KR102655058B1 - Method and apparatus for providing chatbot service using condition expression generated interface - Google Patents

Abstract

A method and device for providing a chatbot service using a conditional expression generation interface are disclosed. A chatbot service provision method according to an embodiment of the present invention includes providing a conditional expression creation interface, and generating an execution block for processing an intent corresponding to a user conversation through the conditional expression creation interface. Accordingly, it is possible for users to effectively write conditional expressions necessary to run a chatbot, thereby providing an improved chatbot service.

Description

Method and device for providing chatbot service using conditional expression generation interface {METHOD AND APPARATUS FOR PROVIDING CHATBOT SERVICE USING CONDITION EXPRESSION GENERATED INTERFACE}

The present invention relates to a method and device for providing a chatbot service, and particularly to a technology for defining a bot driving process by providing an interface for a chatbot designer to create conditional expressions for bot driving in a work area.

Unless otherwise indicated herein, the material described in this section is not prior art to the claims of this application, and is not admitted to be prior art by inclusion in this section.

A chatbot refers to a computer program or artificial intelligence with the ability to communicate through text or voice. It ranges from simple chatbots that only use word or phrase matching to complex and sophisticated natural language processing technology to conduct natural conversations like humans. There are various levels of chatbots, including chatbots that apply .

A conversational AI agent system using a chatbot identifies the service the user wants in response to the user's input and provides a response. At this time, if the user provides input through a natural sentence, natural language processing technology may be applied to process the input. For this purpose, several example utterance sentences with the same user intent are collected and classified into one intent. By mapping the bot response corresponding to each of these classified intents or defining actions such as calling the service API, it is possible to provide a chatbot service that can process a variety of user natural speech.

When mapping one bot response unit (execution block) to one user's speech unit (intent), as service functions increase, bot responses and corresponding intents also increase. They are assigned to intents, and when classifying intent, multiple intents with slight score differences are generated, increasing the possibility of errors occurring in intent classification.

Even if the number of intents does not increase, if you register various speech patterns in intents to recognize various natural utterances of users, similar patterns will be registered in multiple intents, so the natural language processor can accurately detect similar utterances between different intents. It is difficult to process, so errors are likely to occur in intent classification.

In other words, if utterances are classified according to a situation where intent and bot response units are mapped 1:1, the performance of the natural language processor deteriorates as the number of intents increases or the number of example utterance sentences registered in the intent increases. When utterances are classified to ensure the performance of the natural language processor, all similar utterance patterns are defined for one intent, making the implementation of the block mapped to the intent very complicated and many functions actually provided in one block. As the situation becomes mapped, management difficulties are inevitable.

In addition, most chatbots process the bot's actions by searching for all intents when a user speaks. In this case, when searching for all intents, there are problems with processing the results of the natural language processor and bot responses and actions. A problem arises as the configuration becomes more complex.

Therefore, there is an urgent need for a new chatbot service provision technology that can more freely map intent and bot response units.

Korean Patent Publication No. 10-2019-0006403, published on January 18, 2019 (Name: Voice processing method and system supporting the same)

The purpose of the present invention is to reduce errors in conditional expressions and improve readability by providing an interface that applies conditional grammar based on keyword expressions optimized for slot recognition when setting conditions necessary for running a chatbot.

In addition, the purpose of the present invention is to write conditional expressions concisely by limiting the use of parentheses only to setting priorities, and to recommend meaningful parentheses for parentheses written by the bot designer, thereby reducing the number of input steps and eliminating meaningless ones. The use of parentheses prevents conditional expressions from being written in a complicated manner.

In addition, the purpose of the present invention is to enable N:N (N is a natural number) mapping between intents classified appropriately for a natural language processor and execution blocks that process the intents, so that several different utterances can be combined into one. It allows you to execute a function and, conversely, to perform a completely different function with similar utterances.

Additionally, the purpose of the present invention is to enable various intent-execution block mapping relationships by using an intent mapping list that includes an intent identifier and a slot condition for each intent.

In addition, the object of the present invention is not limited to the objects described above, and it is obvious that another object may be derived from the following description.

A method of providing a chatbot service according to an embodiment of the present invention to achieve the above object includes providing a conditional expression creation interface; It includes the step of creating an execution block for processing an intent corresponding to a user conversation through the conditional expression creation interface.

At this time, at least one condition included in the conditional expression created through the conditional expression creation interface may be composed of a combination of a condition variable, a comparison operator, and a condition value.

At this time, the conditional expression creation interface may recommend at least one comparison operator in consideration of the input condition variable and may recommend at least one condition value in consideration of the input comparison operator.

At this time, the conditional expression creation interface may recommend a pair of parentheses entered at an arbitrary position in the conditional expression.

At this time, the conditional expression creation interface may not use parentheses when calling a function.

At this time, the conditional expression creation interface includes a first type parenthesis that can change the result of the conditional expression, a second type parenthesis that does not change the result of the conditional expression but is judged to be unnecessary, and the first type parenthesis that does not change the result of the conditional expression. At least one of a type parenthesis and a third type parenthesis that does not correspond to the second type parenthesis may be recommended as a pair of the parentheses.

At this time, the method of searching for the first type of parentheses may include using a conditional expression result comparison table.

At this time, before recommending a pair of input parentheses, the conditional expression creation interface determines whether the arbitrary position is a position where parentheses can be input by considering the type of the input parentheses, and determines whether the arbitrary position is a position where parentheses can be input. If the position is not correct, the random position can be moved and corrected based on the type of parentheses entered.

At this time, the conditional expression creation interface may display the first type parenthesis, the second type parenthesis, and the third type parenthesis so that they are visually distinguished.

Additionally, a chatbot service providing device according to an embodiment of the present invention includes one or more processors; and an execution memory storing at least one program executed by the one or more processors. It includes, wherein the at least one program provides a conditional expression creation interface and generates an execution block for processing an intent corresponding to a user conversation through the conditional expression creation interface.

At this time, at least one condition included in the conditional expression created through the conditional expression creation interface may be composed of a combination of a condition variable, a comparison operator, and a condition value.

At this time, the conditional expression creation interface may recommend at least one comparison operator in consideration of the input condition variable and may recommend at least one condition value in consideration of the input comparison operator.

At this time, the conditional expression creation interface may recommend a pair of parentheses entered at an arbitrary position in the conditional expression.

At this time, the conditional expression creation interface may not use parentheses when calling a function.

At this time, the conditional expression creation interface includes a first type parenthesis that can change the result of the conditional expression, a second type parenthesis that does not change the result of the conditional expression but is judged to be unnecessary, and the first type parenthesis that does not change the result of the conditional expression. At least one of a type parenthesis and a third type parenthesis that does not correspond to the second type parenthesis may be recommended as a pair of the parentheses.

At this time, the method of searching for the first type of parentheses may include using a conditional result comparison table.

At this time, before recommending a pair of input parentheses, the conditional expression creation interface determines whether the arbitrary position is a position where parentheses can be input by considering the type of the input parentheses, and determines whether the arbitrary position is a position where parentheses can be input. If the position is not correct, the random position can be moved and corrected based on the type of parentheses entered.

At this time, the conditional expression creation interface may display the first type parenthesis, the second type parenthesis, and the third type parenthesis so that they are visually distinguished.

According to the present invention, errors in conditional expressions can be reduced and readability can be improved by providing an interface that applies a keyword expression-based conditional grammar optimized for slot recognition when setting the conditions necessary for running a chatbot.

In addition, the present invention uses parentheses only to set priorities to create conditional expressions concisely, and recommends meaningful parentheses for parentheses written by bot designers, thereby reducing the number of input steps and eliminating the use of meaningless parentheses. This can prevent conditional expressions from being written in a complicated manner.

In addition, the present invention enables N: N (N is a natural number) mapping between intents classified appropriately for a natural language processor and execution blocks that process the intents, allowing one function to be executed with various different utterances. Conversely, similar utterances can perform completely different functions.

Additionally, the present invention makes it possible to define various intent-execution block mapping relationships using an intent mapping list that includes an intent identifier and a slot condition for each intent.

The effects of the present embodiments are not limited to the effects mentioned above, and other effects not mentioned may be clearly understood by those skilled in the art from the description of the claims.

Figure 1 is an operation flowchart showing a method of providing a chatbot service according to an embodiment of the present invention.
Figure 2 is a diagram showing an example of a conditional expression creation interface according to the present invention.
Figure 3 is a diagram showing an example of a conditional expression tokenized and analyzed according to the present invention.
Figures 4 and 5 are diagrams showing an example of recommended pairs of parentheses according to the present invention.
Figure 6 is a diagram showing an example of a conditional result comparison table according to the present invention.
Figures 7 and 8 are diagrams showing an example of position correction of input parentheses and pair recommendation of input parentheses according to the present invention.
Figure 9 is a block diagram showing a chatbot service providing device according to an embodiment of the present invention.
Figure 10 is a block diagram showing a chatbot service system according to an embodiment of the present invention.
Figure 11 is a block diagram of a 'coffee ordering' scenario according to an embodiment of the present invention.
FIG. 12 is a diagram illustrating an example of intents used in the scenario shown in FIG. 11.
Figure 13 is a block diagram showing the connection between a plurality of intents and a plurality of execution blocks.
FIG. 14 is a diagram illustrating an example of execution block coffee selection shown in FIG. 13.
FIG. 15 is a diagram illustrating an example of hot ice confirmation of an execution block shown in FIG. 13.
FIG. 16 is a diagram illustrating an example of execution block takeout confirmation shown in FIG. 13.
Figure 17 is a block diagram showing the connection relationship between execution blocks shown in Figures 13 to 16.
Figure 18 is a diagram showing an example of connecting blocks by defining a user response waiting state in an execution block.
Figure 19 is a diagram showing an example of mapping an intent defined for each speech pattern to a block.
Figure 20 is a diagram showing a case of 1:1 mapping between an intent and an execution block.
Figure 21 is a diagram showing an example of intent classification and block mapping according to an embodiment of the present invention.
FIG. 22 is a diagram showing execution blocks corresponding to the responses shown in FIG. 21.
Figure 23 is a diagram showing an example of priority sorting according to the inclusion relationship of conditions corresponding to the coffee ordering scenarios illustrated in Figures 11 to 17.
Figure 24 is a block diagram showing an example of modules required to run a chatbot.
Figure 25 is a diagram showing the process of searching intents and blocks while running a chatbot.
Figure 26 is an operation flowchart showing a chatbot service provision method according to another embodiment of the present invention.
Figure 27 is a graph showing natural language processor performance as the number of speech patterns increases.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. However, detailed descriptions of known functions or configurations that may obscure the gist of the present invention are omitted in the following description and attached drawings. Additionally, it should be noted that the same components throughout the drawings are indicated by the same reference numerals whenever possible.

Terms or words used in the specification and claims described below should not be construed as limited to their usual or dictionary meanings, and the inventor should use the concept of terminology appropriately to explain his/her invention in the best way. It must be interpreted as meaning and concept consistent with the technical idea of the present invention based on the principle that it can be defined clearly. Therefore, the embodiments described in this specification and the configuration shown in the drawings are only one of the most preferred embodiments of the present invention, and do not represent the entire technical idea of the present invention, so at the time of filing the present application, various alternatives may be used to replace them. It should be understood that equivalents and variations may exist. Additionally, terms such as first and second are used to describe various components, and are used only to distinguish one component from other components and are not used to limit the components.

The present invention relates to a technology for setting various conditions necessary for bot operation to provide a chatbot service and defining a chatbot operation process through these.

Below, the operation flow for defining the bot driving processor will be described in more detail.

Figure 1 is an operation flowchart showing a method of providing a chatbot service according to an embodiment of the present invention.

Referring to Figure 1, the chatbot service provision method according to an embodiment of the present invention provides a conditional expression creation interface (S110).

At this time, the conditional expression creation interface according to an embodiment of the present invention can generate a conditional expression by applying a specialized conditional grammar to compare slots mainly used in chatbot builders or expressions that take readability into account.

For example, the conditional expression for the case where the "membership number" slot is all numbers and the "nickname" slot contains "member" or does not contain "guest" can be expressed in the existing JavaScript language grammar as follows.

[Example of conditional expression expressed in JavaScript language grammar]

/^[0-9]+$/.test(slot_value("Membership number")) && (slot_value("Nickname").includes("member") || !slot_value("Nickname").includes("guest "))

When looking at conditional expressions expressed in JavaScript language grammar, the parentheses used for function calls and the parentheses used to indicate condition priorities are mixed, causing a problem of poor readability. Additionally, because the condition value is written before the condition variable, it is difficult to quickly determine which condition variable the value is.

In the present invention, in order to solve this problem, the above conditional expression can be expressed as follows by applying a specialized conditional grammar to improve readability.

[Examples of conditional expressions expressed using the conditional grammar proposed by the present invention]

$member number regex "^[0-9]+$" and ($nickname not contain "guest" or $nickname contain "member")

At this time, at least one condition included in the conditional expression created through the conditional expression creation interface may be composed of a combination of a condition variable, a comparison operator, and a condition value.

To explain the conditions included in the above conditional expression as an example, the combination of "$member number" corresponding to the condition variable, "regex" corresponding to the comparison operator, and "^[0-9]+$" corresponding to the condition value. You can configure one condition, $member number regex "^[0-9]+$".

At this time, the conditional expression creation interface can create conditions in the order corresponding to the condition variable, comparison operator, and condition value.

If we explain the conditions included in the above conditional expression as an example, "$member number" corresponding to the condition variable is written first, then "regex" is written in place of the comparison operator, and finally "^[0" corresponding to the condition value. You can see that “-9]+$” is displayed.

In this way, the readability of the conditional expression can be improved by creating the conditions included in the conditional expression in an order corresponding to the condition variable, comparison operator, and condition value.

At this time, in the conditional grammar proposed in the present invention, comparison functions can also be expressed in text and used like comparison operators. Taking the above conditional expression as an example, the regular expression test method used in the JavaScript language grammar is written as "regex" according to the conditional expression grammar according to an embodiment of the present invention, and the includes function used in the JavaScript language grammar is an example of the present invention. It can be confirmed that “contain” is displayed according to the conditional grammar according to the embodiment. In addition, functions such as "empty" and "exist" can also be used.

Additionally, in the conditional grammar according to one embodiment of the present invention, the “not” keyword is used to naturally display negative conditions. To explain the above conditional expression as an example, when expressing the case where the "alias" slot does not include "guest", in JavaScript language grammar, it is written as '!slot_value("alias").includes("guest"))'. Alternatively, it can be written as '$alias not contain "guest"' according to the conditional grammar according to an embodiment of the present invention.

At this time, in the conditional grammar according to an embodiment of the present invention, “not exist” and “not empty” are not used, but “empty” and “exist” are used to minimize the use of negative condition marks.

To describe the conditional expression rules in the conditional grammar according to an embodiment of the present invention in more detail, the slot name is $slot, the slot value is "value", and the conditional operations are ==, !=, <, >, <=, It can be expressed as >=, contain, not contain, regex, empty, exist, etc., and logical operations can be used as shown in [Table 1].

comparison operator Conditional expression sample matches: $slot == "value" does not match $slot != "value" greater than $slot > "value greater than or equal to $slot >= "value" less than $slot < "value Less than or equal to $slot <= "value" Included $slot contains "value" Not included $slot　not contain　"value" regular expression $slot regex "[0-9]+" it does not exist $slot empty exists $slot exists

Additionally, a conditional expression written through the conditional expression writing interface according to an embodiment of the present invention can be expressed and used as, for example, the grammar rules (parser grammar rules) shown in [Table 2].

Element Expression and usage examples or_expression or_expression 'or' and_expression
| and_expression and_expression and_expression 'and' compare_expression
| compare_expression compare_expression parameter COMPARE_OPERATOR value
| parameter 'regex' value
| parameter 'contain' value
| parameter 'not''contain' value
| parameter 'exist'
| parameter 'empty'
| '(' expression ')'
| value value STRING
| NUMBER
| PARAMETER COMPARE_OPERATOR '==' | '!= ' | '<=' | '>=' | '<' | '>' STRING Plain string surrounded by quotes: /"([^\\\"]|\\.)*"/ NUMBER Regular numbers (including decimals): /[-+]? (\d*\.\d+|\d+)/ PARAMETER Variables starting with $ (including Korean): /\$[a-zA-Z\d is- \-_\.]+/

Hereinafter, for a clear understanding of the invention, the conditional expression written according to an embodiment of the present invention will be interpreted based on the above description.

For example, in the conditional expression corresponding to ("$svc_cat == "ok cashback points" or "$svc_cat == "points") and "$action" == "accumulation", the condition value of the condition variable "svc_cat" is This may mean a case where the condition value of the condition variable "action" corresponds to "accumulation" while corresponding to "ok cashback points" or "points".

For another example, the conditional expression corresponding to ($svc_cat=="ok cashback points" or $svc_cat regex "[0-9]+") and $action=="Accumulation" is the condition value of the condition variable "svc_cat" This may mean a case where the condition value of the condition variable "action" corresponds to "accumulation" while corresponding to "ok cashback points" or all numbers.

For another example, the conditional expression corresponding to $svc_cat=="ok cashback points" and $location empty and $action =="accumulated" is that the condition value of the condition variable "svc_cat" is "ok cashback points" and the condition variable This may mean a case where the condition value of “location” does not exist and the condition value of the condition variable “action” corresponds to “accumulation.”

At this time, the conditional expression creation interface may support a text-based direct input method and a button-based recommended input method.

For example, in the text-based direct input method, the bot designer enters a conditional expression by typing directly into the input window, and in the button-based recommended input method, the bot designer selects the input button provided in the conditional expression creation interface with the mouse or other selection method. You can click and enter.

The existing text-based UI (User Interface) for entering conditional expressions is a mixture of functional calls, so the positions of variables are not constant, and the parentheses used in functions and the parentheses that group conditions for condition priority are mixed together to distinguish them. The downside is that it is difficult to do.

In addition, the UI (User Interface), which allows conditional expression input without text errors through button input, often does not support nesting even if and/or conditions are possible, so there was inconvenience in writing conditional expressions.

In order to solve these problems, the present invention seeks to provide a conditional expression creation interface that allows direct text input as well as input using buttons, thereby preventing incomplete conditional expressions or grammatical errors from occurring.

Therefore, the conditional expression creation interface according to an embodiment of the present invention may correspond to an interface that allows direct input according to a predefined conditional expression grammar and at the same time supports conditional expression input through mouse control.

At this time, the conditional expression creation interface may recommend at least one comparison operator in consideration of the input condition variable and may recommend at least one condition value in consideration of the input comparison operator.

For example, referring to the conditional expression creation interface according to an embodiment of the present invention shown in Figure 2, when the bot designer inputs a slot name corresponding to a condition variable in the input window 220, the entered slot name is taken into consideration. Therefore, at least one comparison operator that can be applied to this can be automatically recommended. Additionally, assuming that the bot designer has selected and entered a comparison operator through recommendation, at least one slot value corresponding to the condition value can be automatically recommended by considering the entered slot name and comparison operator.

Additionally, the conditional expression creation interface can automatically recommend at least one slot name that can be applied when the bot designer focuses the condition variable area of the input window 220 with the mouse.

Additionally, the conditional expression creation interface can automatically recommend at least one comparison operator that can be applied when the bot designer focuses the comparison operator area of the input window 220 with the mouse.

Additionally, the condition expression creation interface can automatically recommend at least one slot value that can be applied when the bot designer focuses the condition value area of the input window 220 with the mouse.

At this time, depending on which area is the focused area, recommendation may be performed by taking into account what values have been entered in other areas applied to the input window 220.

In addition, the conditional expression creation interface recommends at least one slot name that can be applied when the bot designer enters "$" in the input window 220 using a text-based direct input method, and enters the first letter of the comparison operator to select the entered slot name. By recommending at least one comparison operator that starts with a letter and entering an opening double quote ("), you can recommend at least one slot value that can be applied.

Additionally, the conditional expression creation interface can recommend a pair of parentheses entered at an arbitrary position in the conditional expression.

At this time, the conditional expression creation interface may not use parentheses when calling a function.

For example, in order to prevent confusion between the parentheses used in function calls and the parentheses indicating condition priorities, the conditional expression creation interface according to one embodiment of the present invention uses parentheses only when indicating condition priorities, as in the example below. You can use it. For this purpose, the function that retrieves the slot value corresponding to the condition variable can be expressed in a different way, such as "$slot name."

[Examples of conditional sentences expressed in the conditional grammar proposed by the present invention]

$member number regex "^[0-9]+$" and ($nickname not contain "guest" or $nickname contain "member")

At this time, the conditional expression creation interface according to an embodiment of the present invention may first perform an analysis process on the conditional expression in order to recommend a pair of parentheses entered at an arbitrary position in the conditional expression.

For example, assuming that the bot designer enters a conditional expression and then enters an opening parenthesis to determine condition priority, the conditional expression creation interface can analyze the conditional expression as shown in FIG. 3. At this time, in order to perform analysis in units in which parentheses can be entered, conditional expressions can be classified into logical operators such as and and or and comparison syntax units, and those with the same comparison syntax can be tokenized by considering them as the same condition.

For example, $type == "song" and $action == "play" or $type == "album" and $action == "play" is a conditional expression that includes four comparison operation conditions, but A: $ If you define type == "song", B: $action == "play", C: $type == "album", it can be simplified to A and B or C and B.

Once the conditional expression analysis is completed in this way, all positions of recommendable parentheses can be searched according to the type of parentheses entered, as shown in Figures 4 and 5.

For example, referring to Figure 4, when the input parenthesis is an open parenthesis, the right side of each condition located behind the conditional expression based on the input position of the open parenthesis is the recommended closing parenthesis position (410, 420, 430). It can be searched by .

For another example, referring to FIG. 5, when the input parenthesis is a closing parenthesis, the left side of each condition located in front of the conditional expression based on the input position of the closing parenthesis is the recommended open parenthesis position (510, 520, 530). ) can be searched.

At this time, first-type parentheses that can change the result of the conditional expression, second-type parentheses that do not change the result of the conditional expression but are judged unnecessary, and first-type parentheses and second-type parentheses that do not change the result of the conditional expression. At least one of the non-applicable third type parentheses can be recommended as a pair of parentheses.

At this time, the first type of parentheses may be meaningful parentheses that can change the result of the conditional expression, the second type of parentheses may be unnecessary parentheses that do not need to be present, and the third type of parentheses may be parentheses for readability of the conditional expression. can do.

For example, if the following case applies, it can be judged to be a type 2 parenthesis.

- Full parenthesis surrounding the entire conditional expression

(A and B or C and B)

- Unary parentheses surrounding only one condition

A and B or (C) and B

- Redundant parentheses that surround only parenthetical expressions

A and ((B or C)) and B

At this time, the method of searching for the first type of parentheses may include using a conditional result comparison table.

At this time, the conditional expression result comparison table is a True/False table created based on the conditional expression before the bot designer inputs the parentheses, and a conditional expression with the pair of parentheses recommended by the conditional expression creation interface applied to the parentheses entered by the bot designer. It may correspond to a True/False table created based on , expressed in a comparable form. If there is a difference in the results between the two True/False tables included in the conditional expression result comparison table, it can be determined that the parentheses recommended by the conditional expression creation interface are meaningful parentheses that are the first type of parentheses.

For example, referring to Figure 6, the True/False table shown on the left corresponds to one created based on 'A and B or C and B', which is the conditional expression in the state before the bot designer enters the parentheses, and on the right The True/False table shown may correspond to the conditional expression creation interface created based on the conditional expression 'A and (B or C) and B' with the recommended pair of parentheses applied to the parentheses entered by the bot designer. there is. At this time, in the conditional expression result comparison table shown in FIG. 6, there is a different value 610 among the result values between the two True/False tables, so the pair of parentheses recommended by the conditional expression creation interface is the first type parenthesis. You can judge.

In this way, all first type parentheses can be searched by performing analysis on all pairs of parentheses recommended by the conditional expression creation interface.

In the present invention, the method of using the conditional expression result comparison table to search for the first type of parentheses is explained as an example, but the method of searching for the first type of parentheses is not limited to this, and the recommended pair of parentheses can change the result of the conditional expression. Various methods can be used to determine whether a pair of parentheses is meaningful.

At this time, before recommending a pair of input parentheses, the conditional expression creation interface can determine whether an arbitrary position is a position where parentheses can be input by considering the type of input parenthesis.

At this time, the conditional expression creation interface can correct the position by moving it to a certain position based on the type of parentheses entered if the arbitrary position is not a position where parentheses can be input.

For example, referring to FIG. 7, it can be assumed that the bot designer entered an open parenthesis at the indicated position 710 in the conditional expression. At this time, since the displayed position 710 corresponds to the right side of the condition variable constituting the condition or the left side of the comparison operator, it can be determined that it is not a position where the opening parenthesis can be entered.

Therefore, the conditional expression creation interface can be corrected by moving the position of the opening parenthesis entered by the bot designer so that the opening parenthesis is entered at the corrected position 810, as shown in FIG. 8.

At this time, considering the type of parentheses entered by the bot designer, the opening parenthesis "(" can be corrected by moving its position to the front of the conditional expression, and the closing parenthesis ")" can be corrected by moving it to the back of the conditional expression. .

At this time, the conditional expression creation interface may display the first type parenthesis, the second type parenthesis, and the third type parenthesis so that they are visually distinguished.

For example, you could prioritize type 1 brackets as 1st, type 3 brackets as 2nd, type 2 brackets as 3rd, and indicate the priority number corresponding to the recommended pair of brackets. You can.

For another example, as shown in Figure 8, the first type parentheses 820 and 830, which have a higher priority, may be shown with the most emphasis, and the third type brackets 840, corresponding to the next priority, may be shown with less emphasis. there is. At this time, as shown in FIG. 8, when multiple parentheses of the same type are recommended, they may be highlighted and displayed in close order based on the position of the parenthesis entered by the bot designer. Additionally, if the 1st and 2nd priority types of parentheses are recommended, the 3rd priority type of parentheses may not be recommended for readability.

Below, with reference to FIG. 8, the process of recommending a pair of parentheses, selecting a pair of recommended parentheses, and generating a conditional expression will be described in detail.

First, the conditional expression creation interface can search for all types of parentheses that can be recommended for the input opening parenthesis that has been moved to the corrected position 810 through conditional expression analysis.

Assuming that the first type brackets 820, 830 and the third type brackets 840 are searched as shown in Figure 8, first of the first type brackets 820, 830, the first type closest to the input brackets It can be recommended to display the type brackets (820) with the most emphasis.

At this time, when the bot designer selects the first type parenthesis 820 by touching and clicking with a point device or using left and right arrow keys and tab cycle input with a keyboard device, the opening parenthesis entered at the corrected position 810 is selected. The pair is confirmed as the first type bracket 820 and the recommendation can be ended.

However, because the pair of parentheses that the bot designer intended to enter may not be the first type parenthesis 820, the conditional expression interface replaces the first type parenthesis 830, which is closest to the next input parenthesis, with the first type parenthesis 820. You can express that a choice is possible by showing it with less emphasis than ).

Likewise, if the bot designer selects the first type parenthesis 830, the pair of opening parentheses entered in the corrected position 810 is confirmed as the first type parenthesis 830 and the recommendation may be terminated.

In this way, when there are multiple parentheses of the same type, the conditional expression creation interface according to an embodiment of the present invention can preferentially recommend parentheses that exist in a nearby position based on the position of the input parenthesis.

In addition, referring to FIG. 8, the conditional expression creation interface according to an embodiment of the present invention may indicate that selection is possible while recommending the second type parentheses 840 that are judged to be unnecessary, but the first type parentheses or the third type parentheses 840 may also be selected. It can be displayed so that it is not emphasized compared to the type brackets.

As such, the conditional expression creation interface according to an embodiment of the present invention can validate whether the conditional expression entered by the bot designer has been entered correctly according to the grammar, and if the conditional expression is entered abnormally, the bot designer can re-enter it. You can do it.

For example, whether the parentheses included in the conditional expression are entered in pairs, whether the conditional expression is written in accordance with the notation format for each operator defined in the grammar, and whether the condition variable, comparison operator, and condition value are entered correctly and included in the conditional expression. You can also check whether double quotation marks are entered in pairs.

Additionally, the chatbot service provision method according to an embodiment of the present invention creates an execution block for processing an intent corresponding to a user conversation through a conditional expression creation interface (S120).

For example, the conditional expression creation interface includes at least one of the following: an intent mapping connection condition for mapping between an intent and an execution block, an execution condition for executing a chatbot action defined in an execution block, and a prompt message execution condition for slot filling. It can be provided when creating a conditional expression for running a bot.

Through this chatbot service provision method, errors in conditional expressions can be reduced and readability improved by providing an interface that applies conditional grammar based on keyword expressions optimized for slot recognition when setting the conditions necessary for running the chatbot.

In addition, the use of parentheses is limited to setting priorities to create conditional expressions concisely, and by recommending meaningful parentheses for parentheses written by the bot designer, the number of input steps is reduced and the conditional expression is reduced by using meaningless parentheses. Complex writing can be prevented.

Figure 9 is a block diagram showing a chatbot service providing device according to an embodiment of the present invention.

Referring to FIG. 9, the chatbot service providing device according to an embodiment of the present invention includes a processor 920 and an execution memory 930. Depending on the embodiment, the chatbot service providing device may further include a communication unit 910.

The communication unit 910 may serve to transmit information related to the chatbot service with a number of other devices through a communication network, such as a network. At this time, the network provides a path for transmitting data between devices, and is a concept that encompasses both existing networks and networks that can be developed in the future.

For example, the network is an IP network that provides large-capacity data transmission and reception services and uninterrupted data services through Internet Protocol (IP), and All IP is an IP network structure that integrates different networks based on IP. ) network, etc., and may be a wired network, Wibro (Wireless Broadband) network, 3rd generation mobile communication network including WCDMA, 3.5 generation mobile communication network including HSDPA (High Speed Downlink Packet Access) network and LTE network, and 4 including LTE advanced. It can be achieved by combining one or more of a household mobile communication network, a satellite communication network, and a Wi-Fi network.

For example, the network is a wired and wireless local area network that provides communication of various information devices within a limited area, a mobile communication network that provides communication between moving objects and between moving objects and the outside of the moving object, and a network that provides communication between earth stations using satellites. It may be a satellite communication network, a wired or wireless communication network, or a combination of two or more. Meanwhile, the network transmission method standard is not limited to existing transmission method standards and may include all transmission method standards to be developed in the future.

The execution memory 930 stores at least one program executed by the processor 920.

At this time, at least one program provides a conditional expression creation interface and creates an execution block for processing the intent corresponding to the user conversation through the conditional expression creation interface.

At this time, at least one condition included in the conditional expression created through the conditional expression creation interface may be composed of a combination of a condition variable, a comparison operator, and a condition value.

At this time, the conditional expression creation interface may recommend at least one comparison operator in consideration of the input condition variable and may recommend at least one condition value in consideration of the input comparison operator.

At this time, the conditional expression creation interface may recommend a pair of parentheses entered at an arbitrary position in the conditional expression.

At this time, the conditional expression creation interface may not use parentheses when calling a function.

At this time, the conditional expression creation interface includes a first type parenthesis that can change the result of the conditional expression, a second type parenthesis that does not change the result of the conditional expression but is judged to be unnecessary, and the first type parenthesis that does not change the result of the conditional expression. At least one of a type parenthesis and a third type parenthesis that does not correspond to the second type parenthesis may be recommended as a pair of the parentheses.

At this time, the method of searching for the first type of parentheses may include using a conditional result comparison table.

At this time, before recommending a pair of input parentheses, the conditional expression creation interface determines whether the arbitrary position is a position where parentheses can be input by considering the type of the input parentheses, and determines whether the arbitrary position is a position where parentheses can be input. If the position is not correct, the random position can be moved and corrected based on the type of parentheses entered.

At this time, the conditional expression creation interface may display the first type parenthesis, the second type parenthesis, and the third type parenthesis so that they are visually distinguished.

The device shown in FIG. 9 can receive various information such as number and character information, and transmit signals related to setting various functions and controlling the functions of the device to the processor 920 through the input unit. Additionally, the input unit may be configured to include at least one of a keypad and a touchpad that generate an input signal according to the user's touch or manipulation. At this time, the input unit of the device is configured in the form of a touch panel (or touch screen) together with the display unit and can perform input and display functions simultaneously. Additionally, the input unit may use any type of input means that may be developed in the future, in addition to input devices such as a keyboard, keypad, mouse, joystick, etc.

Additionally, the display unit can display information about a series of operation states and operation results that occur while performing the function of the device. Additionally, the display unit can display menus of the device and user data entered by the user. Here, the display unit includes a liquid crystal display (LCD), an ultra-thin liquid crystal display (TFT-LCD, Thin Film Transistor LCD), a light emitting diode (LED), an organic light emitting diode (OLED, Organic LED), It can be composed of active organic light emitting diode (AMOLED, Active Matrix OLED), Retina Display, flexible display, and 3-dimensional display. At this time, when the display unit is configured in the form of a touch screen, the display unit can perform some or all of the functions of the input unit.

Additionally, the storage unit of the device is a device for storing data, includes a main memory and an auxiliary memory, and can store application programs necessary for the functional operation of the device. This storage unit may largely include a program area and a data area. Here, when the device activates each function in response to the user's request, the device executes the corresponding application programs under the control of the processor 920 to provide each function.

Additionally, the communication unit 910 may perform a function for transmitting and receiving data through a network. Here, the communication unit 910 may include an RF transmitting unit that up-converts and amplifies the frequency of the transmitted signal and an RF receiving unit that low-noise amplifies the received signal and down-converts the frequency. This communication unit 1810 may include at least one of a wireless communication module and a wired communication module. In addition, the wireless communication module is configured to transmit and receive data according to a wireless communication method, and when the device uses wireless communication, data is transmitted to the device using any one of the wireless network communication module, wireless LAN communication module, and wireless fan communication module. It can be sent and received. Additionally, the wired communication module is for transmitting and receiving data by wire. The wired communication module can connect to the network through a wire and transmit and receive data to the device. That is, the device connects to the network using a wireless communication module or a wired communication module and can transmit and receive data with other devices through the network.

Additionally, the processor 920 may be a processing device that drives an operating system (OS) and each component.

With the above-described configuration, the chatbot service providing device according to an embodiment of the present invention provides an interface that applies a keyword expression-based conditional grammar optimized for slot recognition when setting the conditions necessary for running the chatbot, thereby preventing errors in the conditional expression. It can be reduced and readability can be improved.

In addition, the chatbot service providing device according to an embodiment of the present invention uses the use of parentheses only to set priorities to create conditional expressions concisely, and recommends meaningful parentheses for parentheses written by the bot designer. You can reduce the number of input steps and prevent conditional expressions from being written in a complicated manner due to the use of meaningless parentheses.

In addition, the present invention classifies intents corresponding to natural language input according to similar properties according to natural language processing performed by a conversational AI agent system, and maps a plurality of classified intents to blocks that are processing units of the bot. It's about technology.

In order to provide useful chatbot services, it is necessary to continuously improve natural language processing capabilities for terms by business or service domain and processing capabilities for provided functions or tasks.

To this end, it is necessary to train a natural language processor so that various domain-specific terms and sentences can be processed by the natural language processor. At this time, the natural language processor must classify intents to prevent conflicts between similar utterances. In other words, it is necessary to define intent from the perspective of Natural Language Understanding (NLU), rather than defining intent only by function or task unit.

Meanwhile, in order for functions or tasks to be processed effectively, detailed processing steps or flows must be defined as execution blocks, and the connection relationship (sequence relationship) of the defined execution blocks needs to be defined. At this time, it is preferable that the connection relationship be defined in the form of a graph to freely connect multiple execution blocks rather than in the form of a list or tree. At this time, it is preferable that execution blocks are defined appropriately for the functions provided by the service rather than from the perspective of natural language processing.

In this way, only when it is possible to freely (N:N) map intents and blocks defined in a form suitable for each purpose, can a chatbot service that meets the user's intent be efficiently provided. In other words, one intent can be used in multiple blocks, and one block must be able to handle multiple intents, and for this, intents and blocks must be able to be mapped N:N rather than 1:1 mapping. .

When processing a large amount of data where a large amount of speech patterns are classified into various intents, conflicts between similar utterances occur and verification and processing are required.

At this time, similar speech patterns can be grouped into the same intent and conditions can be given based on the slot to enable different bot responses.

In addition, by making it possible to map multiple user speech units to one bot response unit (execution block) so that they can be processed with the same response even though they are not similar utterances, it is possible to process them with the same response even if the utterance type is different. This allows efficient intent management.

Hereinafter, the chatbot service is a concept that covers not only a voice service based on the user's voice, but also a messenger service based on text input from the user.

The design and implementation of a chatbot can be broadly divided into three parts: chatbot builder, Natural Language Understanding (NLU), and Dialog Management (DM). At this time, the chatbot builder structures and defines the data for DM to operate and provides it, and DM processes bot response based on this data.

A chatbot process may require two work areas to run. One is a work area where the bot designer defines the bot drive process, and the other is an execution area where the defined process is executed.

Figure 10 is a block diagram showing a chatbot service system according to an embodiment of the present invention.

Referring to FIG. 10, the chatbot service system includes a chatbot 1010, a natural language processor 1020, a chatbot resource database 1030, and a chatbot builder 1040.

The chatbot 1010 receives a conversational sentence from the user, generates a response, and provides a chatbot response to the user.

The natural language processor 1020 receives speech sentences from the chatbot and intent definition information from the chatbot resource database 1030, generates intent classification results and slot extraction results, and provides them to the chatbot 1010.

The chatbot resource database 1030 receives intent definition information, block information, response information, and scenario information from the chatbot builder, provides block information, response information, and scenario information to the chatbot 1010, and natural language processor 1020. Provides intent definition information.

The chatbot builder 1040 receives input about chatbot resources from the bot designer, creates a chatbot 1010, and provides intent definition information, block information, response information, and scenario information to the chatbot resource database 1030. .

In the chatbot builder (1040), learn utterance sentences and slots, which are important information for bot processing, are defined as intents, map the defined intents to execution blocks, set conditions in the execution blocks, and perform the bot's actions accordingly. It can be defined. Additionally, blocks that can be processed in the response waiting state are set in the chatbot builder 1040, and these blocks can be defined as a scenario in the form of a graph.

At this time, the bot designer works to define the bot driving process through the chatbot builder 1040. First, a block diagram in which bot scenarios for major work processes are defined in graph form may be provided.

Figure 11 is a block diagram of a 'coffee ordering' scenario according to an embodiment of the present invention.

Referring to Figure 11, in the block speech scenario of the coffee ordering chatbot, the user's speech “I will order coffee” is first received (1111).

Additionally, in the block speech scenario of the coffee ordering chatbot, the chatbot executes the speech “What kind of coffee would you like?” (1121).

In response to the chatbot's utterance, in the block utterance scenario of the coffee ordering chatbot, the user utters the name of the coffee desired (1112).

In response to the user's utterance, in the block utterance scenario of the coffee ordering chatbot, the chatbot executes the utterance "Would you like something warm or iced?" (1122).

In response to the chatbot's speech, in the block speech scenario of the coffee ordering chatbot, the user selects hot or ice (1113).

In response to the user's utterance, in the block utterance scenario of the coffee ordering chatbot, the chatbot executes the utterance "Would you like to eat it?" (1123).

In response to the chatbot's speech, in the block speech scenario of the coffee ordering chatbot, the user selects dine-in or takeout (1114).

When all three slots of $(coffee name), $(hot ice), and $(takeout) are entered, in the block speech scenario of the coffee ordering chatbot, the chatbot executes an speech to confirm the order (1124).

In response to the chatbot's utterance, in the block utterance scenario of the coffee ordering chatbot, the user utters 'yes' (1115) or 'no' (1116), and if the user utters 'yes', the chatbot executes the coffee order. Executes a custom action (1125) and executes an utterance to confirm the order details (1126). If the user utters 'No', the chatbot executes the utterance "Cancel the order" (1126). 1127), then the operation ends.

In Figure 11, "User" represents the user's utterance, "templete" represents the chatbot's utterance, and "Custom Action" represents the bot's action processing defined by the execution block.

In the example shown in Figure 11, $ (coffee name), $ (hot ice), $ (takeout), and $ (order number) correspond to required slots, and although not explicitly shown in Figure 11, coffee Option slots can be used more in the block utterance scenario of an ordering chatbot. At this time, the required slot and/or optional slot may be created from the user's utterance, may be a slot indicating the status of another chatbot system, or may be created according to conditions in a block.

Based on the diagram shown in FIG. 11, all speech patterns that can be processed in each block can be collected, and the patterns can be grouped in a form that is easy for a natural language processor to classify. In addition, when the intent is classified based on the grouped information and the intent is set by tagging slots for each speech pattern, the speech patterns based on the grouped and classified intent are used as learning data from a natural language processor (NLU) according to the learning intention. It can be used as

A scenario can be viewed as connecting blocks in a graph, and a list of intents that can be processed can be defined in each block.

FIG. 12 is a diagram illustrating an example of intents used in the scenario shown in FIG. 11.

Referring to FIG. 12, the intents used in the coffee ordering scenario include intent order_coffee (1210), intent order_hotice (1220), intent order_takeout (1230), intent order_togo (1240), and intent order_eatin (1250). It can be seen that it contains

At this time, slot $coffee, slot $hotice, and slot $takeout are defined for intent order_coffee (1210), slot $hotice is defined for intent order_hotice (1220), and slot $hotice is defined for intent order_takeout (1230). $takeout is defined, and slots may not be defined for intent order_togo (1240) and intent order_eatin (1250).

When a bot processes the intents shown in FIG. 12, they are processed in units called blocks (processing blocks), and multiple intents can be mapped and processed in one block. Additionally, one intent may be mapped to multiple blocks.

Figures 13 to 16 are diagrams showing examples of execution blocks used in the scenario shown in Figure 11.

Figure 13 is a block diagram showing the connection between a plurality of intents and a plurality of execution blocks.

Referring to FIG. 13, it can be seen that the intent order_coffee (1210), intent order_hotice (1220), and intent order_takeout (1230) are connected to the execution block coffee selection (1310). At this time, the intent mapping list of the execution block coffee selection 1310 may include intent identifiers corresponding to the intent order_coffee (1210), the intent order_hotice (1220), and the intent order_takeout (1230).

At this time, the intent order_hotice (1220) and intent order_takeout (1230) are also connected to the execution block hot ice check (1320).

At this time, the intent order_takeout (1230) is also connected to the execution block takeout confirmation (1330).

Intent order_togo (1240), intent order_eatin (1250), intent followup.yes (1260), and intent followup.no (1270) are connected to the execution block takeout confirmation (1330).

As shown in Figure 13, intents are mapped to blocks for their execution, and conditions are defined in a grammatical form based on a variable called a slot for each mapped intent, so that the bot utters only by classifying intent for each user utterance. Appropriate detailed processing is possible through conditions without processing.

As shown in FIG. 13, multiple intents may be connected to one block (N:1), and multiple blocks may be connected to one intent (1:N), allowing free communication between intents and blocks. Mapping (N:N) is possible.

Furthermore, one execution block may have a connection relationship with other execution blocks that can be executed later. For example, the execution block coffee selection 1310 may be connected to the execution block hot ice confirmation 1320 and the execution block takeout confirmation 1330, and the execution block hot ice confirmation 1320 may be connected to the execution block takeout confirmation 1330. ) can be connected to.

FIG. 14 is a diagram illustrating an example of execution block coffee selection shown in FIG. 13.

Referring to FIG. 14, the execution block coffee selection 1310 may include a connected intent list, slot, slot filling, and action information.

At this time, the connected intent list may be included in the intent mapping list, and the intent mapping list may include an intent identifier and slot condition for each intent. At this time, the slot condition may correspond to a conditional statement related to one or more slots and may include a null condition.

The slots ($coffee, $hotice, $takeout) shown in FIG. 14 may include the union of all slots corresponding to the intents (order_coffee, order_hotice, order_takeout) to which the execution block coffee selection 1310 is connected.

The slot filling rank information may include information about the filling rank of the slots of the execution block coffee selection 1310 and utterance information required to fill the corresponding slot first.

Action information may consist of a condition and a response when the condition is satisfied. At this time, the condition may correspond to one or more conditional statements using slots, and the response may include a link that calls another block.

For example, the execution block coffee selection 1310 may connect the execution block hot ice confirmation 1320 based on condition determination using the slot $hotice.

For example, the execution block coffee selection 1310 may be connected to the execution block takeout confirmation 1330 based on condition determination using the slot $takeout.

FIG. 15 is a diagram illustrating an example of hot ice confirmation of an execution block shown in FIG. 13.

Referring to FIG. 15, the execution block hot ice check 1320 may include a connected intent list, slot, slot filling, and action information.

The slots ($hotice, $takeout) shown in FIG. 15 may include the union of all slots corresponding to the intents (order_hotice, order_takeout) to which the execution block hot ice check 1320 is connected.

The slot filling rank information may include information about the filling rank of the slots of the execution block hot ice check 1320 and ignition information required to fill the corresponding slot first.

Action information may consist of a condition and a response when the condition is satisfied. At this time, the condition may correspond to one or more conditional statements using slots, and the response may include a link that calls another block.

For example, the execution block hot ice check 1320 can be connected to the execution block takeout check 1330 based on condition determination using the slot $takeout.

FIG. 16 is a diagram illustrating an example of execution block takeout confirmation shown in FIG. 13.

Referring to FIG. 16, the execution block takeout confirmation 1330 may include a connected intent list, slot, slot filling, and action information.

The slot ($takeout) shown in FIG. 16 may include the union of all slots corresponding to the intents (order_takeout, order_togo, order_eatin, followup.yes, followup.no) to which the execution block takeout confirmation 1330 is connected. there is.

The slot filling rank information may include information about the filling rank of the slots in the execution block takeout confirmation 1330 and utterance information required to fill the corresponding slot first.

Action information may consist of a condition and a response when the condition is satisfied. At this time, the condition may correspond to one or more conditional statements using slots, and the response may include a link that calls another block.

FIG. 17 is a block diagram showing the connection relationship between execution blocks shown in FIGS. 13 to 16.

Referring to FIG. 17, it can be seen that the execution block coffee selection (1310), execution block hot ice confirmation (1320), and execution block takeout confirmation (1330) are connected in a graph form.

At this time, the connection between execution blocks may mean connection with the next execution blocks that can process the user's speech when entering the user speech waiting state in a specific current block. In other words, situations that can occur when a user utterance is received in a specific current block are execution blocks connected to the current block, so only intents connected to these execution blocks can be searched as intents that require user utterance processing. When waiting for a user's utterance, only the intents included in the sub-blocks connected based on the current block are collected and the intent is searched through a natural language processor, which corresponds to the utterance much more quickly and accurately than when searching all intents. It is possible to identify the intended intent.

In the example shown in FIG. 17, the execution block coffee selection 1310 is connected to the execution block hot ice selection 1320 and the execution block takeout confirmation 1330, so the user utterance is in the state of the execution block coffee selection 1310. If entered, the next state may be execution block hot ice selection (1320) or execution block takeout confirmation (1330).

Additionally, if a user utterance is input in the state of execution block hot ice selection (1320), the next state may be execution block takeout confirmation (1330).

When a user intent is identified from a user utterance, the blocks can be searched by sorting them in the order of priority of intent conditions of blocks containing the identified intent and/or priority defined by the bot designer.

When a block that meets the conditions is found, you can move to the corresponding block, perform an action defined according to the block settings, and return to the waiting state for user response according to the conditions of the action.

Connecting blocks can be defined as listing the blocks to be processed next in the current block when entering the user response waiting state, and multiple user response waiting states may be defined in one block.

Figure 18 is a diagram showing an example of connecting blocks by defining a user response waiting state in an execution block.

Referring to FIG. 18, it can be seen that one block can be connected to a plurality of other blocks in a graph form.

Block 1 shown in FIG. 18 is connected to Block 2, Block 3, Block 4, and Block 5. At this time, Block 1 is connected to Block 2 and Block 3 by Condition 4, and Block 3 is connected to Block 3 by Condition 5. It can be connected to Block 4 and Block 5. Therefore, after Block 1, Block 2, Block 3, Block 4, and Block 5 situations may occur.

Block 2 is connected to Block 6. Therefore, the situation of Block 6 may occur after Block 2.

Block 3 is connected to Block 6 and Block 7. Therefore, after Block 3, Block 6 and Block 7 situations may occur.

Block 5 is connected to Block 6 and Block 7. Therefore, after Block 5, Block 6 and Block 7 situations may occur.

For example, if the current state corresponds to Block 2, it is sufficient for the user's subsequent utterance to search for intents only for intents corresponding to Block 6, and the probability of occurrence is high. There is no need to perform user utterance matching for intents corresponding to Block 7, which does not exist.

When mapping one intent and one execution block, one intent may need to include multiple speech patterns.

For example, the intent (I1: subscription intention) includes the learning sentences “I want to use it” (pattern classification P1), “How do I use it” (pattern classification P1), and “I am curious about how to register” (pattern classification P3). and “Please inquire about how to sign up” (pattern classification P3), and the intent (I2: intention to withdraw) includes the learning sentences “I want to stop using” (pattern classification P2), “I don’t want to use it anymore” (pattern classification P2), This may include “I’m curious about how to unsubscribe” (pattern classification P3) and “I’m inquiring about how to unsubscribe” (pattern classification P3).

At this time, one block may be connected to the intent (I1: intent to join) and one block may be connected to the intent (I2: intent to leave), but the intents include multiple speech patterns, so the performance of the natural language processor is reduced. may deteriorate.

Therefore, in order to prevent performance degradation of the natural language processor, it is desirable to avoid including multiple speech patterns in one intent. For example, the intent (I1: intention to join) includes only the learning sentences “I want to use it” (pattern classification P1) and “how do I use it” (pattern classification P1), and the intent (I2: intention to withdraw) It includes only the learning sentences “I want to stop writing” (pattern classification P2) and “I don’t want to use it anymore” (pattern classification P2), and the intent (I3: inquiry intent) includes the learning sentences “I’m curious about how to <sign up>” ( Pattern Classification P3), “I’m curious about how to <subscribe>” (Pattern Classification P3), “I’m curious about how to <subscribe>” (Pattern Classification P3), and “I’m inquiring about how to <subscribe>” (Pattern Classification P3). there is.

In this way, when classifying intents so that they can be effectively processed by a natural language processor, intents and processing blocks need to be connected in more diverse forms. To this end, the intents of the block can each have a condition, and this condition can be defined as statements that compare slot values. That is, block and intent mapping information can be defined in the block in the form of a list (intent mapping list) of paired intents and slot conditions.

Figure 19 is a diagram showing an example of mapping an intent defined for each speech pattern to a block.

Referring to FIG. 19, it can be seen that three intents are connected to two blocks as a pair of intent and slot condition.

At this time, the slots used in the slot condition may include not only the slot defined in the current intent, but also slots of other blocks set while executing the previous conversation, or slots provided by the system.

In the example shown in FIG. 19, Intent 1 corresponds to utterance pattern 1 (Pattern 1), Intent 2 corresponds to utterance pattern 2 (Pattern 2), and Intent 3 (Intent 3) can be seen to correspond to speech pattern 3 (Pattern 3).

In the example shown in FIG. 19, the intent mapping list of Block 1 includes a pair of Intent 1 and Condition 1 and a pair of Intent 3 and Condition 2. ) There are two elements of the pair. Therefore, Block 1 is connected to two intents (Intent 1 and Intent 3).

The intent mapping list of Block 2 contains two elements: a pair of Intent 2 and Condition 3 and a pair of Intent 3 and Condition 4. exist. Therefore, Block 2 is connected to two intents (Intent 2 and Intent 3).

Figure 20 is a diagram showing a case of 1:1 mapping between an intent and an execution block.

Referring to FIG. 20, it can be seen that three intents (I1, I2, I3) each include four speech patterns.

In other words, Intent I1 (intent to sign up) includes the speech patterns “I want to use it,” “How do I use it,” “I’m curious about how to sign up,” and “Please inquire about how to sign up.”

Intent I2 (intent to withdraw) includes the speech patterns “I want to stop using it,” “I don’t want to use it anymore,” “I’m curious about how to unsubscribe,” and “I’m asking how to unsubscribe.”

Intent I3 (intent to inquire about regular membership) includes the speech patterns “I want to become a regular member,” “I want to be promoted to a regular member,” “I’m curious about how to sign up for a full member,” and “How to sign up for a full member.”

When intent (intent) and response (execution block) are mapped 1:1, when a sentence such as "*** I'm curious" appears in multiple intents, the natural language processor (NLU) distinguishes between these similar sentences. difficult. For example, a sentence such as “It’s my first time and I would like to inquire about how to sign up” may be processed by a natural language processor as a sentence similar to “I would like to inquire about how to unsubscribe” depending on the situation, resulting in incorrect recognition of intent.

In other words, the user's intention classification may be different from the intention classification that is easy to process by a natural language processor (NLU). In reality, one user's intention may be separated into several intentions that are easy for a natural language processor to process, and conversely, several The user's intentions may be classified into one intention that is easy for a natural language processor to process.

Figure 21 is a diagram showing an example of intent classification and block mapping according to an embodiment of the present invention.

Referring to FIG. 21, it can be seen that sentences are classified into a form that is easy for a natural language processor (NLU) to process, and responses are mapped according to conditions by tagging necessary slots.

In the example shown in Figure 21, four intents (I1, I2, I3, I4) are connected to three responses (R1, R2, R3).

At this time, Intent 1 (I1) is connected to the subscription information response (R1). At this time, the slot condition may be a null condition.

At this time, Intent 4 (I4) is connected to the subscription information response (R1), the withdrawal information response (R2), and the regular member information response (R3). At this time, Intent 4 (I4) can be connected to the subscription information response (R1) by the slot condition $service == "Sign up". At this time, Intent 4 (I4) can be connected to the withdrawal information response (R2) by the slot condition $service == "withdrawal". At this time, Intent 4 (I4) can be connected to the regular member information response (R3) by the slot condition $level == "regular member" and $service == "subscription".

At this time, Intent 2 (I2) is connected to the withdrawal information response (R2). At this time, the slot condition may be a null condition.

At this time, Intent 3 (I3) is connected to the regular member information response (R3). At this time, the slot condition may be a null condition.

The responses shown in FIG. 21 may each correspond to one execution block.

At this time, the execution block can be defined including intent, slot conditions, and chatbot response processing.

FIG. 22 is a diagram showing execution blocks corresponding to the responses shown in FIG. 21.

Referring to FIG. 22, each execution block includes an intent mapping list and operation information including at least one pair of intent and slot condition.

As shown in FIG. 22, the slot condition may include a NULL condition corresponding to “no condition” and may include a conditional statement using one or more slots.

When a chatbot performs conversation processing, the natural language processor first classifies intents and extracts slots from the sentences entered by the user, then checks the conditions listed in the blocks containing the classified intents to find a matching block. .

In the examples shown in Figures 21 and 22, when the intent is classified as "I4: Inquiry Intent", the blocks containing this intent are B1, B2, and B3, and the blocks matching according to the conditions are listed in the table below. Same as 3.

condition block Custom condition check order $service == "Sign up" B1: Registration instructions One $service == "Withdraw" B2: Withdrawal notice 2 $service == "regular member" and $service == "subscription" B3: Regular member information 3

The order in which conditions are checked can be determined by the user specifying the order in which blocks are listed. If it is simply processed in the order listed by the user as shown in Table 3, a sentence such as signing up as a regular member satisfies the conditions $level == "regular member" and $service == "subscription", but at the same time, logically, $service = = The condition “join” is also satisfied, and the B1 block with a high condition check rank can be executed, which may not be a desirable result.

To solve this problem, priorities can be given again by considering the inclusion relationship of the conditions, as shown in Table 4 below.

condition block Custom condition check order Reorder rankings considering inclusion relationships $service == "regular member" and $service == "subscription" B3: Regular member information 3 One $service == "Sign up" B1: Registration instructions One 2 $service == "Withdraw" B2: Withdrawal notice 2 3

At this time, the inclusion relationship of the condition can be determined by the inclusion relationship of the set of slots used in the conditional expression. The included side (large set) is treated preferentially over the included side (small set), and if there is no inclusion relationship between them, The priority specified by the user can be maintained. In the example of Table 4, the condition of B3, which checks both level and service, has a higher condition inclusion relationship ranking than B1 and B2, which only check service, so B3 can be processed preferentially over B1 or B2.

Figure 23 is a diagram showing an example of priority sorting according to the inclusion relationship of conditions corresponding to the coffee ordering scenarios illustrated in Figures 11 to 17.

Referring to Figure 23, it can be seen that the conditions corresponding to two coffee slots and takeout slots have higher priority than the conditions corresponding to one coffee slot.

Therefore, when the user utters “I would like to take out Americano,” the slot coffee is filled with “Americano” and the slot takeout is filled with “takeout,” so the coffee order in the execution block is executed with priority over the takeout confirmation in the execution block.

When the chatbot process is executed, intents are collected from connected blocks while currently waiting for user response, and an appropriate intent can be searched among the intents collected through a natural language processor.

When searching for a block with an intent, you can search for a suitable block by sorting it in the order of the priority of the intent conditions of the blocks containing the intent and the priority defined by the user.

If a suitable block is not found because the conditions are not met, the above method can be repeated using the next suitable intent.

When you find a block that meets the conditions, you can move to that block, perform actions determined by the chatbot's response process, and enter a state waiting for the next user response depending on the conditions.

Figure 24 is a block diagram showing an example of modules required to run a chatbot.

Referring to Figure 24, the modules required when a chatbot executes the resources defined by the bot designer are roughly divided around the conversation flow processor: a part that controls message input and output, a part that manages block status, a part that processes natural language, and a block. It can be seen as a part that processes conditions and a part that executes blocks.

In the example of Figure 24, the modules required to run the chatbot are conversation flow processor, block state manager, message input gateway, message output gateway, response message generator, block process executor, NLU searcher, block searcher, intent classifier, slot extractor, and natural language processor. , may include an NLU priority manager, a condition interpreter, and a block condition priority manager.

Figure 25 is a diagram showing the process of searching intents and blocks while running a chatbot.

Referring to Figure 25, when a user's speech is input while the chatbot is waiting for user input, blocks that can be executed at the current location are first searched. At this time, sub-blocks connected to the current (execution) block may be searched first, and then the main blocks of the scenario may be searched. At this time, the main block refers to a list of blocks to be operated in the initial state of conversation, and becomes the target blocks that the bot searches for when the user first speaks. In other words, efficient conversation flow processing is possible if intent is identified only for sub-blocks that are likely to be executed after the current block and main blocks that are newly executable.

When a block is searched, the intents mapped to the searched block are listed and the processing priority is checked. At this time, the intents defined in the searched blocks are listed, and the matching score of the intents corresponding to the user's utterance is determined using a natural language processor. By sequentially searching the defined intent pattern threshold score list, comparing the names and matching scores of the intents, the intent with the highest matching score among the intents that satisfy the pattern and threshold score is selected first. Extract.

If there is more than one block containing a matched intent, the processing priority is sorted according to the defined intent mapping conditions. If the conditions have the same priority, the priority of the scenario and block defined by the user is followed. At this time, the priority of the conditions may be prioritized according to the inclusion relationship of the condition comparison statement. In other words, when constructing a condition based on slots extracted from a user's utterance, the comparison sentence that constitutes the condition is composed mainly of slots mentioned by the user and slots not mentioned are omitted, so a condition containing many slots It can be seen as a better reflection of the user's intention. That is, if there are multiple true conditions within a condition statement, the condition containing more constraints or slots may be considered to have higher priority.

The sorted conditions are checked in order of priority, and the block containing the condition that satisfies the first order is processed as an execution block. If a block that satisfies the condition cannot be found, the blocks to which the intent belongs are searched again using the intent of the next priority, and the process of determining the priority of the condition and checking the condition is repeated to find the block.

Figure 26 is an operation flowchart showing a chatbot service provision method according to another embodiment of the present invention.

Referring to FIG. 26, the chatbot service providing method according to another embodiment of the present invention determines an intent corresponding to a user conversation (S2610).

At this time, the intent corresponding to the user conversation may be selected from within one or more candidate blocks connected to the current block through a graph or from intent candidates connected to the main block.

At this time, the intent corresponding to the user conversation may be determined by considering the first intent candidate connected to the candidate block as priority over the second intent candidate connected to the main block.

Additionally, the chatbot service provision method according to another embodiment of the present invention uses an intent mapping list to select an execution block corresponding to the intent (S2620).

At this time, the intent mapping list is provided for each execution block and may correspond to at least one intent.

At this time, the intent mapping list may include an intent identifier and slot condition for each intent.

At this time, the slot condition may correspond to a conditional statement related to one or more slots.

At this time, the slot condition may include a null condition.

At this time, the intent mapping list may include intent identifiers and slot conditions corresponding to a plurality of intents (N:1), and the intent is one in which different execution blocks have the same intent identifier. In this case, if the slot conditions are different, they may be connected (1:N) to different execution blocks.

At this time, slots corresponding to the execution block may include slots of intents connected to the execution block through the intent mapping list.

At this time, when the intent corresponding to the user conversation is connected to a plurality of execution blocks, the execution block corresponding to the intent is the first execution block having the inclusion relationship of the slot condition corresponding to the large set. It may be selected with priority over the second execution block having an inclusion relationship of slot conditions corresponding to a small set included in . At this time, the inclusion relationship of the slot condition may be the inclusion relationship of slots included in the slot condition. For example, an execution block corresponding to $level == "regular member" and $service == "subscription" may be selected with priority over an execution block corresponding to $service == "subscription".

Additionally, the chatbot service provision method according to another embodiment of the present invention performs conversation processing corresponding to the intent based on the execution block (S2630).

By the above-described configuration, the chatbot service provision method according to an embodiment of the present invention has an N:N (N is a natural number) mapping between intents classified appropriately for a natural language processor and execution blocks that process the intents. By making this possible, one function can be performed with several different utterances, and conversely, a completely different function can be performed with similar utterances.

Additionally, the chatbot service provision method according to an embodiment of the present invention allows defining various intent-execution block mapping relationships using an intent mapping list that includes an intent identifier and a slot condition for each intent.

In addition, the method of providing a chatbot service according to an embodiment of the present invention identifies candidate blocks that can be executed after the current execution block when identifying an intent corresponding to a user conversation, and only intents connected to the candidate blocks are used. By targeting identification, fast and accurate intent identification is possible.

Figure 27 is a graph showing natural language processor performance as the number of speech patterns increases.

Referring to Figure 27, when the intents are classified according to the functions provided by the chatbot, when various speech sentences are added, similar speech patterns frequently appear in several intents, and the intent of the natural language processor (NLU) as shown in A. It can be seen that the classification performance deteriorates rapidly.

To solve this problem, by allowing the natural language processor to define intents for each similar speech pattern in a form suitable for processing, the performance of the natural language processor can be maintained as high as B even if the speech pattern increases.

Additionally, by mapping different speech patterns to a unit called 'block' so that the bot can properly process them according to conditions, various speech patterns can be processed efficiently.

Blocks are connected in a graph form and multiple response waiting states can be defined depending on conditions, so work processing with various processing flows or steps depending on conditions can be visually defined with a small number of blocks.

Implementations of the functional operations and subject matter described in this specification may be implemented as digital electronic circuits, computer software, firmware or hardware including the structures disclosed herein and their structural equivalents, or a combination of one or more of these. It is possible to implement. Implementations of the subject matter described herein may comprise one or more computer program products, that is, one or more modules of computer program instructions encoded on a tangible program storage medium for execution by or for controlling the operation of a processing system. It can be implemented.

The computer-readable medium may be a machine-readable storage device, a machine-readable storage substrate, a memory device, a composition of matter that affects a machine-readable radio wave signal, or a combination of one or more of these.

In this specification, 'system' or 'device' includes all instruments, devices, and machines for processing data, including, for example, programmable processors, computers, or multiple processors or computers. In addition to the hardware, the processing system may include code that forms an execution environment for computer programs on demand, such as code making up processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of these. .

A computer program (also known as a program, software, software application, script, or code) may be written in any form of a programming language, including compiled, interpreted, a priori, or procedural languages, as a stand-alone program, or as a module. It can be deployed in any form, including components, subroutines, or other units suitable for use in a computer environment. Computer programs do not necessarily correspond to files in a file system. A program may be stored within a single file that serves the requested program, or within multiple interacting files (e.g., files storing one or more modules, subprograms, or portions of code), or as part of a file that holds other programs or data. (e.g., one or more scripts stored within a markup language document). The computer program may be deployed to run on a single computer or multiple computers located at one site or distributed across multiple sites and interconnected by a communications network.

Meanwhile, computer-readable media suitable for storing computer program instructions and data include semiconductor memory devices such as EPROM, EEPROM, and flash memory devices, magnetic disks such as internal hard disks and external disks, magneto-optical disks, and CDs. -Can include all forms of non-volatile memory, media, and memory devices, including ROM and DVD-ROM disks. The processor and memory may be supplemented by, or integrated into, special-purpose logic circuitry.

Implementations of the subject matter described herein may include backend components, such as a data server, middleware components, such as an application server, or, for example, a web browser or graphical user through which a user may interact with an implementation of the subject matter described herein. It may be implemented in a front-end component, such as a client computer with an interface, or in a computing system that includes any combination of one or more of such back-end, middleware, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication, such as a telecommunications network.

Although this specification contains details of numerous specific implementations, these should not be construed as limitations on the scope of any invention or what may be claimed, but rather as descriptions of features that may be unique to particular embodiments of particular inventions. It must be understood. Likewise, certain features described herein in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination. Furthermore, although features may be described as operating in a particular combination and initially claimed as such, one or more features from a claimed combination may in some cases be excluded from that combination, and the claimed combination may be a sub-combination. It can be changed to a variant of a sub-combination.

Additionally, although operations are depicted in the drawings in a specific order herein, this should not be understood to mean that such operations must be performed in the specific order or sequential order shown or that all illustrated operations must be performed to obtain desirable results. Can not be done. In certain cases, multitasking and parallel processing may be advantageous. Additionally, the separation of various system components in the above-described embodiments should not be construed as requiring such separation in all embodiments, and the described program components and systems may generally be integrated together into a single software product or packaged into multiple software products. You must understand that you can

As such, this specification is not intended to limit the invention to the specific terms presented. Accordingly, although the present invention has been described in detail with reference to the above-described examples, those skilled in the art may make modifications, changes, and variations to the examples without departing from the scope of the present invention. The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention. do.

As described above, the method and device for providing a chatbot service using a conditional generation interface according to the present invention is not limited to the configuration and method of the embodiments described above, and the embodiments can be modified in various ways. All or part of each embodiment may be selectively combined to allow for configuration.

According to the present invention, a method and device for providing a chatbot service using a conditional expression generation interface are disclosed. A chatbot service provision method according to an embodiment of the present invention includes providing a conditional expression creation interface, and generating an execution block for processing an intent corresponding to a user conversation through the conditional expression creation interface. Accordingly, when a user sets a conditional expression for chatbot execution using a chatbot builder or a chatbot service provider, it is possible to reduce errors in the conditional expression and improve readability, making it possible to effectively rescue the chatbot.

910: Communication unit 920: Processor
930: Memory 1010: Chatbot
1020: Natural language processor 1030: Chatbot resource DB
1040: Chatbot Builder

Claims (5)

Generating a conditional expression and providing a conditional expression creation interface that recommends a pair of parentheses entered at an arbitrary position in the conditional expression; and
Creating an execution block for processing an intent corresponding to a user conversation through the conditional expression creation interface.
Including,
The conditional expression creation interface is
A chatbot service providing method using a conditional expression generation interface, characterized in that the arbitrary position is corrected based on the type of the input parenthesis by considering whether the arbitrary position is a position where parentheses can be input. delete ◈Claim 3 was abandoned upon payment of the setup registration fee.◈ In claim 1,
The conditional expression creation interface is
A method of providing a chatbot service using a conditional expression generation interface, characterized in that after correcting the random position by moving it, a pair of parentheses entered at the corrected position is recommended. ◈Claim 4 was abandoned upon payment of the setup registration fee.◈ In claim 3,
The conditional expression creation interface is
If the type of the input parenthesis is an open parenthesis, correction is performed by moving the position of the input parenthesis to the front of the conditional expression, and if the type of the input parenthesis is a closing parenthesis, the position of the input parenthesis is performed as above. A method of providing a chatbot service using a conditional expression creation interface, characterized by performing correction while moving the conditional expression to the back. ◈Claim 5 was abandoned upon payment of the setup registration fee.◈ In claim 1,
The above conditional expression is
A method of providing a chatbot service using a conditional expression generation interface, characterized in that it includes at least one condition consisting of a combination of a condition variable, a comparison operator, and a condition value.

Images