KR20200023620A - An apparatus for rule-based user inference reasoning for conversation awareness - Google Patents

Abstract

Provided is a rule-based user intention inference apparatus for conversation recognition. According to the present invention, the apparatus is configured to create rules for conversation recognition and infer user′s intention based on the rules. The apparatus comprises: a pre-processing unit which extracts input keyword information including information about a keyword set which is a set consisting of at least one keyword from conversation data including information about user′s conversation; a determination unit which determines, based on an inclusion relation between sets, whether there is an intention cluster corresponding to the input keyword information among a plurality of intention clusters, which are clusters of at least one keyword set in which an inclusion relation between the sets is not established; an updating unit which updates, based on the determination results, a knowledge base (KB) including the plurality of intention clusters by using the input keyword information; and a deciding unit which decides a keyword set including a keyword set included in the input keyword information among one or more keyword sets included in each of the plurality of intention clusters. According to the present invention, the apparatus is able to secure completeness and soundness when inferring user′s intention.

Description

Rule-based user intention inference device for dialogue recognition {AN APPARATUS FOR RULE-BASED USER INFERENCE REASONING FOR CONVERSATION AWARENESS}

The present invention relates to a rule-based intention inference device for recognizing a user's conversation, and more particularly, to automatically generate a rule in order to correctly recognize a user's command or query, and to generate a rule based on the rule. A device for inferring a user's intention.

In a general rule system, a data analyst analyzes and selects distinguishable key words to create a rule in order to create a rule. In other words, the rules could be created only by experts extracting the keywords for the specific knowledge related to the data, rather than automatically generating the rules. In addition, the relationship between the rules generated in this process could not be defined by extracting the main keywords from the user's conversation sentences and creating rules and analyzing intentions.

As a result, there was a problem that the rule was used only to determine whether a simple keyword appeared and it was difficult to infer because it was difficult to detect and resolve the contradiction between the rules. In addition, there is a problem in that it is difficult to accept various vocabularies because the stored rules are used without inferring the relationship between the rules when deducing the intention of the query.

Therefore, there is a need for an automatic rule generation method, a rule-based inference reasoning method, and an apparatus that can supplement the disadvantages of the rule generation method of the existing rule system.

Related prior arts include Korean Patent No. 10-1187425 (name of the invention: an interactive service providing system and method, and a registration date: September 25, 2012).

The present invention grasps the user's conversation intention, and systematically configures a set of keywords included in the user's conversation according to the inclusion relation between the sets of keywords, thereby automatically generating a rule for recognizing the conversation and generating the generated rule. It is intended to provide a device for inferring the user's intention by using.

The problem to be solved by the present invention is not limited to the above-mentioned task (s), another task (s) not mentioned will be clearly understood by those skilled in the art from the following description.

An apparatus for inferring reasoning for achieving the above object is a device for generating a rule for recognizing a conversation and inferring a user's intention based on the rule, comprising: at least one of conversational data including information about a conversation of the user; A plurality of preprocessing units for extracting input keyword information including information on a keyword set, which is a set of keywords, and a group of groups of at least one keyword set based on an inclusion relationship between the sets, wherein at least one keyword set is not established; A determination unit that determines whether there is an intention cluster corresponding to the input keyword information among the intention clusters of the at least one of the intention clusters, and a knowledge base (KB) including the plurality of intention clusters using the input keyword information according to the determination result. An update unit for updating and at least one keyword set included in each of the plurality of intention groups; And a determination unit for determining a keyword set including a keyword set included in the input keyword information, wherein the determination unit is based on an inverse order of the order according to an inclusion relationship of at least one keyword set between the plurality of intention groups. The keyword set is determined, and the determining unit includes determining the keyword set in the order of the intention group including the least keywords in the intention cluster including the most keywords.

Preferably, the determination unit may determine whether the intention cluster exists, based on an order according to the inclusion relationship of the at least one keyword set between the plurality of intention clusters.

Preferably, the determination unit inputs one intention group among the plurality of intention clusters, in which an inclusion relationship between each of the at least one keyword set included in the intention group and the keyword set included in the input keyword information does not hold. It may be determined that it corresponds to the information.

Preferably, the determining unit determines whether each of the at least one keyword set included in the intention group and the intention of the keyword set included in the input keyword information are identical in each of the plurality of intention clusters according to a predetermined criterion. If the determination result is the same, based on the inclusion relationship between two keyword sets having the same intention, the keyword set included in the intention group may be replaced with the keyword set included in the input keyword information.

Preferably, the update unit, if there is an intention group corresponding to the input keyword information among the plurality of intention clusters, adds the keyword set included in the input keyword information to the intention group corresponding to the input keyword information. If there is no intention group corresponding to the input keyword information among a plurality of intention clusters, a new intention group including a keyword set included in the input keyword information may be added to the KB.

An inference method for achieving the object of the present invention is a method of generating a rule for recognition of a conversation and inferring user intention based on the rule, comprising a plurality of intention clusters composed of at least one keyword set. A method of inferring a user's intention using a knowledge base, the method comprising: extracting input keyword information including information on a keyword set, which is a set composed of at least one keyword, from conversation data including information about a user's conversation step; Determining whether there is an intention group corresponding to the input keyword information among a plurality of intention clusters formed of at least one keyword set based on an inclusion relationship between sets; Updating a knowledge base (KB) including the plurality of intention clusters using the input keyword information according to the determination result; And determining a keyword set including a keyword set included in the input keyword information among at least one keyword set included in each of the plurality of intention groups, wherein the determining of the keyword set includes the plurality of intents. The keyword set is determined based on the reverse order of the order according to the inclusion relationship of the at least one keyword set between clusters.

Preferably, the determining of the existence of the intention cluster may determine whether the intention cluster exists, based on an order according to the inclusion relationship of the at least one keyword set among the plurality of intention clusters.

Preferably, the determining of the existence of the intention group does not include an inclusion relationship between each of the at least one keyword set included in the intention group and the keyword set included in the input keyword information. One intention cluster may be determined to correspond to the input keyword information.

Preferably, the determining of the existence of the intention cluster includes: in each of the plurality of intention clusters, each of at least one keyword set included in the intention cluster and the intention of the keyword set included in the input keyword information. Determining whether the same according to the; And if the determination result is the same, selectively replacing the keyword set included in the intention group with the keyword set included in the input keyword information based on the inclusion relationship between two keyword sets having the same intention. .

Preferably, in the updating of the KB, if there is an intention group corresponding to the input keyword information among the plurality of intention clusters, the keyword set included in the input keyword information is assigned to the intention group corresponding to the input keyword information. Adding; And if there is no intention group corresponding to the input keyword information among the plurality of intention clusters, adding a new intention cluster including a keyword set included in the input keyword information to the KB.

According to an embodiment of the present invention, a method for generating a rule for recognizing a conversation may be performed by grasping a user's intention of a conversation and systematically configuring a set of keywords included in the user's conversation in an order according to an inclusion relationship between the sets of keywords. There is an effect that a rule can be automatically generated by separating a user's conversation intent into a cluster unit.

In addition, the rule-based intention inference method according to an embodiment of the present invention infers the intention by applying a rule divided in a cluster unit in the reverse order with respect to the user's conversation, in order, the completeness (completeness) And soundness.

1 is a flowchart illustrating a rule generating method for dialogue recognition according to an embodiment of the present invention.
2 is a block diagram of an apparatus for generating rules for dialogue recognition according to an embodiment of the present invention.
3 is a flowchart of a rule-based intention inference method according to an embodiment of the present invention.
4 is a block diagram of a rule-based intention reasoning apparatus according to an embodiment of the present invention.
5 is a pseudo code of a rule generation method according to an embodiment of the present invention.
6 is a pseudo code of an intention reasoning method according to an embodiment of the present invention.
7 is a view for explaining the configuration of the intention group according to an embodiment of the present invention.
8 is a view for explaining a user intention reasoning process using the intention cluster according to an embodiment of the present invention.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. In describing the drawings, similar reference numerals are used for similar elements.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component. The term and / or includes a combination of a plurality of related items or any item of a plurality of related items.

When a component is said to be "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that another component may be present in the middle. Should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that there is no other component in between.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art, and shall not be construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention may be applied to conversations (eg, voice recognition artificial intelligence (AI) speakers, unmanned telephone service) as well as conversations made through words as well as conversations made through a text (eg, chat service, messenger service). It will be apparent to those skilled in the art that the present invention can be applied to other forms of dialogue. In addition, the conversation of the present invention may be a conversation not only in Korean but in all kinds of foreign languages.

1 is a flowchart illustrating a rule generating method for dialogue recognition according to an embodiment of the present invention.

In operation S110, the rule generating device extracts input keyword information including information about a keyword set, which is a set composed of at least one keyword, from conversation data including information about a user's conversation.

For example, the rule generating device may use POS tagging as a preprocessing process for rule generation to detect unnecessary parts of speech such as interjection, punctuation, investigation, adverbs, etc., which do not play a large role as keywords among words included in the conversation data. You can extract the keyword set by removing it. In addition, in the case of a conversation performed in a foreign language other than Korean, the rule generator may extract a keyword set by performing an analysis such as a part-of-speech analysis to match the characteristics of the foreign language.

In operation S120, the rule generating apparatus determines whether an intention group corresponding to the input keyword information exists among a plurality of intention groups composed of at least one keyword set based on the inclusion relation between sets.

Here, the intention cluster may be a cluster of at least one keyword set in which an inclusion relationship does not hold.

For example, the keyword set A of {'Nowon-gu', 'candidate', 'teach'} and the keyword set B of {'Yangcheon-gu', 'gu mayor', 'candidate', 'teach'} can be assumed. In this case, the keyword sets A and B may be included in one intention group because the inclusion relations of A′B do not hold and the inclusion relations of A′B do not hold at the same time.

On the other hand, in the case of keyword set C of {'Nowon-gu', 'candidate', 'teach'} and keyword set D of {'nowon-gu', 'gu office', 'candidate', 'teach'}, C⊂ Since the inclusion relationship of D holds, the keyword sets C and D cannot be included in one intent group.

In this case, the rule generating apparatus may determine an inclusion relationship with all of at least one keyword set included in each of the plurality of intention clusters with respect to the keyword set included in the input keyword information. The rule generating device may determine whether an intention group corresponding to the input keyword information exists among the plurality of intention groups by using the determination result regarding the inclusion relationship.

At this time, the result of the determination may include i) the presence of an intention cluster corresponding to the input keyword information among the plurality of intention clusters, and ii) the intention cluster corresponding to the input keyword information among the plurality of intention clusters. have.

In another embodiment, the rule generating apparatus may determine whether the intention cluster exists based on an order according to the inclusion relationship of the at least one keyword set among the plurality of intention clusters.

For example, the rule generating device may start the determination from the intention group including the keyword set including the smallest number of elements among the plurality of intention groups. That is, referring to FIG. 7, the rule generating apparatus may start the determination from the intention group # 1 including {'candidate', 'teach'}, which is a keyword set including only two elements.

And, the rule generating device is {'Yangcheon-gu', 'gujangjang', 'candidate', 'teach'}, which is a keyword set that is a subset of {'candidate', 'teach'} keyword set of intent group # 1, It is possible to judge the intention cluster # 2 including {'Nowon-gu', 'candidate', 'teach'}, {'wonju', 'market', 'candidate', 'teach'}. In addition, the rule generator is a keyword set that is a subset of the keyword set {"Nowon-gu", "candidate", "teach"} of the intent group # 2 {"Nowon-gu", "market", "candidate", "teach" Can be determined with respect to the intention group # 3 including the "

At this time, the order according to the inclusion relationship of at least one keyword set between the plurality of intention groups has a relationship in which any one keyword set included in the intention group # 1 is included in any one keyword set included in the intention group # 2. When there is a relationship in which any one keyword set included in the intention group # 2 is included in any one keyword set included in the intention group # 3, the intention group # 1, the intention group # 2, and the intention group # 3 It can mean the order.

As such, the rule generating apparatus determines whether the intention cluster exists in the order according to the inclusion relationship of at least one keyword set among the plurality of intention clusters, starting from the intention cluster including the keyword set including the smallest number of elements. As a result, the existence of the intention cluster can be determined in the order of the intention cluster including the keyword set on the more general content to the intention cluster including the keyword set on the more specific content. That is, the rule generation device may determine whether or not there is an intention cluster corresponding to an intention cluster composed of a keyword set having a relatively small number of keywords, in an order cluster consisting of a keyword set having a relatively large number of keywords.

As a result, the rule generating device can determine the intention cluster including the most general contents as the corresponding intention cluster with respect to the keyword set included in the input keyword information. That is, the rule generating device can systematically maintain the plurality of intention clusters in the order of intention clusters including keyword sets on more specific contents in the intention cluster including keyword sets on more general contents.

In another embodiment, the apparatus for generating a rule may determine that an intention cluster in which the inclusion relationship between each of the at least one keyword set and the keyword set included in the input keyword information does not hold among the plurality of intention clusters corresponds to the input keyword information. Can be.

For example, referring to FIG. 7, the rule generating apparatus includes intention group # 1 and intention group # 2, respectively, for keyword sets {'Gangnam-gu', 'candidate', and 'teach'} included in the input keyword information. The inclusion relationship with the at least one keyword set can be determined.

More specifically, the rule generation device includes the keyword set {'Gangnam-gu', 'candidate', 'teach'} included in the input keyword information with the keyword set {'candidate', 'teach'} of the intent group # 1. The relationship is established, and the keyword set of Intent Group # 2 {'Yangcheon-gu', 'gujangjang', 'candidate', 'teach'}, {'nowon-gu', 'candidate', 'teach'}, {'won-ju' , 'Market', 'candidate', 'teach'} and the inclusion relationship does not hold. In this case, the rule generating device may determine that the keyword set {'Gangnam-gu', 'candidate', 'teach'} included in the input keyword information corresponds to the intention group # 2.

In another embodiment, the rule generating apparatus may update the general-specific rule corresponding to the keyword set included in each of the plurality of intention clusters based on the keyword set included in the input keyword information.

First, the rule generation device may determine whether each of the at least one keyword set included in the intention group and the intention of the keyword set included in the input keyword information are the same according to a predetermined criterion.

For example, the rule generating apparatus may include a keyword set included in the intention group {"Yangcheongu", "candidate", "some", "teach"}, and the keyword set included in the input keyword information is {"Yangcheongu", " Assuming "candidate" and "teach me", the intention of the two keyword sets can be judged to be the same as "tell me the candidate of Yangcheon-gu". This is because the rule generating device may determine that the keyword 'some' does not include a meaningful user's intention.

In this case, the apparatus for generating a rule may grasp the intention through collective analysis, such as reasoning of conceptual subsumption. The conceptual inclusion relationship may refer to a relationship between concepts in a taxonomy of ontology of description logic. For example, the Tabau algorithm of the male concept: subconcept logic of the human concept engine uses the definition of a concept as a set of instances for mathematical interpretation. have.

Secondly, if the determination result is the same, the rule generating apparatus selectively replaces the keyword set included in the intention group with the keyword set included in the input keyword information based on the inclusion relationship between two keyword sets having the same intention. can do.

For example, the rule generating apparatus may include a keyword set {'Yangcheon-gu', 'candidate', 'something', 'teach'} included in the intention group, and a keyword set {'Yangcheon-gu', 'candidate', 'Teach'} can be judged as two sets of keywords with the same intent. If the keyword set included in the input keyword information is a subset of the keyword set included in the intention group, the rule generating device may replace the keyword set included in the intention group with the keyword set included in the input keyword information.

More specifically, the rule generating device may include a keyword set {'Yangcheon-gu', 'candidate', 'teach'} included in the input keyword information in a keyword set {'Yangcheon-gu', 'candidate', 'something' , 'Teach'}, so that the keyword set included in the intention group can be replaced with the keyword set included in the input keyword information.

As a result, the rule generating device can secure the general-detailed rules corresponding to the most concise and clear keyword sets among the keyword sets having the same intention. This, referring to Fig. 5, shows a pseudo code indicating the contents related to the sixth to eighth lines.

Finally, in step S130, the rule generating device updates the knowledge base (KB) including the plurality of intention clusters using the input keyword information according to the determination result.

That is, the rule generating device may update the KB including the plurality of intention clusters according to the determination result of the previous step S120.

In another embodiment, the rule generating device may update the KB in different ways depending on whether there is an intention group corresponding to the input keyword information among the plurality of intention groups.

First, if there is an intention group corresponding to the input keyword information among the plurality of intention clusters, the rule generating device may add the keyword set included in the input keyword information to the corresponding intention group.

In other words, the rule generating apparatus may include the keyword generation included in the input keyword information when the inclusion relationship between the keyword set included in the input keyword information and all of the at least one keyword set included in any one intention group among the plurality of intention clusters does not hold. You can add a set of keywords to any one intent group.

For example, the rule generator may include the keyword set {'Gangnam-gu', 'candidate', 'teach'} included in the input keyword information with the keyword set {'candidate', 'teach'} of the intent group # 1. And the keyword set of Intent Group # 2 {'Yangcheon-gu', 'gu-jangjang', 'candidate', 'teach'}, {'nowon-gu', 'candidate', 'teach'}, {'won-ju', ' Market ”,“ candidate ”and“ teach me ”}. In this case, the apparatus for generating a rule may determine that the keyword set {'Gangnam-gu', 'candidate', 'teach'} included in the input keyword information corresponds to the intention group # 2, and add the keyword set to the intention group # 2.

Next, if there is no intention group corresponding to the input keyword information among the plurality of intention clusters, the rule generating apparatus adds a new intention cluster including the keyword set included in the input keyword information to the KB including the plurality of intention clusters. can do.

That is, when the inclusion relationship is established between the keyword set included in the input keyword information and one of the keyword sets included in each of the plurality of intention clusters, the rule generating device generates a new intention cluster and generates a keyword set included in the input keyword information. You can add to that new intent cluster. In addition, the rule generating device may add the new intention cluster to the KB. At this time, the number of intention clusters included in the KB may increase by +1.

For example, the rule generation device may include a keyword set {'infant', 'water play', 'product', 'recommend'} in the input keyword information, and a keyword set {'product', 'recommend'} of the intent group # 1.} The inclusion relation is established, and the inclusion relation is also established with the keyword set {'water play', 'product', 'recommend'}, {'vacation', 'product', 'recommend'}. In this case, the rule generating device includes the keyword set {'infant', 'water play', 'product', 'recommend'} included in the input keyword information, so that the inclusive group # 1 and # 2 are included in the rule. Create a cluster and add the keyword set {'infant', 'water play', 'product', 'recommend'} included in the input keyword information to the new intention cluster.

On the other hand, when the rule generating device does not establish any inclusion relationship between the keyword set included in the input keyword information and at least one keyword set included in each of the plurality of intention clusters, the rule generating apparatus uses a new keyword set included in the input keyword information. You can generate a KB.

For example, the rule generating apparatus assumes that KB # 1 related to content inquiring about candidates running for an election exists, when the input keyword information is requesting product recommendation, the keyword set included in the input keyword information and Since no inclusion relationship is established between at least one keyword set included in each of the plurality of intention groups, a new KB # 2 may be generated. In this case, an intention cluster including a keyword set included in the input keyword information may be newly generated and added to KB # 2.

In another embodiment, the rule generating device may update the KB including the plurality of intention clusters by using the keyword set included in the input keyword information as follows.

Here, the rule generating device may be regarded as a contradiction where the intention corresponding to the intention group and the intention corresponding to the input keyword information do not coincide. In addition, it may be assumed that each of the plurality of intention clusters is configured in an order according to the inclusion relationship of keyword sets.

For example, the rule generator may include the jth intention group R _j (where j <i) when the total i + 1 intention clusters from the intention group R ₀ to R _i are configured in the order according to the inclusion relationship of the keyword set. It may be assumed that one of the keyword sets r _j is a subset (ie, r _j ⊂ t _tokens ) of the keyword set t _tokens included in the input keyword information. In this case, the rule generating apparatus may determine the inclusion relationship between each of the at least one keyword set r _{j + 1} included in the j + 1 th intention group R _{j + 1} and t _tokens . Similarly, the rule generating device may sequentially determine the inclusion relationship with t _tokens for R _{j + 2} , R _{j + 3,} etc., which is a process of determining the inclusion relationship for all intention clusters or the inclusion relationship exists. This may continue until a cluster of intents that do not exist are found. This content is shown in the 11th to 12th lines with reference to FIG.

In this case, the rule generating apparatus, if the j + 1 when the r _{j + 1} is t _tokens, one of the set of keywords contained in the second intended Community as a subset (i.e., t _tokens ⊂ r _{j + 1),} t _tokens include The new intent cluster may be referred to as the j + 1st intention cluster, and the existing j + 1st intention cluster as R _{j + 1} may be referred to as the j + 2nd intention cluster R _{j + 2} . This content is shown in the 13th-14th line with reference to FIG.

As described above, the rule generation method for dialogue recognition according to an embodiment of the present invention grasps the user's conversation intention, and systematically sets a set of keywords included in the user's conversation in the order of inclusion relation between the sets of the keywords. By constructing, it is possible to automatically generate a rule by dividing the user's intentions into groups.

2 is a block diagram of an apparatus for generating rules for dialogue recognition according to an embodiment of the present invention.

2, the rule generating apparatus 200 for recognizing a conversation according to an embodiment of the present invention includes a preprocessor 210, a determiner 220, and an updater 230.

In this case, the rule generator 310 may be mounted on a PC, a notebook computer, a server computer, a smart phone, a tablet, and an AI speaker, and used to generate a rule for recognizing a user's conversation.

The preprocessor 210 extracts input keyword information including information on a keyword set, which is a set composed of at least one keyword, from conversation data including information about a user's conversation.

The determination unit 220 determines whether there is an intention group corresponding to the input keyword information among a plurality of intention groups composed of at least one keyword set based on the inclusion relationship between sets.

In another embodiment, the determination unit 220 may determine whether the intention cluster exists based on an order according to the inclusion relationship of the at least one keyword set among the plurality of intention clusters.

In another exemplary embodiment, the determination unit 220 may select one intention group from among a plurality of intention clusters in which an inclusion relationship between each of the at least one keyword set included in the intention group and the keyword set included in the input keyword information does not hold. It may be determined that it corresponds to the input keyword information.

In another embodiment, the determination unit 220 determines whether each of the at least one keyword set included in the intention group and the intention of the keyword set included in the input keyword information are the same according to a predetermined criterion in each of the plurality of intention clusters. If the determination result is the same, based on the inclusion relationship between two keyword sets having the same intention, the keyword set included in the intention group can be replaced with the keyword set included in the input keyword information.

The updater 230 updates the knowledge base (KB) including the plurality of intention clusters by using the input keyword information according to the determination result.

In another embodiment, if there is an intention group corresponding to the input keyword information among the plurality of intention clusters, the updater 230 adds the keyword set included in the input keyword information to the intention cluster corresponding to the input keyword information, If there is no intention group corresponding to the input keyword information among the intention clusters of, a new intention cluster including a keyword set included in the input keyword information may be added to the KB.

3 is a flowchart of a rule-based intention inference method according to an embodiment of the present invention.

In this case, the intention reasoning apparatus may perform rule-based intention inference using a KB including a plurality of intention clusters composed of at least one keyword set.

In operation S310, the intention reasoning apparatus extracts input keyword information including information about a keyword set composed of at least one keyword from conversation data including information about a user's conversation.

For example, the intention reasoning device uses POS tagging as a preprocessing process of intention reasoning to detect unnecessary parts of speech such as interjection, punctuation, investigation, adverbs, etc., which do not play a large role as keywords among conversation words. You can extract the keyword set by removing it. In addition, if the conversation is performed in a foreign language other than Korean, the intention reasoning apparatus may extract a keyword set by performing an analysis such as a part-of-speech analysis in accordance with the characteristics of the foreign language.

Finally, in step S320, the intention reasoning apparatus determines a keyword set having a subset of the keyword set included in the input keyword information among at least one keyword set included in each of the plurality of intention groups.

In this case, the intention reasoning apparatus may determine the keyword set based on the reverse order of the order according to the inclusion relationship of the at least one keyword set among the plurality of intention groups.

On the other hand, the reverse order of the order according to the inclusion relationship of at least one keyword set between the plurality of intention group is a relationship in which any one keyword set included in the intent group # 1 is included in any one keyword set included in the intent group # 2 Exists, and when there is a relationship in which any one keyword set included in intent group # 2 is included in any one keyword set included in intent group # 3, intention cluster # 3, intention cluster # 2, and intent cluster # It can mean the order of 1.

That is, the intention reasoning apparatus inputs the plurality of intention clusters in order from the intention cluster containing the most keywords to the intention cluster containing the fewest keywords in the reverse order of the order of the inclusion of the keyword sets. By determining the inclusion relationship between each of the keyword sets corresponding to and at least one keyword set included in the intention group, a keyword set having a subset of the keyword sets included in the input keyword information may be determined. The user's intention can be deduced from the determined keyword set.

In other words, the intention inference apparatus includes an intention cluster including the most specific rule among a plurality of intention clusters included in the KB in order to more accurately infer the intention of the keyword set included in the input keyword information and accommodate various vocabularies. Inference can be proceeded in order of intent clusters containing the most general rules.

For example, referring to FIG. 5, the intention inference apparatus includes a keyword set having a subset of the keyword set included in the input keyword information in a reverse order (using the reverse_sort function on the fifth line) according to the inclusion relationship of the keyword set. Decision (line 6), the intention can be deduced.

More specifically, referring to FIG. 8, the intention reasoning apparatus starts from the intention cluster # 3, the intention cluster # 2, for the keyword set {'wonju', 'candidate', 'teach'} which is a keyword set included in the input keyword information. In order of intent group # 1, inference about intention clusters having the same intention may be performed in the reverse order of the order of inclusion of keyword sets. As a result, the intention inference apparatus is based on the keyword set included in the intention cluster # 2, {'wonju', 'market', 'candidate', 'teach'}, which is a keyword set included in the input keyword information. ',' Candidate ',' teach '} as a subset, it can be determined by the keyword set corresponding to the user's intention.

As described above, the rule-based intention reasoning method according to an embodiment of the present invention infers the intention by applying a rule divided in a cluster unit in the reverse order with respect to the user's conversation and infers the intention thereof, thereby completing the completeness of the inference. And it is effective to ensure soundness.

4 is a block diagram of a rule-based intention reasoning apparatus according to an embodiment of the present invention.

Referring to FIG. 4, the rule-based intention reasoning apparatus 400 according to an embodiment of the present invention includes a preprocessor 210 and a determiner 410.

In this case, the intention reasoning device 400 may be mounted on a PC, a notebook computer, a server computer, a smartphone, a tablet, and an AI speaker, and used to infer the conversation intention of the user.

The preprocessor 210 extracts input keyword information including information on a keyword set composed of at least one keyword from conversation data including information about a user's conversation.

The determination unit 410 determines a keyword set having a subset of the keyword set included in the input keyword information among at least one keyword set included in each of the plurality of intention groups.

In this case, the determination unit 510 may determine the keyword set based on the reverse order of the order according to the inclusion relationship of the at least one keyword set among the plurality of intention groups.

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed in a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium.

The computer-readable recording medium may include a magnetic storage medium (for example, a ROM, a floppy disk, a hard disk, etc.) and an optical reading medium (for example, a CD-ROM, a DVD, etc.).

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

Claims (5)

In the apparatus for generating a rule for recognition of a conversation, and inferring the user's intention based on the rule,
A pre-processing unit for extracting input keyword information including information on a keyword set, which is a set composed of at least one keyword, from conversation data including information about a user's conversation;
A determination unit that determines whether there is an intention cluster corresponding to the input keyword information among a plurality of intention clusters, which are a collection of at least one keyword set in which an inclusion relation between each other is not established, based on an inclusion relation between sets;
An update unit for updating a knowledge base (KB) including the plurality of intention clusters using the input keyword information according to the determination result;
A determination unit for determining a keyword set including a keyword set included in the input keyword information among at least one keyword set included in each of the plurality of intention groups;
The determining unit,
Determining the keyword set based on an inverse order of an order according to an inclusion relationship of at least one keyword set among the plurality of intention groups
The determining unit determines the keyword set in the order of the intention cluster containing the least keywords in the intention cluster containing the most keywords, characterized in that
Rule-based user intention inference device for dialogue recognition. The method of claim 1,
The determination unit
And a rule-based user intention inference apparatus for determining the presence of the intention cluster, based on an order according to the inclusion relationship of the at least one keyword set between the plurality of intention clusters. The method of claim 2,
The determination unit
Among the plurality of intention clusters, it is determined that one intention cluster corresponding to the input keyword information does not include an intention relationship between each of the at least one keyword set included in the intention group and the keyword set included in the input keyword information. Rule-based user intention reasoning device for dialogue recognition. The method of claim 1,
The determination unit
In each of the plurality of intention groups, it is determined whether each of the at least one keyword set included in the intention group and the intention of the keyword set included in the input keyword information are the same according to a predetermined criterion,
If the determination result is the same, based on the inclusion relationship between two keyword sets having the same intention, dialogue recognition, wherein the keyword set included in the intention group is selectively replaced with the keyword set included in the input keyword information. Rule-based user intention reasoning device. The method of claim 1,
The update unit
If there is an intention group corresponding to the input keyword information among the plurality of intention clusters, the keyword set included in the input keyword information is added to the intention group corresponding to the input keyword information.
If there is no intention group corresponding to the input keyword information among the plurality of intention clusters, a new intention group including a keyword set included in the input keyword information is added to the KB. Based user intention inference device.

Images