KR102099670B1 - The syntax grammar rules automatic generation method of understanding user query intention - Google Patents

Abstract

The present invention relates to a method for automatically generating a syntax rule to respond to a query having a variety of meanings with a higher probability in a query response system such as a chatbot system or a speech recognition system.
The automatic syntax rule generation method of the present invention is a method of automatically generating a syntax rule in advance while constructing a query response system that recognizes a query of a query terminal and generates response data through a language processing process in advance.
(a) The administrator terminal inputs the sentence including the object tag, (b) extracts the object tag from the input sentence, remembers the location, executes the morpheme analysis, and (c) places the object tag in the morpheme analysis result. To create a basic syntax rule, and (d) create a rule that expands a tag syntax by generating a rule that compares only vocabulary of nouns and verbs regardless of the morpheme result among morpheme analysis results, and (e) constructs a syntax for a predicate After going through a pre-determined process of executing rule expansion,
And creating a syntax rule and storing it in the question-and-answer system.

Description

Syntax rule auto-generating device and method for understanding the interrogator intent {THE SYNTAX GRAMMAR RULES AUTOMATIC GENERATION METHOD OF UNDERSTANDING USER QUERY INTENTION}

The present invention relates to a language processing technology, and in particular, to a language processing technology that is the basis of a chatbot system, a voice recognition system, etc. for generating a response to a user's query.

Recently, many companies have introduced a messenger-based question and answer system to respond to consumers outside of business hours of corporate customer centers. In addition, AI speakers are emerging that recognize the voice of the user's query and analyze the content of the question and respond appropriately. Examples of such systems include Amazon Echo, Google Home, SKT Anyone, KT GiGA Genie, and Kakao Mini. Such a question-and-answer system has been discussed and developed for a long time, but the range that can be used has been greatly expanded due to recent DB accumulation and technological development.

The Q & A system is a computer system in which a program that acquires patterns and rules of conversation based on a large number of conversation data between people determines the range of responses to user queries and responds to queries that fall within this range.

In order to analyze a user's query, a natural language processing process is required to recognize and process the language used by humans, such as morpheme analysis and syntax analysis. Among the various methods of implementing a question and answer system, a method using a syntax rule analyzes a user's query through a natural language processing process and then responds using a predefined syntax rule (Syntax Grammar).

Syntax rules are to create objects by collecting words from various sentences based on various criteria, and to define repetitive phrases as rules consisting of morphemes, vocabulary, parts of speech, and objects. In order to provide a service with such a question-and-answer system using syntax rules, many syntax rules must be created and managed continuously.

Chatbot systems, speech recognition systems, etc. using existing syntax rules show tremendous performance response rates for typical queries established by rules, but there is a problem in that the response rate is significantly reduced even if the patterns are not matched slightly. And even if it means the same, there is a limit that the probability that a pattern is established gradually decreases as the queries diversify.

To solve these problems and limitations, it is necessary to increase the probability of responding by generating the corresponding syntax rules for various queries. However, syntax rules are grammar rules that can express information and syntax structures such as morphemes, vocabulary, parts of speech, objects, etc. In the past, experts with knowledge in the field had to work directly.

While researching to solve the above problems, the inventors of the present invention have long researched and developed and tried to complete the present invention.

Even if an expert with knowledge in the field creates a syntax rule based on a lot of data when constructing an initial question-and-answer system, it costs a lot of money because the expert needs constant management such as adding and modifying syntax rules again while operating the system. Is done.

The object of the present invention is that anyone who is not an expert with knowledge in the field in managing a syntax rule can easily create an extended syntax rule, and in one way, the syntax rules are more diverse than the existing syntax rules. It provides an environment and methodology that can generate the same extended syntax rules.

On the other hand, other objects not specified in the present invention will be additionally considered within a range that can be easily deduced from the following detailed description and its effects.

The first aspect of the present invention is a method of automatically generating a syntax rule in advance while constructing a query response system that generates a response data through a language processing process by recognizing a query of the interrogator terminal in advance:

(a) The administrator terminal inputs the sentence including the object tag, (b) extracts the object tag from the input sentence, remembers the location, executes morpheme analysis, and (c) places the object tag in the morpheme analysis result. To create a basic syntax rule, and (d) create a rule that expands a tag syntax by generating a rule that compares only vocabulary of nouns and verbs regardless of the morpheme result among morpheme analysis results, and (e) constructs a syntax for a predicate After going through a pre-determined process of executing rule expansion,

And generating a syntax rule and storing it in the Q & A system.

In a method for automatically generating syntax rules in a question-and-answer system according to an exemplary embodiment of the present invention, it is preferable to restore a verb in step (d) to the vocabulary of the input sentence in step (a) before morpheme analysis.

In addition, in the method of automatically generating syntax rules in the question-and-answer system according to an exemplary embodiment of the present invention, it is preferable that step (e) extends the syntax rules by removing adverbs.

The second aspect of the present invention is software including data and program code to automatically generate a syntax rule in advance while recognizing a query of a queryer terminal and constructing a query response system that responds to response data through a language processing process in advance. A module for automatically generating syntax rules,

An input sentence module for inputting a sentence including an object tag by an administrator terminal, an object tag extraction module for extracting an object tag from an input sentence and storing its location, and then performing a morpheme analysis, an object in the morpheme analysis result of the object tag extraction module The basic syntax rule generation module that inserts tags into the original position to generate basic syntax rules, and the tag expansion syntax by creating a rule that compares only vocabulary of nouns and verbs regardless of the morpheme results among the morpheme analysis results of the object tag extraction module It includes a tag extension syntax rule generation module that generates rules, and a syntax rule extension module that executes an extension of syntax rules for predicates,

After passing through a predetermined process, it is characterized by generating a syntax rule and storing it in the question-and-answer system.

In the apparatus for automatically generating a syntax rule in the question and answer system according to an exemplary embodiment of the present invention, the apparatus for automatically generating the syntax rule is a module included in the question and answer system, or a network installed as a module installed outside the question and answer system. It can be connected via communication.

Through the above-described problem solving means of the present invention, the present invention has an advantage that anyone who is not an expert for defining syntax rules in a query response system can effectively define syntax rules.

In addition, when creating syntax rules, it is possible to respond to various query patterns compared to the existing ones by expanding tags and extending parts of speech.

In addition, the present invention has an advantage that maintenance and syntax rule management can be performed more efficiently than before.

On the other hand, even if the effects are not explicitly mentioned herein, it is noted that the effects described in the following specification expected by the technical features of the present invention and the potential effects thereof are handled as described in the specification of the present invention.

1 shows a schematic network configuration of a question and answer system 10 according to an exemplary embodiment of the present invention.
2 schematically shows an example of the system configuration of the apparatus 100 for automatically generating roles and syntax rules performed by the administrator terminal 15 in the present invention.
FIG. 3 is a schematic diagram showing the overall configuration and process of the apparatus 100 for automatically generating syntax rules according to an exemplary embodiment of the present invention.
4 is a diagram showing a basic syntax rule generation process in a preferred embodiment of the present invention.
5 is a diagram illustrating a process of generating a tag extension syntax rule in a preferred embodiment of the present invention.
6 is a diagram illustrating a process of extending a predicate rule in any preferred embodiment of the present invention.
7 is a diagram illustrating a process of adverb extension rules in a preferred embodiment of the present invention.
※ The accompanying drawings indicate that they are exemplified by reference for understanding the technical idea of the present invention, and the scope of the present invention is not limited thereby.

Hereinafter, a configuration of the present invention guided by various embodiments of the present invention and effects resulting from the configuration will be described with reference to the drawings. In the description of the present invention, when it is determined that the subject matter of the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

1 shows a schematic network configuration of a question and answer system 10 according to an exemplary embodiment of the present invention. The interrogator terminal 1 queries the query response system 10 through a communication network.

The interrogator terminal 1 is a device that generates and transmits voice through a wired or wireless communication network, and is a device capable of generating and transmitting voice, including a cordless phone, a wireless phone such as a smart phone, and the like. The query response system 10 is a computer system capable of recognizing and responding to a query when the interrogator terminal 1 accesses and queries, and uses various electronic devices and software used in web-based, application-based, and communication systems. It includes. In addition, it is a system that responds to query of interrogator based on speech recognition and language processing. It can also be expressed or configured as a chatbot system.

The question and answer system 10 outputs a response generated by a predetermined process through a speaker through voice. The query response system 10 includes hardware including a storage device, a communication device, an I / O interface, software for processing a process for recognizing a query and generating a response, and a server for managing execution and resources of the software. It includes.

The question and answer system server 11 is a server device that includes one or more hardware / software equipment, and manages the question and answer system 10.

The question and answer system 10 includes various databases. The database 12 records and manages language resources that recognize queries and generate responses.

The administrator terminal 15 connects to the server 11 and manages the query response system 10. In addition, an extended syntax rule is generated through the automatic generation device of the syntax rule of the present invention.

FIG. 2 schematically shows a system configuration of an automatic generating device 100 of roles and syntax rules performed by the administrator terminal 15 in the present invention.

The administrator terminal 15 accesses the query response system 10 to define and manage input sentences and entities required for responses. Then, the syntax rule automatic generation device 100 generates an extended syntax rule described below using an input sentence and a defined object. In addition, the response generation unit 150 generates response data by executing the syntax rule automatic generation device 100.

The syntax rule automatic generation device 100 builds data and program codes necessary for driving in a storage device. In the embodiment of Fig. 2 (a), it is configured to be included in the question and answer system 10. As shown in the embodiment of FIG. 2 (b), it may be separately configured outside the question and answer system 10, and data may be exchanged through network communication. In this case, the syntax rule generation unit 101 and the syntax rule automatic generation device 100 communicate in real time.

The automatic syntactic rule generation device 100 can effectively define anyone by simply defining the syntactic rules used in the query response system, such as a chatbot system or speech recognition system using the syntactic rules, without inputting expertise and defining input sentences and objects required for response. Device. The disadvantage of having to have basic knowledge of language grammar and morpheme analysis in order to define and build syntax rules by yourself, and that in order to respond to various query patterns, the syntax rules of the present invention must be added and extended one by one. It can be resolved by the generating device 100.

3 shows a schematic overall configuration and process of the apparatus 100 for automatically generating syntax rules according to a preferred embodiment of the present invention.

As shown in the figure, the syntax rule automatic generation device 100 executes a morpheme analysis after an input sentence module 110 in which an administrator terminal inputs a sentence including an object tag, extracts an object tag from the input sentence, stores the location thereof, and stores the location tag. The object tag extraction module 120, the basic syntax rule generation module 130, and the object tag extraction module 120 for generating basic syntax rules by inserting the object tag into the original location in the morpheme analysis result of the object tag extraction module 120 ), The tag expansion syntax rule generation module 140 that generates a tag expansion syntax rule by generating a rule that compares only vocabulary of nouns and verbs regardless of the morpheme result among the morpheme analysis results of), executes the expansion of syntax rules for predicates The syntax rule expansion module 150 may include a syntax rule generation module 160.

Each component of these syntax rule automatic generation devices 100 executes a time-series process as shown.

The above configurations will be described again. The input sentence module 110 of the apparatus 100 for automatically generating a syntax rule for an object defined by the administrator terminal accessing the query response system 10 in an administrator mode automatically generates a syntax rule based on an input sentence including an object tag. Is created by

Next, the object tag extraction module 120 extracts the object tag from the input sentence, remembers the location of the tag, and analyzes the text from which the object tag has been removed using a morpheme analyzer.

The basic syntax rule generation module 130 generates a basic syntax rule by inserting the object tag into the original location in the morpheme analysis result of the object tag extraction module 120.

The tag extension syntax rule generation module 140 generates rules for comparing vocabulary only with respect to nouns and verbs among morpheme analysis results in order to generate syntax rules applicable to various object values. At this time, the verb is restored to the vocabulary of the input sentence before morphological analysis.

The syntax rule expansion module 150 expands syntax rules for predicates to generate syntax rules applicable to various predicates in a sentence. In addition, the adjectives, verbs, and adverbs of sentences are explained in more detail and adverbs are removed. Semantically, since adverbs are less important in sentences, syntax rules can be extended by removing them.

The present invention extends syntax rules through such a process, and has an advantage that anyone who is not an expert for defining syntax rules can effectively define syntax rules. In addition, when creating syntax rules, it is possible to respond to a variety of query patterns compared to the existing ones by expanding tags and expanding suppliers.

Now, a detailed principle and content of the present invention will be described using the input sentence scenario in earnest. 4 schematically shows how basic syntax rule generation is achieved as an embodiment of the present invention.

The administrator terminal inputs a sentence for automatic generation of syntax rules to be constructed. Syntax rules are automatically generated through this input sentence, and the input sentence at this time becomes a general string type including the object tag. It is as follows [Example 1].

[Example 1] <WHERE> Seoul City Hall </ WHERE> Give me the location

Next, after finding the object tag in the input sentence, save the object name and object value as shown in Example 2 below.

[Example 2] WHERE = [Seoul City Hall]

Next, the basic syntax rule generation module remembers the location of the object tag stored in the input sentence and removes only the object tag while leaving the object value (S101). The extracted object tag separately stores the location value (S102).

Next, the morpheme analysis is performed through a morpheme analyzer on a sentence composed of only a general string with the object tag removed (S103). See Example 3 and Example 4 below.

[Example 3] 0 ~ 4: WHERE = [Seoul City Hall]

[Example 4] Give me the location of Seoul City Hall> Seoul / nng + City Hall / nng Location / nng Ali / vv + uh / ec + juju / vx + uh / ec

The tag storing the location value is relocated to the corresponding location value in the morpheme analyzed sentence (S105). The tag is expressed as "$ variable = @ object name", and the variable name uses the object name.

[Example 5] $ date = @ DATE

[Example 6] $ where = @ WHERE location / nng ali / vv + uh / ec + week / vx + uh / ec

When the position value of the morpheme and the position value of the tag collide during tag relocation (S107), the corresponding morpheme is separated into syllables (S108) to generate a basic syntax rule (S109). Consider Example 7 and Example 8.

[Example 7] <WHERE> Seoul City </ WHERE> Please let me know your location

[Example 8] $ where = @ WHERE + Blue location / nng ali / vv + uh / ec + week / vx + uh / ec

When the system matches the syntax rule pattern, the object value is included as a variable name in the result, so if there is the same object tag, the variable is added with a number to distinguish it. That is, the format is $ variable = @ object name, $ variable1 = @ object name.

[Example 9] Give <WHERE> Seoul </ WHERE> the location of <WHERE> Watch </ WHERE>

[Example 10] $ where = @ WHERE + to / j $ where1 = @ WHERE location / nng ali / vv + uh / ec + week / vx + uh / ec

5 schematically illustrates the process of the tag extension syntax rule generation module as an embodiment of the present invention.

Removing the object tag from the input sentence (S110), storing the location value of the extracted object tag (S111), and performing morpheme analysis (S112) are the same as the basic syntax rule generation process of FIG. 4 above. However, since the individual tag part can contain various vocabulary, the result of morpheme analysis of a sentence can be varied.

For this reason, in order to generate syntax rules applicable to various individual values, rules for comparing only vocabulary words for nouns and verb morphemes, which are major parts of speech, are generated (S113). Then, the object tag is rearranged (S115). At this time, the verb is restored to the original form of the vocabulary of the input sentence before the morpheme analysis (S116).

[Example 11] <WHERE> Seoul City Hall </ WHERE> Give me the location

[Example 12] $ where = @ WHERE location / * Ali / * + uh / ec + week / vx + uh / ec

Since even the sentence with the same meaning can vary in various parts of the predicate, it is preferable to expand the syntax rules for the predicate so that it can respond to it. 6 shows such an embodiment. 4 and 5, a basic syntax rule was generated and a tag extension syntax rule was generated (S120). Then expand the predicate rules.

The extension of the predicate rules can be sequentially expanded for forms such as the main language + auxiliary language (S121), long language + ending (S123), noun + verb derivative suffix + ending (S125), noun + positive noun + ending (S127). have. Through such a process, it is possible to generate the main verb auxiliary verb extension rule (S121a), the verb ending rule (S123a), the verb derived suffix extension rule (S125a), and the positive designator extension rule (S127a).

[Example 13] Main usage + auxiliary usage

$ where = @ WHERE location / nng ali / vv + uh / ec + main / vx + uh / ec

$ where = @ WHERE location / nng ali / vv + *

[Example 14] Longan + Mother

$ where = @ WHERE location / nng ali / vv + uh / ec + main / vx + uh / ec

$ where = @ WHERE location / nng ali / vv + uh / ec + share / vx + *

[Example 15] Noun + Verb Derivative Suffix + Mother

$ where = @ WHERE location / nng report / nng + ha / xsv + ah / ec

$ where = @ WHERE location / nng report / nng + *

[Example 16] Noun + Positive Designator + Mother

$ where = @ WHERE location / nng + / j seoul / nng + square / nng next / nng + i / vcp + ni / e

$ where = @ WHERE location / nng + / j seoul / nng + square / nng next / nng + *

Adverbs serve to explain and decorate adjectives, verbs, and adverbs of sentences in more detail. As shown in Fig. 7, such adverbs are reduced in importance in sentences, thereby eliminating syntax rules. The "some" of Example 17 was removed from Example 18.

[Example 17] $ where = @ WHERE location / nng some / mag tell me / v + uh / e

[Example 18] $ where = @ WHERE location / nng tell / v + uh / e

For reference, in the question and answer system according to an embodiment of the present invention, a method for automatically generating syntax rules may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, or the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and available to those skilled in computer software.

Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs, DVDs, magnetic-optical media such as floptical disks, and ROM, RAM, flash Hardware devices specifically configured to store and execute program instructions, such as memory, may be included. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine codes such as those produced by a compiler. The above-described hardware device may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

The scope of protection of the present invention is not limited to the description and expression of the embodiments explicitly described above. In addition, it is pointed out again that the protection scope of the present invention may not be limited by obvious changes or substitutions in the technical field to which the present invention pertains.

Claims (5)

As a method of automatically constructing a syntax rule in advance while constructing a query response system that recognizes a query of a query terminal and generates response data through a language processing process in advance:
(a) The administrator terminal inputs the sentence including the object tag, (b) extracts the object tag from the input sentence, remembers the location, executes the morpheme analysis, and (c) places the object tag in the morpheme analysis result. To create a basic syntax rule, and (d) create a rule that expands the syntax of a tag by generating a rule that compares only vocabulary of nouns and verbs regardless of the morpheme result among morpheme analysis results, and (e) adverbs for a predicate After going through a pre-determined process to run the extension of the syntax rules by removing
And generating a syntax rule and storing it in the question-and-answer system.
Syntax rule automatic generation device, which is a software module that includes data and program code that automatically generates syntax rules in advance while constructing a query response system that recognizes the query of the interrogator terminal and responds to the response data through a language processing process in advance. Is,
An input sentence module for inputting a sentence including an object tag by an administrator terminal, an object tag extraction module for extracting an object tag from an input sentence and storing its location, and then performing a morpheme analysis, an object in the morpheme analysis result of the object tag extraction module A basic syntax rule generation module that inserts tags into the original position to generate basic syntax rules, and a tag expansion syntax by creating a rule that compares only vocabulary of nouns and verbs regardless of morpheme results among morpheme analysis results of the object tag extraction module Includes a tag extension syntax rule generation module that generates rules, and a syntax rule extension module that executes the extension of syntax rules by removing adverbs for predicates,
A device for automatically generating syntax rules of a question-and-answer system, characterized in that the syntax rules are generated and stored in the question-and-answer system.
According to claim 2,
The apparatus for automatically generating the syntax rules is a module included in the question and answer system, or a module installed outside the question and answer system and connected through network communication.
delete delete

Images