecting price recommendations for runs with discounts above 40%, or approving all batch runs that show a 5% increase in revenue.
[0040] Recalculate key performance metrics. For example, reject recommendations with discounts above 70% and recalculate the metric.
[0041] The identified intent can be associated with a data document (e.g., a markup language document or a data description language document) that defines the dialog flow for the intent. For example, intent 302 can be associated with a data document that includes variables, states, and transitions. Entity 306 can be the context of intent 302, and the context is used with the data document to navigate the dialog flow. For example, when intent maps words and phrases to specific actions, entities add context to the intent itself. They help describe intent more fully and enable intelligent robot programs to complete user requests. An example of entity 306 is the location Boston in the utterance "Show me optimized results for location Boston".
[0042] In some embodiments, the data document may include variables for navigating the flow of dialogue. For example, the flow of dialogue may involve retrieving a specific representative of enterprise data. Therefore, the variables may include parameters that define the representative of enterprise data. In an example, the data document may be structured so that the intelligent robot program obtains one or more values for the parameters in order to construct a retrieval request requesting the return of the desired data.
[0043] In some embodiments, user variables can also be used to navigate the dialog flow. For example, user variables can include user identity, role, preferences, etc. Other variables, such as context variables of the software application or the state of the software application, can also be passed to navigate the dialog flow. For example, session and / or user-scoped variables can be defined in the dialog flow. These variables can be part of a custom component that is used to communicate information between the software application and the intelligent robot program engine (e.g., via a REST API).
[0044] In some embodiments, variables may include alphanumeric values, email addresses, specific times, durations, phone numbers, uniform resource locators, etc. Variables may also include lists of values, such as a list of user inputs that may indicate similar concepts; lists of entities, such as a superset of entities; derived variables, where one variable is derived from another; or regular expressions, which may allow intelligent robot programs to identify predefined patterns in user input. Entity 306 may include one or more of these variables.
[0045] Business logic 308 may include business intelligence used by the intelligent robot program to navigate the dialog flow and perform actions for the user. For example, certain states, transitions, and / or functions called within the data document may be structured according to business considerations (e.g., according to the data schema of the enterprise data expected to be retrieved, according to the structure of the batches configured by navigating the dialog flow, or any other appropriate considerations). In some embodiments, when returning data from the transaction data store and / or when executing business logic, the intelligent robot program may use custom components, such as an example of an intelligent robot program that obtains summary results for the location Boston from the enterprise database.
[0046] In some embodiments, the data document may be a markup language document. For example, the data document may implement a predefined YAML Ain't Markup Language ("YAML") structure, wherein the intelligent robot program is configured to navigate the dialog flow based on the predefined YAML structure. Any other suitable data document may be implemented similarly.
[0047] The following is an example data document definition of a parsed intent related to a simplified version of ordering a pizza, where the data document is used to configure an intelligent bot to navigate a conversation flow with the user to complete a pizza order:
[0048] platformVersion:"1.0"
[0049] main:true
[0050] name:"PizzaBot"
[0051] parameters:
[0052] age:18
[0053] context:
[0054] variables:
[0055] size:"PizzaSize"
[0056] type:"PizzaType"
[0057] crust:"PizzaCrust"
[0058] iResult:"nlpresult"
[0059] sameAsLast:"YesNo"
[0060] states:
[0061] intent:
[0062] component:"System.Intent"
[0063] properties:
[0064] variable:"iResult"
[0065] confidenceThreshold:0.4
[0066] transitions:
[0067] actions:
[0068] OrderPizza:"checklastorder"
[0069] CancelPizza:"cancelorder"
[0070]
[0071] to:"size,type,crust"
[0072] transitions:{}
[0073] The following is another example data document definition of a parsed intent related to enterprise software implementation as described in various embodiments of the present disclosure.
[0074] platformVersion:"1.0"
[0075] main:true
[0076] name:"optimizationFlow"
[0077] context:
[0078] #Define variables such as iResult,iResult2,commonActionType,nextAction,taskflowId,moduleExtKey,locationExtKey,productExtKey,seasonExtKey,runTypeExtKey,runActionExtKey,runExtKey,yesOrNoExtKey,actionType
[0079] primaryKey,copyFlag,recommendationType,recommendationAction,recomDiscountCriteria,recomTypeExtKey,metricsTypeExtKey,metricCriteria,metricType,runApprovalOption
[0080] states:
[0081] init:
[0082] component:"commonGreeting"
[0083] properties:
[0084] botType:"CHATBOT"
[0085] commonActionType:"INITIALIZE"
[0086] transitions:{}
[0087] Intent:
[0088] component:"System.Intent"
[0089] properties:
[0090] confidenceThreshold:0.4
[0091] variable:"iResult"
[0092] transitions:
[0093] actions:
[0094] unresolvedIntent:"Unresolved"
[0095] Greeting:"greeting"
[0096] Reset:"reset"
[0097] SummaryOptimization:"displayOptimization"
[0098] CopyOptimizationRun:"askLocationRun"
[0099] ApproveOptimizationRun:"askLocationRun"
[0100]
[0101]
[0102]
[0103]
[0104] In some embodiments, the example dialog flow may include context, default transitions, and states, and may also include custom components (e.g., for communicating with external components and / or components of an enterprise software application). The following are examples of custom components in a dialog flow (e.g., commonGreeting and retailOptimization). For example, each component may have attributes as variables that are passed to the custom component (e.g., from a robot service) to take action. In some embodiments, the custom component, after processing a request, may update context variables and send them to the robot service in a response.
[0105] init:
[0106] component:"commonGreeting"
[0107] properties:
[0108] botType:"CHATBOT"
[0109] commonActionType:"INITIALIZE"
[0110]
[0111] In combination, intent 302, utterance 304, entity 306, and business logic 308 can be used to navigate a dialog flow 310 and / or perform related actions for a user. In some embodiments, an intelligent robot program can be a transactional robot program such as in a retail conversational AI setting, which interacts with a retail data warehouse (e.g., a database storing enterprise data) to review key performance metrics ("KPIs") and business performance by asking questions. These intelligent robot programs can maintain context on the conversation and can be integrated into the retail enterprise application flow. These intelligent robot programs enable retailers to conduct interactive conversations within a UI workflow, where various responses are provided to end users within the enterprise application, such as a simple text message with a KPI, multiple choices for the user to choose, or a guide to other screens based on system status. Figures 6A-6C Illustrations of example implementations are provided.
[0112] In some embodiments, the chat session may be coupled with voice assistants (e.g., Echo, Google Home, Chrome voice assistant, etc.), machine learning techniques (e.g., natural language processing), and enterprise software (e.g., retail quote optimization, etc.). Such implementations can reduce the workload for typical users of enterprise software applications and can include handling complex business process use cases.
[0113] In current versions of enterprise retail software, complex business use cases such as pricing decisions / optimization are not integrated with voice assistants / intelligent bots to handle complex workflow / business process problems. For example, a buyer or manager managing retail apparel price recommendations would be interested in use cases such as exception management (e.g., based on price recommendations, show me the locations of departments with profit growth below 5%), complex approval processes (e.g., reject price recommendations above 40%, recalculate key performance metrics, send targeted quotes to my platinum customers), or instant rollups (e.g., what is the profit growth in the United States? Germany?). In some embodiments, users can use natural language processing (“NLP”) algorithms (e.g., such as A specific question (utterance) is verbally expressed to a voice-enabled device by an intelligent robot program, and the utterance can be converted to text and mapped to intents / entities. In some examples, once the intents / entities are resolved and the dialog flow is navigated, a pricing solution (e.g., Retail Quote Optimization), the solution can execute the appropriate process and return the information.
[0114] Embodiments can reduce the amount of work required for users to conduct certain complex business processes, while also adding the ability for users to speak to enterprise business software applications in natural language based on voice-enabled functionality. For example, as a typical user of such software, a business user can speak their business use cases to the enterprise business software application.
[0115] Embodiments include trained and integrated intelligent robot programs (e.g., using Digital assistant implementations) to answer specific questions generated for a typical user in, for example, a pricing software solution. For example, a buyer or manager managing retail clothing price recommendations may be interested in use cases such as the following:
[0116] Management by exception. For example, a user can ask to see the position of departments with profit growth below 5%.
[0117] Approval processes. For example, rejecting price recommendations for runs with discounts above 40%, or approving all batch runs that show a 5% increase in revenue.
[0118] Recalculate key performance metrics. For example, reject recommendations with discounts above 70% and recalculate the metric.
[0119] ·Q&A. This integration helps answer the questions most frequently asked by users. For example, a user can ask about promotions, and the Q&A tab displays relevant questions and answers related to the item "Promotions".
[0120] Some embodiments may include the following:
[0121] 1. The user can enable the voice assistant (e.g., by clicking on the microphone symbol).
[0122] 2. The user can speak specific questions (utterances) to the voice-enabled device.
[0123] 3. NLP algorithms can be used to convert utterances into text, and then the intent and corresponding entities can be parsed from the utterances. For example, a trained intelligent robot program can be used to identify the intent of the above use case. The trained intelligent robot program can also be used to ask suitable follow-up questions to achieve the intent.
[0124] 4. Once the intent / entity is resolved, the conversation flow can be navigated and enterprise software applications can be invoked (e.g. Retail offer optimization), the enterprise software application then executes the appropriate process.
[0125] 5. Once the process is complete, the enterprise software application can return information.
[0126] One or more intents can be created for different use cases to help users achieve relevant functions. In addition, one or more intelligence robot programs can be trained to recognize the created intents. For example, a user can ask to see the location of departments that do not meet approval criteria, or a user can ask to see the location of departments with profit growth below 5%. As described in this article, example utterances for each intent can be provided to train the intelligent robot program.
[0127] Embodiments can implement a variety of training techniques. For example, Trainer Ht technology can be used, which uses NLP-based linguistics to derive a combination of derived rule templates based on a provided set of example utterances. This can help intelligent robot programs to be deployed quickly and can produce highly accurate predictions. Trainer Ht can be used with a small training corpus (e.g., 10-20 utterances for the created intent), can utilize NLP-based linguistics to identify intent, and can use matching rules to create templates and distinguish intent.
[0128] When larger amounts of conversational data are available, another training technique, Trainer Tm, can be used. Trainer Tm is well suited for situations where generalization of intent prediction is needed and the implementer has built a good quality initial dataset for the conversation flow. These models can use a combination of algorithms that perform unsupervised learning from large amounts of unlabeled text to generate context-sensitive mappings for user intent prediction based on data collected over time from end users in retail enterprise applications.
[0129] In addition to providing training data specific to the pricing solution, other appropriate NLP processing techniques may be used, such as techniques to identify sample phrases belonging to each intent and phrases similar to those phrases. NLP technology implemented by digital assistants. In some embodiments, NLP algorithms that perform word analysis and semantic analysis can be used. For example, the matching engine can grammatically and semantically analyze the user's expression or statement to understand the user's intention.
[0130] In some embodiments, NLP techniques may include spelling correction, lemmatization (e.g., the form of a word can be associated with a general or base form of the word called a root), stemming (e.g., the ending of a word can be removed to obtain the base word, e.g., cats becomes cat), synonym lookup, hypernym and hyponym analysis (e.g., parent-child relationships, or types of something, such as the color is a hypernym of red, and spoon is a hyponym of cutlery (a type of cutlery)), detection of multiple word units, resolution of lexical ambiguity (e.g., using knowledge of specific industries and users to resolve ambiguity, such as in the financial field, bank refers to a financial institution, not a river bank), semantic analysis, and other suitable NLP techniques.
[0131] In addition to training, entities can also be defined for intents. For example, in some cases, context can be used to enable intelligent bots to complete user requests. Depending on the use case and request, entities such as location, item can be added. For example, for an intent related to finding runs that do not meet approval criteria, the following entities can be included: location, department.
[0132] Once training and entities are defined, custom components such as custom business logic and / or functionality (e.g., REST API calls) can be integrated. Embodiments include defining a dialog flow using a data document (e.g., a markup language data document or similar data definition document), wherein the intelligent robot program can use the training, entities, custom components / business logic, and the defined data documents to navigate the dialog flow with the user.
[0133] Figure 4 A system for implementing a conversation flow using trained intelligent bots and business logic according to an example embodiment is shown. The system includes an enterprise AI client 402, a network logic container 404, a bot service 406, an advanced science cloud service chat component 408, an AI chat server 410, a bot request 412, an event webhook 414, a channel configuration 416, a service configuration 418, a training component 420, a conversation flow 422, a business logic component 424, a cache 426, a database 428, and a custom component API 430.
[0134] The enterprise AI client 402 may have an integrated voice-enabled solution that allows the user to either ask questions or type questions to get responses, and displays a history of questions and responses for the user to review (e.g., a chat display). The enterprise AI client 402 may use, for example, The speech recognition kit in the Chrome browser converts speech to text, or any other suitable technology, service, or framework may be used. The enterprise AI client 402 may be displayed in a user interface (e.g., in a context area) that allows a user to ask questions in the context of a user workflow. An example of such a query from a client is "What promotion is running at store XYZ" while viewing targeted offers in an offer optimization workflow. The enterprise AI client 402 may also provide the ability to reset and restart a conversation with an enterprise application by resetting the context from the UI.
[0135] The network logic container 404 and the AI chat server 410 can relay client queries to the robot service 406 by attaching application context and / or security role to the client request. For example, a customer query can be a simple request or have detailed criteria for obtaining, an example of such query criteria can be to show promotions in store YY or department XX. The network logic container 404 and the AI chat server 410 can also provide enterprise data to the robot service 406 using the basic custom component API 430, the business logic component 424, the cache 426, and the database 428.
[0136] The robot program service 406 may include a composite of dialog flows that define dialogues and transition states. These dialog flows can be configured for business use cases and can rely on built-in components and / or can have component-controlled navigation that relies on enterprise-specific components to display messages in sequence and transition to the next state. The robot program service 406 can use a training component 420 to periodically batch train models using data collected from past use of the robot program. In some embodiments, these component layers can interact with each other (e.g., using a REST application programming interface ("API")) and share messages (e.g., JSON object messages) with an AI chat server 410 that acts as a mediator, which can be deployed as a network application on a Java enterprise server. In some embodiments, enterprise applications can share application contexts with the robot program service 406, such as logged-in users, their roles, and applications or modules currently displayed / navigated to in the UI.
[0137] In some embodiments, the enterprise AI client 402 may send a bot request to the network logic container 404, which may relay the request to the bot service 406. Channel configuration 416 and / or service configuration 418 may be used to configure a chat session with an intelligent bot. In some embodiments, navigation of a dialog flow (e.g., dialog flow 422) may invoke a business logic component, and the base custom component API 430, business logic component 424, cache 426, and database 428 may be used to perform business operations and / or return enterprise data. Event network hooks 414 may be configured and used to relay information back to the enterprise AI client 402, such as Figure 4 shown.
[0138] In some embodiments, an example call to the base custom component API 430 may include:
[0139]
[0140]
[0141] In some embodiments, an example response to a call from base custom component API 430 (e.g., after retrieving data and / or performing an action via business logic component 424, cache 426, database 428) may include:
[0142]
[0143]
[0144] In some embodiments, the robot service 406 may send a request and one or more parameters and context variables maintained by the robot service through the custom component. The base custom component API 430 may retrieve the relevant information (e.g., from a transaction database using JDBC), or may communicate with other components to obtain the data. In some embodiments, based on the passed parameters, the base custom component API 430 may then update the context variables and construct a response to send back to the robot service 406. The response may include data, a success or error response, and / or the next action that the robot service 406 should take.
[0145] In some embodiments, user chat can start from a user prompt sent to robot program service 406, where NLP algorithms or machine learning models are used to parse intent. After parsing the intent, robot program service 406 can evoke one or more components to obtain real-time enterprise data, where such data and the next conversion state can be returned to robot program service 406 as a response. The conversion state can be parsed to take the next action and / or provide a response message to the user, which is displayed in the chat window. In some embodiments, the response message can be a multiple choice, a simple message, or it can guide the user to another screen in the enterprise application. The dialog flow in the robot program service 406 can be designed to have an exception flow, which is configured to guide the user when the intent is not parsed or the user is prompted to provide further clarification.
[0146] Figure 5A System components for implementing a dialog flow using a cloud service client are shown according to an example embodiment. Figure 5A Depicted is an Advanced Science Cloud Service chat component 408, which may include a UI component 450, a speech recognition component 452, a custom change component 458, a UI to ADF interaction 454, and a network socket 456. The UI component 450 may display a chat client (e.g., a chat window, such as a web browser) in an interface of a client and / or within a UI of an enterprise application. Figures 6A-6C ). The speech recognition component 452 can recognize natural language speech and can include Speech recognition suite in Chrome or any other suitable component. Custom changes 458 may include a custom implementation for a chat session at the client.
[0147] For example, a chat may start with an enterprise application greeting and by setting the context for the conversation, which may then be utterances from the user that are parsed and converted to a conversation at the bot service 406. Custom changes 458 may implement custom greetings and / or custom reset conditions to reset the context. For example, a JavaScript reset method may be added to clear messages from the chat client. Once the message is cleared, a "reset" utterance mapped to a reset intent may be pushed to the bot service 406 to reset the variables. In some embodiments, a JavaScript initialization greeting message may be added to the chat client when the network socket connection is initialized.
[0148] In some embodiments, the application data framework ("ADF") can interact with application / UI components. For example, when talking to an intelligent robot program, a message can be sent from a custom component to the robot program server 406, and then the robot program server returns the result as a callback to the channel network hook (Webhook). These parameters can then be passed to the composite component, which can pass these parameters to the JavaScript parent window (e.g., ADF component). This component can then call the baking bean (e.g., to open a task flow). In some embodiments, JavaScript is added to convert speech to text, and then the text is pushed to the robot program server 406, which can be similar to the way a typed text message is passed. The chat client can communicate with the chat server using a network socket 456.
[0149] Figure 5B System components for implementing a conversation flow using a network logic container are shown according to an example embodiment. Figure 5B A network logic container 404 is depicted, which may include chat configuration 460, task flow 462, chat context 464, chat JSON model 466, chat security 468, message events 470, web socket mediator 472, base custom component API 474, web hooks 476, component registry 478, credential store 480, security context 482, business services and components 484, and roles and permissions 486.
[0150] The chat configuration 460 may use the channel ID and other relevant configuration information to configure the chat. The task flow 462 may include task flow metadata that may be used for integration with UI components and enterprise software applications. The chat context 464 may include context data for the chat session, including user context, enterprise application context, etc. The chat JSON model 466 may include a data description for configuring an intelligent robot program to guide a user (e.g., a data description for a dialog flow data document).
[0151] Chat security 468 may include administrative roles, such as view roles and service roles (e.g., granted to limited users registered with bot service 406). Credential store 480 may store channel keys used to maintain security of chat sessions, and security context 482 may ensure that user roles are verified by external custom components (e.g., custom components of an enterprise application).
[0152] Message events 470 may represent messages between a user and an intelligent robot (e.g., messages received as callbacks from robot service 406). In some embodiments, this is implemented using the Secure Sockets Layer (“SSL”) protocol. Web socket mediator 472 may receive messages from a chat client and push data to robot service 406. Basic custom component API 474 may be used to read message payloads from custom components (e.g., components of an enterprise application that returns enterprise data). Web hooks 476 are configured as part of the channel configuration. Component registry 478 may use a Spring application context as a registry that may indicate which service / component to invoke from the common layer and delegate to the business layer.
[0153] In some embodiments, a question and answer ("Q&A") intelligent bot is also configured. The Q&A bot can rely on a predefined set of questions and answers to guide users and customer service within contextual issues. The Q&A bot can rely on elastic search to parse intent using built-in rules and inference rules. This type of bot can minimize customer service engagement because they get an immediate response from the enterprise solution and also reduce the involvement of the enterprise development team in responding to repetitive questions from multiple retailers.
[0154] A Q&A bot can be implemented using a main bot and child bots, which can resolve based on a set of contexts. In an embodiment, if the enterprise application is retail, and it has modules such as forecasting and quote optimization, the main bot can be configured with logic to switch to a specific Q&A bot for a specific intent solution. Thus, forecasting and quote optimization are not mixed together, but separated in different bots.
[0155] In some embodiments, custom components can be used to resolve intent after component navigation. For example, some conversation navigation can be resolved by an intelligent robot program. An example of component navigation is a situation where a custom component displays multiple messages in sequence and then requires the intelligent robot program to use an action to transition to the next state. An example embodiment is shown below:
[0156] User: Show me the optimization results.
[0157] Bot: (Calls the custom component and the custom component returns a list of messages with the optimization result metrics and a prompt for the user’s next action via a question (do they want to approve or copy the result?))
[0158] User: Show me the optimization results (again)
[0159] Bots: Use controlled customization - Navigational smart bots can correctly interpret intent and display
[0160] (The custom component is called and returns a list of messages with metrics, and a question prompting the user's next action (do they want to approve or copy the results?))
[0161] In some embodiments, navigating a conversation flow may include conversation navigation and custom component navigation. For example, in conversation navigation, the intent may be resolved by an intelligent robot program (e.g., using a defined conversation flow and example utterances), while in custom component navigation, the intent resolution may be forwarded from the custom component to the intelligent robot program (as shown in the present disclosure). In the example of custom component navigation above, the custom component displays multiple messages in sequence (e.g., without user actions in between), and when prompted by the user, the custom component is responsible for configuring the intelligent robot program to resolve the user intent. In some embodiments, navigation of a conversation flow may include conversation navigation and custom component navigation (e.g., in a combined form).
[0162] Additional example text illustrating navigation of several example conversation flows follows:
[0163] Sample dialogue: National view for all sectors
[0164] User: “I want to see how my store in the US is performing.”
[0165] Bot: "Okay. It looks like you want to see country-level metrics for your store. Please specify the location from: <United States, Europe, US eCommerce>?"
[0166] User: "United States".
[0167] Bot: "Do you want to review a specific department or all departments?"
[0168] User: "All Departments".
[0169] Bot: "The aggregated metrics for the specified location(s) and all departments are: Revenue of 2756451.736, Revenue change of -1.49%, Gross Profit of 1868901.003, Gross Profit change of -3%, and Sell-through Rate of 0.6861, Sell-through Rate change of 1.28%."
[0170] Sample dialogue: National view for selected sectors
[0171] User: “I want to see how my store in the US is performing.”
[0172] Bot: "Okay. It looks like you want to see country-level metrics for your store. Please specify the location from: <United States, Europe, US eCommerce>?"
[0173] User: "United States".
[0174] Bot: "Do you want to review a specific department or all departments?"
[0175] User: "Specific Department".
[0176] Bot: "Which department are you interested in?"
[0177] User: "Women's Sportswear"
[0178] Bot: "The aggregated metrics for the specified location(s) and department(s) are: Revenue of 2756451.736, Revenue change of -1.49%, Gross Profit of 1868901.003, Gross Profit change of -3%, and Sell-through Rate of 0.6861, Sell-through Rate change of 1.28%."
[0179] Sample Dialogue: Management by Exception
[0180] User: "I want to see locations that do not meet approval criteria."
[0181] Bot: "Okay. You seem to be interested in finding exceptions. First, which department are you interested in?"
[0182] User: "Women's sportswear".
[0183] Bot: “What metric are you looking for: revenue growth, profit growth, sales rate growth, or sales rate?”
[0184] User: "Sell-through rate".
[0185] Bot: "Please specify a threshold for identifying poorly performing locations (e.g., 15%)."
[0186] User: “10%”.
[0187] Bot: "Positions in the women's activewear department with a sell-through rate of less than 10% are corporate."
[0188] Bot: "Would you like to see a summary of operations for this department and this location?"
[0189] User: “Yes”.
[0190] Bot: "The metrics for running Copy_Of_test are: Revenue is 2756451.736, Revenue Change is -1.49%, Gross Profit is 1868901.003, Gross Profit Change is -3%, and Sell-through Rate is 0.6861, Sell-through Rate Change is 1.28%.
[0191] Bot: "Do you want to approve this run - yes / no?"
[0192] Depending on what the user answers here, the flow will be different. If the user says "yes", the user is taken to the approval use case. If the user says "no", the user is taken to the recalculation use case.
[0193] Sample dialogue: Approval use case
[0194] User: "I want to review the metrics and want to run approval today."
[0195] Bot: "Okay. You seem interested in an approval run. First, which position are you interested in?"
[0196] User: "Company".
[0197] Bot: "Which department are you interested in?"
[0198] User: "Women's sportswear".
[0199] Bot: "The metrics from running Copy_Of_test are: Revenue is 2756451.736, Revenue Change is -1.49%, Gross Profit is 1868901.003, Gross Profit Change is -3%, and Sell-through Rate is 0.6861, Sell-through Rate Change is 1.28%."
[0200] Bot: "Do you want to approve this run - yes / no"
[0201] User: “Yes”.
[0202] Bot: “The run has been approved and the quote has been sent to the customer.”
[0203] Bot: “Please let me know how I can help you with the Quote Optimization App.”
[0204] Sample Dialogue: Recalculation Use Case
[0205] User: "I want to review the metrics and see if there are any changes to the recommendations."
[0206] Bot: "Okay. You seem interested in accepting or declining the recommendation. First, which position are you interested in?"
[0207] User: "Company".
[0208] Bot: "Which department are you interested in?"
[0209] User: "Women's sportswear".
[0210] Bot: "The metrics from running Copy_Of_test are: Revenue is 2756451.736, Revenue Change is -1.49%, Gross Profit is 1868901.003, Gross Profit Change is -3%, and Sell-through Rate is 0.6861, Sell-through Rate Change is 1.28%."
[0211] Bot: "Do you want to approve this run - yes / no?"
[0212] User: “No”.
[0213] Bot: "Do you want to accept / reject the available recommendation? - Accept / Reject."
[0214] User: “Accept.”
[0215] Bot: "Would you like to receive a promotion / price reduction recommendation? - Promotion / price reduction."
[0216] User: "Promotion".
[0217] Bot: "What is the maximum discount percentage metric to accept a promotional recommendation?"
[0218] User: “40%”.
[0219] Bot: "Promotional price recommendation below 40% has been accepted."
[0220] Bot: "Do you want to recalculate the metrics - yes / no?"
[0221] User: “Yes”.
[0222] Robot: "A run has been submitted to recalculate metrics. To see the new metrics, query the robot."
[0223] Bot: “Please let me know how I can help you next with the Quote Optimization app.”
[0224] In some embodiments, a chat session can be held in the context of an enterprise application (e.g., to perform actions and / or retrieve data associated with the enterprise application). For example, an enterprise application may include a quote optimization function (e.g., a user interface with quote optimization). Retail Advanced Science Cloud Service). In some embodiments, the enterprise application may include and / or access enterprise data, such as a multidimensional enterprise data model. Such a model may include data dimensions, such as products, metrics, locations, calendars, seasons, departments, etc. In some embodiments, the dimensions may each include multiple members (e.g., multiple products, organizational metrics, locations, calendar timings, seasons, departments, etc.), and in some embodiments, one or more of these members may have multiple levels of organizational structure or may be organized in a hierarchy (e.g., products organized by product line, specific organizational metrics organized by more general metrics, such as in-store returns under returns, specific locations under larger geographic locations, finer calendar timings under larger calendar timings, such as weeks under months, specific departments under more general departments, etc.).
[0225] In some embodiments, the representative representations of these dimensions may describe useful information, such as the total sales of a particular product line in stores in a particular geographic location during a particular time period of the year. In this example case, this information may be retrieved based on the specified representative representations of the members of the product, measure, location, and calendar dimensions (e.g., the specified representative representations of each individual member in each of these dimensions of interest). Thus, in some cases, accessing enterprise data may involve specifying parameters of members in a multidimensional data model.
[0226] In some embodiments, offer optimization can be used to determine recommended pricing for promotions, price reductions, or targeted offers. For example, promotions and price reductions can be at a location level (e.g., price zone), and targeted offers can take location into account and / or can be specific to each customer. The timing and depth of promotions, price reductions, or targeted offers may impact inventory management over the life cycle of a product.
[0227] For example, quote optimization can be used in the following example scenarios:
[0228] · Increase inventory to ideal levels, not only during full-price sales but also during clearance sales.
[0229] Maximize total gross profit throughout the product life cycle.
[0230] Evaluate seasonal performance.
[0231] Recommendations are updated weekly. This helps make decisions based on recent data including new sales, inventory, price levels, planned promotions, and other relevant data.
[0232] Provide targeted price recommendations at the market segment level.
[0233] In some embodiments, quote optimization can be used to enhance the buyer's business strategy, such as a business strategy for a department, by generating price recommendations. The buyer can specify a business strategy and may want to determine when to offer promotions, the depth of promotions, which items are considered part of the promotion, and similar issues about price reductions, while keeping in mind certain business rules and inventory levels. In some embodiments, quote optimization can run different price combinations, considering the benefits of offering promotions or price reductions and whether to provide any customer-specific promotions that can achieve optimization goals in an efficient manner. Once the optimization is completed, the run can enter the "ready for review" state. The user can review the recommendation (for example, at each item level) and decide to accept, reject or overturn the recommendation. Once completed, the run state can be changed to "reviewed". The reviewed run can reach the buyer for approval. If the buyer likes the recommendation, the buyer can change the run state to "submitted". For example, "submit" can indicate that the price recommendation is sent to a price execution system or software application.
[0234] In some embodiments, optimization can be performed at a configured process (or operational) location, item level, and calendar level. Once optimization is complete, recommendations can be generated at a level lower than the process level, which is referred to as a recommendation level for the item level and location level. The location and item level can be any level in the location hierarchy and item hierarchy, respectively. Embodiments include optimizations that take into account the following: optimizations that can be performed at a level below the operational item level (for example, if the operational item level is the department, then each job is at the class level); at a configured recommended item level, inventory can be rolled to the desired recommended item level and / or inventory can be aggregated across multiple or all locations to the operational location level; price recommendations can be generated at the recommended item, location and calendar level, and at the customer group level.
[0235] Figures 6A-6C An example conversation flow according to an example embodiment is shown. For example, enterprise application 600 depicts an enterprise application, such as a retail offer optimization application. Chat window 602 describes the navigation of the conversation flow between the intelligent robot program and the user, including parsing intent, navigating the custom data layer of enterprise data, retrieving enterprise information, and triggering enterprise batch processing.
[0236] Figure 7 A flow chart for navigating a conversation flow using a trained intelligent robot program according to an example embodiment is shown. In one embodiment, Figure 7The functions of are implemented by software stored in a memory or other computer-readable or tangible medium and executed by a processor. In other embodiments, each function may be performed by hardware (e.g., by using an application specific integrated circuit (“ASIC”), a programmable gate array (“PGA”), a field programmable gate array (“FPGA”), etc.), or any combination of hardware and software. In an embodiment, Figure 7 The function can be Figure 2 The system 200 may be implemented by one or more elements of the system 200.
[0237] At 702, upon initiation of a chat session between a user and a trained intelligent bot, one or more utterances may be received from the user. For example, a chat session may be initiated between a user interacting with an enterprise application and an intelligent bot provided by a bot service. Utterances or language / text may be received from the user that indicate the user's intent or use case.
[0238] At 704, the utterance may be processed using a trained intelligent robot program to parse an intent from a plurality of predefined intents, wherein the intelligent robot program is trained to parse the predefined intent from the user utterance based on training data associated with each predefined intent. For example, each intent that the intelligent robot program is trained to parse may be associated with a series of example utterances or training data. The intelligent robot program may be trained using these examples so that the intent may be distinguished from the plurality of intents when performing parsing.
[0239] At 706, the intelligent robot program may generate a message, for example, in response to the user's utterance. At 708, a response from the user may be received, for example, in response to a prompt or message from the intelligent robot program. At 710, the communication may involve navigating a predefined dialog flow associated with the resolved intent.
[0240] For example, an intent may be associated with a data document (e.g., a predefined dialog flow) that defines states, transitions, and variables that an intelligent robot program uses to navigate a dialog flow. In an embodiment, the predefined dialog flow configures the intelligent robot program to obtain parameters for a retrieval request from a user, and the parameters are based on the parsed intent associated with the predefined dialog flow. For example, navigation of a dialog flow may involve securing parameters for an enterprise action (e.g., retrieving enterprise data or performing an enterprise batch process).
[0241] In an embodiment, navigating a predefined dialog flow includes generating a preconfigured message from an intelligent robot program that prompts a user to enter parameters for a retrieval request. For example, the intelligent robot program may generate a preconfigured message when transitioning between states within a data definition document that defines a dialog flow. In an embodiment, in response to a first preconfigured message generated by the intelligent robot program in a first state, a transition may be made from at least a first state of a plurality of states to a second state of a plurality of states based on input from a user. For example, when, in response to the first preconfigured message, the input from the user has a first category, the data definition document may define a transition from a first state to a second state, and when, in response to the first preconfigured message, the input from the user has a second category, a transition from a first state to a third state is defined.
[0242] At 712, a retrieval request can be generated based on the navigation dialog flow to retrieve enterprise data using parameters. For example, the enterprise data can include a typical representation of multi-dimensional data, and the navigation dialog flow can include an intelligent robot program that obtains parameters for retrieving the multi-dimensional data from the user. A custom function (e.g., a REST API) can then be called to retrieve the enterprise data, such as from an enterprise data store or database.
[0243] At 714, batch processing can be triggered while navigating the dialog flow. For example, the batch processing can include approving recommendations for optimizing retail offers, running the offer optimization process after editing parameters for the process (e.g., promotion thresholds, price reductions, etc.), rejecting recommended offers or a subset of recommended offers, and other suitable batch processing.
[0244] Figure 8 An integrated supplier, inventory and logistics system including inventory management is shown according to an example embodiment. Figure 8 As shown, system 800 may include an enterprise business system 870 that executes code to manage product inventory at enterprise locations 801-804 using warehouses 880 and to ship products directly to consumers from warehouses 880. Enterprise business system 870 communicates with one or more inventory systems 820 via a cloud network 850 or other type of communication network. In some embodiments, retail offer optimization software of enterprise business system 870 may generate optimized offers for various products. Inventory system 820 and warehouse 880 may perform transportation to and from enterprise locations 801-804 based on these optimized offers. In some embodiments, the use of customized and trained models may provide improvements to these inventory levels, resulting in a more efficient transportation process.
[0245] Inventory system 820 stores inventory and provides transportation logistics to deliver items to business locations 801-804 and consumer locations (e.g., a consumer's home) using trucks 810-813 or some other transportation mechanism. In one embodiment, inventory system 820 implements an enterprise application-specific computer system or a dedicated inventory control system that uses input from enterprise business systems 810, such as retail offer optimization software, to determine inventory levels and the quantity and timing of product deliveries to business locations 801-804.
[0246] Warehouse 880 may be a fulfillment warehouse that supplies one or more products to enterprise locations 801-804 based on inventory system 820 and ships the products to consumer locations (e.g., the consumer's home). In one embodiment, warehouse 880 implements a dedicated computer system or dedicated supplier system (e.g., retail offer optimization software) that uses input from enterprise business system 810 to determine the quantity and timing of product shipments to inventory system 820 and / or enterprise locations 801-804. In some embodiments, warehouse 880 may receive shipments from enterprise locations, e.g., based on returns to enterprise locations 801-804, e.g., to ensure that the enterprise locations are not oversupplied and have adequate product variety.
[0247] Embodiments use trained intelligent robot programs to navigate conversation flows. In some embodiments, a chat session can be initiated between a user and a trained AI component (e.g., a trained intelligent robot program). For example, a user can interact with a software application such as an enterprise software application with a given configuration. The software application configuration may include data displayed by the software application, a representative representation of the data that the user is viewing, analyzing, or otherwise selecting, a user identification or role, and any other suitable configuration information. In some embodiments, such software application configuration information may be referred to as a context.
[0248] When chatting with a trained intelligent robot program, the user may have an intention. For example, when using an enterprise application, the user may want to retrieve certain enterprise data, process or analyze certain data, run a batch software process or some other function, a combination of these functions, or any other suitable intention. In some embodiments, the intelligent robot program can be trained based on the text entered by the user or the user's speech to parse the intent from multiple intents. For example, the intelligent robot program can be trained with example text for each of the multiple intents, and the examples can therefore be used to parse the intent from the user's speech. The software application context that can be passed to the intelligent robot program can also be used to parse the intent.
[0249] In some embodiments, the intelligent robot program can navigate a predefined dialog flow associated with the parsed intent. For example, the dialog flow associated with the intent can be defined in a markup language data document (or other data description language document). In some embodiments, the data document can define variables, states, and transformations of the dialog flow. The intelligent robot program can exchange messages with the user based on the predefined dialog flow. In some embodiments, the state transitions followed by the intelligent robot program to navigate the dialog flow are based on responses provided by the user to the messages of the intelligent robot program.
[0250] In some embodiments, the navigation of the dialog flow can trigger the execution of functions. For example, a retrieval request can be generated to obtain enterprise data, wherein the parameters for the retrieval request are based on the responses provided by the user when navigating the dialog flow and / or the software application context. In another example, a software process (e.g., a batch process) can be triggered, such as a process that is adjusted based on user intervention / editing or includes user approval, wherein the dialog flow can involve user intervention, editing, and / or approval related to the software process. In some embodiments, one or more of these functions can be included in the navigation of the dialog flow.
[0251] The features, structures, or characteristics of the present disclosure described throughout the specification may be combined in any suitable manner in one or more embodiments. For example, throughout the specification, the use of "one embodiment", "some embodiments", "an embodiment", "certain embodiments", or other similar language refers to the fact that a particular feature, structure, or characteristic described in conjunction with an embodiment may be included in at least one embodiment of the present disclosure. Therefore, throughout the specification, the phrases "one embodiment", "some embodiments", "an embodiment", "certain embodiments", or other similar language do not necessarily all refer to the same set of embodiments, and the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
[0252] Those of ordinary skill in the art will readily appreciate that the embodiments discussed above may be implemented with steps in a different order and / or with elements in a configuration different from that disclosed. Thus, while the present disclosure contemplates the embodiments outlined, it will be apparent to those skilled in the art that certain modifications, variations, and alternative configurations will be apparent while still being within the spirit and scope of the present disclosure. Therefore, in order to determine the scope and limits of the present disclosure, reference should be made to the appended claims.
Claims
1. A method for navigating a conversation flow using a trained intelligent robot program, the method comprising: When a chat session is initiated between a user and a trained intelligent robot program, one or more utterances are received from the user, wherein The chat session is initiated when the user interacts with an enterprise application, The enterprise application is configured to retrieve and display one or more representative representations of a hierarchical multidimensional enterprise data model including a plurality of data dimensions, the data dimensions including data members organized in a hierarchical structure, and providing enterprise application context data regarding a state of the enterprise application to the intelligent robot program, at least a portion of the enterprise application context data comprising a representative representation of the hierarchical multidimensional enterprise data model selected for display at the enterprise application; Processing the utterance using the trained intelligent robot program to resolve intent from a plurality of predefined intents, wherein the intelligent robot program is trained based on training data associated with each of the predefined intents to resolve the predefined intents from the user utterance; navigating a dialog flow associated with the resolved intent using the intelligent robot program, such that the intelligent robot program guides the user through the dialog flow using context variables, wherein at least one context variable is defined based on the enterprise application context data provided to the intelligent robot program; providing the user with enterprise data retrieved by the intelligent robot program using a retrieval request generated based on navigation of the dialog flow and the context variables, wherein the retrieval request includes parameters for selecting representative representations across a plurality of dimensions of the hierarchical multidimensional enterprise data model, the parameters being specified by the dialog flow through which the intelligent robot program guides the user using the context variables; and A batch process is triggered in response to user input while navigating the dialog flow through which the intelligent robot program guides the user, wherein the batch process performs a series of calculations using data from the hierarchical multidimensional enterprise data model, and the batch process uses one or more data values of the hierarchical multidimensional enterprise data model modified by the user while navigating the dialog flow.
2. The method according to claim 1, wherein: The training data associated with the parsed intent includes multiple examples of user utterances that are indicative of the parsed intent.
3. The method according to claim 1, wherein: The dialog flow includes a predefined dialog flow that configures the intelligent robot program to obtain the parameters for the retrieval request from the user, and the parameters are based on the parsed intent associated with the predefined dialog flow.
4. The method according to claim 3, wherein: The dialog flow is defined by a data definition document, which includes states, transitions, and variables for the dialog flow.
5. The method according to claim 4, wherein: Navigating the predefined dialog flow includes generating a preconfigured message from the intelligent robot program, the preconfigured message prompting the user to enter parameters for the retrieval request.
6. The method according to claim 5, wherein: The preconfigured messages are generated by the intelligent robot program when transitioning between states within a data definition document defining the dialog flow.
7. The method according to claim 6, wherein: Transitioning from at least a first state of a plurality of states to a second state of the plurality of states is based on input from the user in response to a first preconfigured message generated by the intelligent robot program in the first state.
8. The method according to claim 7, wherein: The data definition document defines a transition from the first state to the second state when the input from the user in response to the first preconfigured message has a first category, and the data definition document defines a transition from the first state to a third state when the input from the user in response to the first preconfigured message has a second category.
9. The method according to claim 1, wherein: The batching is dependent upon user approval being received when navigating the dialog flow.
10. A system for navigating a conversation flow using a trained intelligent robot program, the system comprising: processor; as well as a memory storing instructions for execution by the processor, the instructions configuring the processor to: When a chat session is initiated between a user and a trained intelligent robot program, one or more utterances are received from the user, wherein The chat session is initiated when the user interacts with an enterprise application, The enterprise application is configured to retrieve and display one or more representative representations of a hierarchical multidimensional enterprise data model including a plurality of data dimensions, the data dimensions including data members organized in a hierarchical structure, and providing enterprise application context data regarding a state of the enterprise application to the intelligent robot program, at least a portion of the enterprise application context data comprising a representative representation of the hierarchical multidimensional enterprise data model selected for display at the enterprise application; Processing the utterance using the trained intelligent robot program to resolve intent from a plurality of predefined intents, wherein the intelligent robot program is trained based on training data associated with each of the predefined intents to resolve the predefined intents from the user utterance; navigating a dialog flow associated with the resolved intent using the intelligent robot program, such that the intelligent robot program guides the user through the dialog flow using context variables, wherein at least one context variable is defined based on the enterprise application context data provided to the intelligent robot program; providing the user with enterprise data retrieved by the intelligent robot program using a retrieval request generated based on navigation of the dialog flow and the context variables, wherein the retrieval request includes parameters for selecting representative representations across a plurality of dimensions of the hierarchical multidimensional enterprise data model, the parameters being specified by the dialog flow through which the intelligent robot program guides the user using the context variables; and A batch process is triggered in response to user input while navigating the dialog flow through which the intelligent robot program guides the user, wherein the batch process performs a series of calculations using data from the hierarchical multidimensional enterprise data model, and the batch process uses one or more data values of the hierarchical multidimensional enterprise data model modified by the user while navigating the dialog flow.
11. The system according to claim 10, wherein: The training data associated with the parsed intent includes multiple examples of user utterances that are indicative of the parsed intent.
12. The system according to claim 10, wherein: The dialog flow includes a predefined dialog flow that configures the intelligent robot program to obtain the parameters for the retrieval request from the user, and the parameters are based on the parsed intent associated with the predefined dialog flow.
13. The system according to claim 12, wherein: The dialog flow is defined by a data definition document, which includes states, transitions, and variables for the dialog flow.
14. The system according to claim 13, wherein: Navigating the predefined dialog flow includes generating a preconfigured message from the intelligent robot program, the preconfigured message prompting the user to enter parameters for the retrieval request.
15. The system of claim 14, wherein: The preconfigured messages are generated by the intelligent robot program when transitioning between states within a data definition document defining the dialog flow.
16. The system of claim 15, wherein: Transitioning from at least a first state of a plurality of states to a second state of the plurality of states is based on input from the user in response to a first preconfigured message generated by the intelligent robot program in the first state.
17. The system of claim 16, wherein: The data definition document defines a transition from the first state to the second state when the input from the user in response to the first preconfigured message has a first category, and the data definition document defines a transition from the first state to a third state when the input from the user in response to the first preconfigured message has a second category.
18. A non-transitory computer-readable medium having instructions stored thereon, which, when executed by a processor, cause the processor to navigate a conversation flow using a trained intelligent robot program, wherein: The instructions, when executed, cause the processor to: When a chat session is initiated between a user and a trained intelligent robot program, one or more utterances are received from the user, wherein The chat session is initiated when the user interacts with an enterprise application, The enterprise application is configured to retrieve and display one or more representative representations of a hierarchical multidimensional enterprise data model including a plurality of data dimensions, the data dimensions including data members organized in a hierarchical structure, and providing enterprise application context data regarding a state of the enterprise application to the intelligent robot program, at least a portion of the enterprise application context data comprising a representative representation of the hierarchical multidimensional enterprise data model selected for display at the enterprise application; processing the utterance using the trained intelligent robot program to parse intent from a plurality of predefined intents, wherein the intelligent robot program is trained to parse the predefined intent from the user utterance based on training data associated with each of the predefined intents; navigating a dialog flow associated with the parsed intent using the intelligent robot program such that the intelligent robot program guides the user through the dialog flow using context variables, wherein at least one context variable is defined based on the enterprise application context data provided to the intelligent robot program; providing the user with enterprise data retrieved by the intelligent robot program using a retrieval request generated based on navigation of the dialog flow and the context variables, wherein the retrieval request includes parameters for selecting representative representations across a plurality of dimensions of the hierarchical multidimensional enterprise data model, the parameters being specified by the dialog flow through which the intelligent robot program guides the user using the context variables; and A batch process is triggered in response to user input while navigating the dialog flow through which the intelligent robot program guides the user, wherein the batch process performs a series of calculations using data from the hierarchical multidimensional enterprise data model, and the batch process uses one or more data values of the hierarchical multidimensional enterprise data model modified by the user while navigating the dialog flow.
19. A device for navigating a dialog flow using a trained intelligent robot program, the device comprising components for executing the steps of any one of the methods of claims 1 to 9.
Citation Information
Patent Citations
Systems and methods for providing automated natural language dialogue with customers
CA2995929A1
Intelligent operation and maintenance robot
CN108170734A
Language based solution agent
US10013416B1
System and method for providing an intelligent multi-step dialog with a user
US20010049688A1
Creating coordinated multi-chatbots using natural dialogues by means of knowledge base
US20180025726A1
