Conversation system using mapping of user intent
By mapping NLU intentions to DM intentions and performing multi-intention mapping, the expressive limitations of the dialogue system when expressing user intentions are solved, and efficient and intuitive generation and multi-intention processing of dialogue scenarios are achieved.
Patent Information
- Application Number
- CN202380035047.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-22
- Filing Date
- 2023-03-30
- Publication Date
- 2024-12-06
AI Technical Summary
Existing dialogue systems have expressive limitations when expressing user intentions, especially when complex dialogue flow design and combination of multiple user intentions, it is difficult to describe dialogue scenarios efficiently and intuitively, resulting in complex and unintuitive dialogue generation.
By mapping natural language understanding (NLU) intent to dialogue manager (DM) intent and performing multi-intention mapping in multi-intention situations, generating a list of DM intents to process user input mode, using the intent mapping module and the multi-intention mapping module Abstract and refine intent, reduce branching conditions and improve intuitiveness.
It realizes efficient and intuitive expression of user intentions in the generation of dialogue scenarios, can handle multiple user intentions, simplify branch conditions, and improve the intuitiveness and generation efficiency of the dialogue system.
Smart Images

Figure CN119096247A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to a dialog system using user intent mapping. Background Art
[0002] The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
[0003] like Figure 1 As shown, conventional dialogue systems (e.g., Google's Dialogflow CX, Amazon's Lex, etc.) express user intent as intent based on natural language understanding (NLU) analysis. This intent can be used as a key condition for the chatbot to perform specific actions or dialogue state transitions.
[0004] It is possible to design a conversation flow using only intents based on NLU analysis. However, when designing complex conversation flows, there may be limitations in expressiveness. When expressing user intent, there may be cases where the expression is too detailed or abstract. When the intent occurs in a specific context, there may also be situations where it cannot be expressed. For example, "like" in a response to a confirmation request and "like" in a chat can have different meanings. In addition, there are cases where the expression intent related to a negative question is reversed depending on the state of the conversation. For example, in the response to "Aren't you Hong Gil-dong?", "yes" can be interpreted as a negation. In other words, it may be difficult to make a distinction based only on NLU intents. Due to these limitations in the expressiveness of intent, when implementing a conversation system, the generation of conversation scenarios becomes complicated, and the resulting conversation may become unintuitive.
[0005] The above problems may become more serious when analyzing multiple user intents. Multiple user intents are analyzed in the form of combinations of intents (e.g., lists). As the combinations become more complex, the expressiveness problem of intents can become more serious. When generating dialogue scenarios using combinations of intents, a variety of scenario branching conditions need to be described to respond to various input combinations.
[0006] Therefore, it is necessary to consider a dialogue system that uses methods to effectively express user intent to overcome the limitations of intent expression based on NLU analysis and handle multiple user intents. Summary of the invention
[0007] Technical issues
[0008] The present disclosure seeks to provide a dialogue system that uses a method for expressing user intention based on mapping in the generation of dialogue scenarios. The dialogue system can efficiently and intuitively describe dialogue scenarios corresponding to various user input modes.
[0009] Technical Solution
[0010] At least one aspect of the present disclosure provides a method for expressing an intent contained in a user's speech by a dialogue system. The method includes: extracting at least one natural language understanding (NLU) intent from the user's speech. The method also includes: mapping each of the at least one NLU intent to a dialogue manager (DM) intent. When multiple NLU intents are extracted from the user's speech, the method also includes: generating a DM intent list. Here, the DM intent list includes a DM intent mapped to each of the multiple NLU intents. The method also includes: performing multi-intent mapping on the multiple DM intents included in the DM intent list.
[0011] Another aspect of the present disclosure provides a dialogue system. The dialogue system includes a natural language understanding (NLU) module configured to extract at least one NLU intent from a user's speech. The dialogue system also includes an intent mapping module configured to map each of the at least one NLU intent to a dialogue manager (DM) intent, and when multiple NLU intents are extracted from the user's speech, generate a DM intent list. Here, the DM intent list includes a DM intent mapped to each of the multiple NLU intents. The dialogue system also includes a multi-intent mapping module configured to perform multi-intent mapping on multiple DM intents included in the DM intent list.
[0012] Another aspect of the present disclosure provides a computer-readable recording medium storing instructions. When the instructions are executed by a computer, the computer extracts at least one natural language understanding (NLU) intent from the user's speech. The instructions also cause the computer to map each of the at least one NLU intent to a dialogue manager (DM) intent. When multiple NLU intents are extracted from the user's speech, the instructions also cause the computer to generate a DM intent list. Here, the DM intent list includes DM intents mapped to each of the multiple NLU intents. The instructions also cause the computer to perform multi-intent mapping on multiple DM intents included in the DM intent list.
[0013] Beneficial Effects
[0014] As described above, the present disclosure provides a dialogue system using a method for expressing user intention based on mapping in the generation of dialogue scenarios. Therefore, the dialogue system efficiently and intuitively describes dialogue scenarios corresponding to various user input modes.
[0015] In addition, the present disclosure provides a dialogue system using a method of expressing user intention based on mapping in the generation of a dialogue scenario, so that multiple user intentions can be processed efficiently. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 is an example diagram illustrating a conventional dialogue system.
[0017] Figure 2 is an exemplary diagram illustrating a dialog system according to an embodiment of the present disclosure.
[0018] Figure 3 is an example diagram illustrating intent mapping according to an embodiment of the present disclosure.
[0019] Figure 4 is an example diagram illustrating multi-intent mapping according to an embodiment of the present disclosure.
[0020] Figure 5 is a flowchart illustrating a method of expressing intention according to an embodiment of the present disclosure. DETAILED DESCRIPTION
[0021] Hereinafter, some embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In the following description, although the elements are shown in different drawings, the same reference numerals preferably refer to the same elements. In addition, in the following description of some embodiments, for the purpose of clarity and brevity, the detailed description of related known components and functions will be omitted when it is considered to obscure the subject matter of the present disclosure.
[0022] In addition, various terms such as first, second, A, B, (a), (b), etc. are only used to distinguish one component from another component, and do not imply or suggest the substantive content, order or sequence of the components. Throughout the specification, when a part "includes" or "contains" a component, the part means to also include other components, but not to exclude other components, unless explicitly mentioned to the contrary. Terms such as "unit", "module", etc. refer to one or more units for processing at least one function or operation, which can be implemented by hardware, software or a combination thereof.
[0023] The present embodiment discloses a dialogue system using user intent mapping. More specifically, the present embodiment provides a dialogue system using the following method, which is used to express user intent based on mapping in the generation of dialogue scenarios so as to efficiently and intuitively describe dialogue scenarios corresponding to various user input modes.
[0024] Figure 2 is an exemplary diagram illustrating a dialog system according to an embodiment of the present disclosure.
[0025] The dialogue system according to the embodiment of the present disclosure includes Figure 1 All or part of the natural language understanding (NLU) module 202, the natural language generation (NLG) module 204, the dialog manager (DM), and the knowledge database 206 are shown. Here, in addition to the dialogue state tracking (DST) module 210 and the dialogue policy (DP) module 212, the DM also includes an intent mapping module 220 and a multi-intent mapping module 222.
[0026] Figure 2 The diagram in FIG. 1 is an exemplary configuration according to an embodiment of the present disclosure. Various dialog systems including different components or different connections between components can be implemented according to the structure and operation of the NLU module, the structure and operation of the NLG module, the structure and operation of the DM, and the structure of the knowledge base.
[0027] In the dialogue system, the NLU module 202 analyzes the user's input utterance and derives an intent indicating the purpose of the user's utterance. The NLU module 202 may derive an entity as additional information in addition to the intent. Analysis data such as the intent and the entity are transmitted to the DM. Figure 2 In the example of , the NLU module 202 generates "Chinese" indicating the type of restaurant as the intent of the user utterance.
[0028] The DM operates the DST module 210 and the DP module 212 based on the analysis data transmitted from the NLU module 202. The DST module 210 maintains state variables such as the history of the dialog state, the most recent unanswered user questions, etc. The DP module 212 generates semantic representations using the state variables tracked by the DST module 210 and queries using the knowledge database 206 to generate answers that meet the user's intent. Figure 2 In the example of , DM generates a semantic representation including the location of the restaurant conditioned on the user intent.
[0029] The NLG module 204 generates a text sentence based on the semantic representation generated from the DM, and provides the text sentence to the user.
[0030] The knowledge database 206 stores the knowledge required to generate a semantic representation that matches the user's intent based on the entity and the intent. For example, the intent and the entity may correspond to a condition, and the semantic representation may be an answer that matches the condition. The knowledge database 206 may include various data to mediate between the two.
[0031] Hereinafter, operations of the intent mapping module 220 and the multi-intent mapping module 222 as components added to the DM according to an embodiment of the present disclosure are described.
[0032] As described above, the NLU module 202 derives an entity as additional information in addition to the intent indicating the user's utterance intent. When the user utterance includes multiple intents, the NLU module 202 may generate an analysis result corresponding to the number of intents. For example, in response to the user utterance "I have a fever of about 38 degrees, please wait", the NLU module 202 may generate the following analysis result.
[0033]
[0034] In the above example, it is assumed that the NLU module 202 analyzes multiple intents for a given sentence, but this may vary depending on the structure and operation characteristics of the NLU module 202. For example, when using an NLU module 202 that analyzes one intent per sentence, the same result can be obtained by first performing sentence separation and then obtaining intent for each separated sentence. As described above, the analysis result is transmitted to the intent mapping module 220 within the DM.
[0035] The intent mapping module 220 maps each of the at least one intent analyzed in the NLU module 202 (hereinafter referred to as "NLU intent") to an intent in the DM (hereinafter referred to as "DM intent"). Therefore, the intent mapping can be a decision to express the interpretation of the NLU intent in the current context.
[0036] Figure 3 is an example diagram illustrating intent mapping according to an embodiment of the present disclosure.
[0037] Based on the specific context (e.g., dialogue scene and dialogue state), conditional information obtained during the dialogue, etc., Figure 3 After defining the preset mapping rules regarding which DM intent an NLU intent corresponds to as shown, the intent mapping module 220 can perform intent mapping using these mapping rules.
[0038] By performing intent mapping, the intent mapping module 220 can generate a DM intent that is abstract and refined compared to the NLU intent. Therefore, the DM intent can better express the user intent. When multiple NLU intents are interpreted as the same intent in a specific context, the multiple NLU intents can be abstracted into a single DM intent. On the other hand, when the NLU intents are the same but need to be interpreted differently based on context or conditional information, the DM intent can be used to refine the user intent.
[0039] Intents can be used as branch conditions for conversation scenarios within DM. As described above, when DM intents are introduced and used, there are two advantages as follows. First, the number of required branch conditions can be greatly reduced when describing conversation scenarios. This is an advantage that can be obtained through intent abstraction. The second is the intuitiveness of branch conditions. For example, since "like" needs to be interpreted differently in confirmation request scenarios and chat scenarios, the branch conditions described using DM intents can be more intuitive than those described using NLU intents. This is an advantage that can be obtained by refining intents.
[0040] When multiple user intentions are input, the mapping result of the intention mapping module 220 may be given in the form of a DM intention list including multiple DM intentions. The DM intention list may be transmitted to the multi-intent mapping module 222.
[0041] When multiple DM intents are analyzed based on multiple intents included in the user's speech, DM can basically perform actions corresponding to each DM intent in sequence. However, depending on the combination of input DM intents, DM can select one of several DM intents (for example, in the case of a confirmation request, when the response is "OK, please wait", the system ignores the response and waits), or can convert it into a new DM intent, and then can perform the corresponding action.
[0042] In the following, use Figure 4 An example of the multi-intent mapping module 222 performing multi-intent mapping is described with reference to FIG.
[0043] Figure 4 is an example diagram illustrating multi-intent mapping according to an embodiment of the present disclosure.
[0044] Assume a scenario in which a slot filling dialogue holds "fever", "cough", or "sore throat" to check coronavirus symptoms. In addition, assume that a system action when a user responds to a fever symptom (guiding utterance for fever symptom) and a system action when a user responds to a cough symptom (guiding utterance for cough symptom) are defined. In this regard, when a user responds to both fever and cough symptoms at the same time, the actions corresponding to the respective symptoms may be performed sequentially (e.g., if you have a fever, please...; if you cough, please...). However, generating and providing separate guiding utterances (e.g., when you have a fever and cough, please...) may provide a better user experience. In this regard, if Figure 4 As shown in the second row of the illustrated multi-intent mapping rule example, the multi-intent mapping module 222 can integrate the fever symptom related intent (yes_fever) and the cough symptom related intent (yes_cough) to generate a new DM intent (yes_fever_cough), which can then be transmitted to the next stage. In this regard, the action corresponding to the new DM intent can be a separate guided utterance for both fever and cough symptoms.
[0045] As described above, when processing a DM intent list and performing multi-intent mapping, the multi-intent mapping module 222 may select a DM intent based on whether the DM intent list input in a specific context is constructed by a specific combination. As another example, the multi-intent mapping module 222 may map the specific combination of the DM intent list to a new DM intent.
[0046] like Figure 4As shown, multiple mapping rules can be defined for multi-intent mapping. In other words, for conditions including a specific context (e.g., a conversation scene and a conversation state) and a pattern (multiple DM intents) of a DM intent list, when the conditions are met, the multi-mapping rules assign the DM intent to be mapped.
[0047] The pattern of the DM intent list may be defined as shown in the example of Table 1 (multiple DM intents).
[0048] [Table 1]
[0049]
[0050] As exemplified in Table 1, the pattern may be constructed regardless of the order, and in this regard, each DM intent may be separated by a “,”. As another example, the pattern may be constructed in the order described by the user, wherein each DM intent may be separated by, for example, a “>”. Table 1 is an example of the simplest pattern, and more diverse patterns may be defined and used as needed.
[0051] For Figure 4 DM intent list and multiple multi-mapping rules shown, the multi-intention mapping module 222 can apply the mapping rule that first matches the pattern of the DM intent list. As another example, after applying one mapping rule, the multi-intention mapping module 222 can additionally check whether to apply the remaining mapping rules. In this regard, the mapping rules checked for later application can contain more DM intents than the mapping rules previously applied. When no more mapping rules are applied to the DM intent list, the multi-intention mapping module 222 transmits the checked mapping results to the subsequent DM stage (e.g., DST module 210). When no multiple mapping rules are applied to the DM intent list, the multi-intention mapping module 222 will not generate results based on the multi-intention mapping. In this regard, as described above, the DM can sequentially perform actions corresponding to each DM intent.
[0052] Using this multi-intent mapping, the combination of each DM intent that meets the pattern can be converted into a DM intent and processed. In addition, using this mapping process, the multi-intent mapping module 222 can abstract the DM intent list again.
[0053] A DM intent obtained based on intent mapping and multi-intent mapping according to this embodiment can cover a very large number of user input patterns. As a result, this embodiment can process multiple user intents in a more intuitive and efficient manner, rather than processing multiple user intents with NLU intents.
[0054] In the following, use Figure 5 The diagram of describes the method performed by the dialog system to express the intent contained in the user's utterance.
[0055] Figure 5 is a flowchart illustrating a method of expressing intention according to an embodiment of the present disclosure.
[0056] The dialogue system extracts at least one NLU intent from the user utterance (S500). The dialogue system analyzes the user's input utterance and derives an intent indicating the purpose of the user utterance. The dialogue system may derive an entity as additional information in addition to the intent. Analysis data such as the intent and the entity are transmitted to the DM in the dialogue system.
[0057] The dialog system maps each of the at least one NLU intent to a DM intent (S502). The dialog system maps the NLU intent to a DM intent according to a preset mapping rule for a specific context and condition acquired during a dialog with a user.
[0058] In the case where multiple NLU intents are interpreted as the same intent in a specific context, the preset mapping rules can abstract the multiple NLU intents into a single DM intent. Additionally, when the same NLU intent is interpreted differently depending on a specific context or condition, the preset mapping rules can use the DM intent to refine the user intent.
[0059] The dialog system checks the number of NLU intents ( S504 ).
[0060] When multiple NLU intents are extracted from the user utterance (“No” in S504 ), the dialog system may additionally perform the following stages.
[0061] The dialog system generates a DM intent list (S506). Here, the DM intent list includes DM intents mapped to each of the plurality of NLU intents.
[0062] The dialog system performs multi-intent mapping on a plurality of DM intents included in the DM intent list ( S508 ).
[0063] The dialogue system performs multi-intent mapping according to the matched multi-mapping rules among the preset multi-mapping rules for the conditions corresponding to the patterns in the DM intent list under the specific context. The dialogue system can select a DM intent from the DM intent list, or map multiple DM intents to a new DM intent according to the matched multi-mapping rules.
[0064] The dialogue system may apply the first matched multi-mapping rule among the preset multi-mapping rules to the pattern in the DM intent list. When none of the preset multi-mapping rules for the DM intent list matches, the dialogue system does not generate a result based on the multi-intent mapping. In this regard, as described above, the dialogue system may sequentially execute the actions corresponding to each DM intent.
[0065] Although the steps in each flowchart are described as being performed sequentially, these steps are merely technical concepts that instantiate some embodiments of the present disclosure. Therefore, a person of ordinary skill in the art to which the present disclosure belongs can perform these steps by changing the order described in each figure or by performing two or more steps in parallel. Therefore, the steps in each flowchart are not limited to the time sequence shown.
[0066] The various components of the device or method according to the present disclosure can be implemented as hardware or software, or as a combination of hardware and software. In addition, the function of each component can be implemented as software, and a microprocessor can also be implemented to execute the function of the software corresponding to each component.
[0067] Various implementations of the systems and methods described herein may be implemented by digital electronic circuits, integrated circuits, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include those implemented in one or more computer programs executable on a programmable system. A programmable system includes: at least one programmable processor connected to receive data and instructions from a storage system and to send data and instructions to a storage system; at least one input device; and at least one output device, wherein the programmable processor may be a dedicated processor or a general purpose processor. A computer program (also referred to as a program, software, software application, or code) contains instructions for a programmable processor and is stored in a "computer-readable recording medium."
[0068] The computer-readable recording medium includes any type of recording device on which data that can be read by a computer system can be recorded. Examples of computer-readable recording media include nonvolatile or non-transient media such as ROM, CD-ROM, magnetic tape, floppy disk, memory card, hard disk, optical / magnetic disk, storage device, etc. In addition, the computer-readable recording medium can be distributed in computer systems connected via a network, where computer-readable codes can be stored and executed in a distributed manner.
[0069] Although the embodiments of the present disclosure have been described for illustrative purposes, it should be understood by those skilled in the art that various modifications, additions and substitutions may be made without departing from the concept and scope of the present disclosure. Therefore, for the sake of brevity and clarity, the embodiments of the present disclosure have been described. The scope of the technical concept of the embodiments of the present disclosure is not limited by the examples. Therefore, it should be understood by those skilled in the art that the scope of the present disclosure should not be limited by the embodiments explicitly described above, but by the claims and their equivalents.
[0070] (reference numerals)
[0071] 202: NLU module
[0072] 204: NLG module
[0073] 206: Knowledge Database
[0074] 210: DST module
[0075] 212: DP module
[0076] 220: Intent Mapping Module
[0077] 222: Multi-Intent Mapping Module
[0078] CROSS-REFERENCE TO RELATED APPLICATIONS
[0079] This application claims priority to and the benefit of Korean Patent Application No. 10-2022-0049933, filed on April 22, 2022, the disclosure of each application is incorporated herein by reference in its entirety.
Claims
1. A method for expressing the intention contained in a user's speech by a dialogue system, the method comprising: Extracting at least one natural language understanding (NLU) intent from the utterance of the user; as well as mapping each of the at least one NLU intent to a dialog manager DM intent, Wherein, when extracting multiple NLU intents from the utterance of the user, the method further includes: generating a DM intent list, wherein the DM intent list includes a DM intent mapped to each of the plurality of NLU intents; and Multi-intent mapping is performed on the plurality of DM intents included in the DM intent list.
2. The method according to claim 1, wherein: Mappings to the DM intent include: With respect to the specific context and conditions acquired during the conversation with the user, the NLU intent is mapped to the DM intent according to a preset mapping rule.
3. The method according to claim 2, wherein: When the multiple NLU intents are interpreted as the same intent in the specific context, the preset mapping rule abstracts the multiple NLU intents into a single DM intent.
4. The method according to claim 2, wherein: When the same NLU intent is interpreted differently depending on the specific context or the condition, the preset mapping rule refines the intent of the user using the DM intent.
5. The method according to claim 1, wherein: Executing the multi-intent mapping includes: For a condition corresponding to a pattern of the DM intent list in a specific context, the multi-intent mapping is performed according to a matched multi-mapping rule among preset multi-mapping rules.
6. The method according to claim 5, wherein: Executing the multi-intent mapping includes: A DM intent is selected from the DM intent list according to the matched multi-mapping rule, or the multiple DM intents are mapped to a new DM intent.
7. The method according to claim 5, wherein: The mode of the DM intention list is characterized in that the plurality of DM intentions included in the DM intention list are presented regardless of the order of the user description.
8. The method according to claim 5, wherein: Executing the multi-intent mapping includes: The multi-mapping rule that is matched first among the preset multi-mapping rules is applied to the pattern of the DM intention list.
9. The method according to claim 5, wherein: Executing the multi-intent mapping includes: When none of the preset multi-mapping rules matches the DM intent list, no result is generated according to the multi-intent mapping.
10. A dialogue system, comprising: A natural language understanding (NLU) module, wherein the NLU module is configured to extract at least one NLU intent from the user's speech; an intent mapping module configured to map each of the at least one NLU intent to a dialog manager DM intent, and generate a DM intent list when a plurality of NLU intents are extracted from the utterance of the user, wherein the DM intent list includes the DM intent mapped to each of the plurality of NLU intents; as well as A multi-intent mapping module is configured to perform multi-intent mapping on a plurality of DM intents included in the DM intent list.
11. The dialogue system according to claim 10, wherein: The intent mapping module is configured to map the NLU intent to the DM intent according to a preset mapping rule for a specific context and condition acquired during a conversation with the user.
12. The dialogue system according to claim 10, wherein: The multi-intent mapping module is configured to perform the multi-intent mapping according to a matched multi-mapping rule among preset multi-mapping rules for a condition corresponding to a pattern of the DM intent list in a specific context.
13. A computer-readable recording medium storing instructions, wherein: When executed by a computer, the instructions cause the computer to: Extract at least one Natural Language Understanding (NLU) intent from the user’s utterance; and mapping each of the at least one NLU intent to a dialog manager DM intent, Wherein, when extracting multiple NLU intents from the utterance of the user, the instructions further cause the computer to: generating a DM intent list, wherein the DM intent list includes a DM intent mapped to each of the plurality of NLU intents; and Multi-intent mapping is performed on the plurality of DM intents included in the DM intent list.
