KR20200074603A - Sentences automatically generating apparatus for generating a corpus associated with the new object name word and operating method thereof - Google Patents

Abstract

The present invention provides an apparatus for automatically generating a sentence for generating a corpus associated with a new object name and an operating method thereof. A user can be supported to easily compose the corpus used in a language model of a speech recognizer.

Description

{SENTENCES AUTOMATICALLY GENERATING APPARATUS FOR GENERATING A CORPUS ASSOCIATED WITH THE NEW OBJECT NAME WORD AND OPERATING METHOD THEREOF}

The present invention relates to techniques for constructing a corpus used in a language model of a speech recognizer.

Recently, as electronic devices equipped with artificial intelligence functions are released, a technology that recognizes a user's voice and performs control when a user applies a command by voice has attracted attention.

In order to implement a speech recognition device having high performance, it is necessary to secure a large number of corpuses. The corpus is a collection of samples of languages with a specific purpose for the study of natural languages. With the development of computers, corpus analysis has been facilitated, and the natural language is often morphologically analyzed for analysis accuracy. Probability/statistical techniques and time-series approaches can be used to grasp the whole, confirm the frequency and distribution of language, and are essential for modern linguistics research. It can be seen that the natural science methodology is most successfully applied to the humanities.

In a speech recognizer using a language model based on probability distribution, a language model based on probability distribution is created based on a corpus and calculates a predicted probability of a specific word by reflecting a probability distribution for various word sequences in the corpus. For this reason, in order to reflect the new word in the language model, it is necessary to make a corpus containing a large number of sentences of various types containing the new word and reflect it in the language model.

In particular, when such a new individual name is created in that the individual name such as a person's name, a place name, and an organization name is often newly generated, a corpus consisting of sentences including the new entity name must be created and reflected in the language model to reflect the new entity name. The included voice can be accurately recognized.

However, in order to generate a corpus associated with a new entity name, it is inevitable that many difficulties exist for the developer to create a sentence containing the new entity name. Therefore, there is a need for a study on an automatic sentence generation technique capable of automatically generating new sentences for generating a corpus associated with a new individual name.

The present invention provides a sentence automatic generation device and an operation method for automatically generating new sentences for generating a corpus associated with a new object name word, so that a user can more easily construct a corpus used for a language model of a speech recognizer. I want to support.

An automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention includes a plurality of predetermined entity name words and a predetermined category identifier for identifying each category of the plurality of entity name words A word table holding unit that stores and maintains a word table recorded in correspondence, a sentence database storing a plurality of preset sentences including one or more of the plurality of object name words in a sentence, a new object name word from a user, and When a plurality of example sentences in which the new entity name word is inserted and a first category identifier indicating a category of the new entity name word are input, an automatic sentence generation command for generating a corpus associated with the new entity name word is applied. , An entity name word extracting unit extracting a plurality of first entity name words recorded corresponding to the first category identifier from the word table, and when the plurality of first entity name words are extracted, the plurality from the sentence database A sentence extracting unit for extracting a plurality of first sentences including one or more first entity name words in a sentence, and when the plurality of first sentences are extracted, the plurality of example sentences among the plurality of first sentences A candidate that selects sentences that have a sentence similarity to and exceeds a preset threshold, and selects words included in the selected sentences one or more times among the plurality of first object name words as a plurality of candidate object name words When the word selector and the plurality of candidate entity name words are selected, sentences including one or more of the plurality of candidate entity name words are extracted from the sentence database as a plurality of candidate sentences, and the plurality of candidate sentences are And a new sentence generating unit generating a plurality of new sentences including the new object name word by replacing the included plurality of candidate object name words with the new object name word.

In addition, an operation method of an automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention identifies a plurality of predetermined entity name words and categories of each of the plurality of entity name words Storing and maintaining a word table in which a predetermined category identifier for correspondence is recorded, and maintaining a sentence database in which a plurality of preset sentences including one or more words of the plurality of object names are stored in a sentence, a user A new entity name word, a plurality of example sentences in which the new entity name word is inserted, and a first category identifier indicating a category of the new entity name word are input, and a sentence for generating a corpus associated with the new entity name word Extracting a plurality of first entity name words recorded corresponding to the first category identifier from the word table when the automatic generation command is applied, and when the plurality of first entity name words are extracted, from the sentence database Extracting a plurality of first sentences including one or more words of the plurality of first entity names in a sentence, and when the plurality of first sentences are extracted, the plurality of example sentences among the plurality of first sentences Selecting sentences that have a sentence similarity with and exceeding a preset threshold, and selecting words included in the selected sentences one or more of the plurality of first object name words as a plurality of candidate object name words And when the plurality of candidate entity name words are selected, extracts sentences including one or more of the plurality of candidate entity name words from the sentence database into a plurality of candidate sentences, and is included in the plurality of candidate sentences. And generating a plurality of new sentences including the new entity name word by replacing the plurality of candidate entity name words with the new entity name word.

The present invention provides a sentence automatic generation device and an operation method for automatically generating new sentences for generating a corpus associated with a new object name word, so that a user can more easily construct a corpus used for a language model of a speech recognizer. Can help.

1 is a diagram illustrating a structure of an automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention.
2 is a flowchart illustrating an operation method of an automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention.

Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings. It should be understood that this description is not intended to limit the invention to specific embodiments, and includes all modifications, equivalents, or substitutes included in the spirit and scope of the invention. In describing each drawing, similar reference numerals have been used for similar components, and unless otherwise defined, all terms used in this specification, including technical or scientific terms, have common knowledge in the technical field to which the present invention pertains. It has the same meaning as commonly understood by people who have it.

In this document, when it is said that a part "includes" a certain component, this means that other components may be further included rather than excluding other components unless specifically stated to the contrary. In addition, in various embodiments of the present invention, each component, functional blocks, or means may be composed of one or more sub-components, and the electrical, electronic, and mechanical functions performed by each component are electronic Circuits, integrated circuits, ASIC (Application Specific Integrated Circuit) may be implemented with various known elements or mechanical elements, each of which may be implemented separately or may be implemented by integrating two or more into one.

On the other hand, the steps of the blocks or flow charts of the accompanying block diagrams are computer program instructions that are mounted on a processor or memory of data processing equipment such as a general purpose computer, a special purpose computer, a portable notebook computer, and a network computer to perform designated functions. It can be interpreted as meaning. Since these computer program instructions can be stored in a memory provided in a computer device or in a computer readable memory, the functions described in the blocks of the block diagrams or steps in the flowchart are produced as an article containing instruction means for performing them. It may be. In addition, each block or step may represent a module, segment, or portion of code that includes one or more executable instructions for executing the specified logical function(s). It should also be noted that in some alternative embodiments, the functions mentioned in blocks or steps may be executed out of order. For example, two blocks or steps shown in succession may be performed substantially simultaneously, or may be performed in reverse order, and in some cases, some blocks or steps may be omitted.

1 is a diagram illustrating a structure of an automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention.

Referring to FIG. 1, the automatic sentence generation device 110 for generating a corpus associated with a new object name word according to the present invention includes a word table holding unit 111, a sentence database 112, and an object name word extraction unit 113 , A sentence extraction unit 114, a candidate word selection unit 115, and a new sentence generation unit 116.

The word table holding unit 111 stores and maintains a word table in which a plurality of predefined entity name words and a predetermined category identifier for identifying each of the plurality of entity name words are recorded in correspondence.

Here, the word “individual name” means a proper noun or a noun phrase such as a person's name, a place name, an organization name, and a name of an object. In addition, the plurality of pre-specified object name words refer to the object name words in which the corpus for voice recognition has already been generated, and the category refers to the classification to which each object name word belongs, and the specific object name word is'Bulgogi Pizza' If you say'Bulgogi Pizza' can belong to the category of'Food', so the category for the object name'Bulgogi Pizza' can be'Food'.

At this time, the predefined category identifier for the category of'food' is called'identifier 1', the predefined identifier for the category of'electronic products' is called'identifier 2', and the predefined name for the category of'mobile phone accessories'. If the identifier is'identifier 3', information may be recorded in the word table as shown in Table 1 below.

Multiple entity words Category identifier Black Tiger Shrimp Pizza Identifier 1 hamburger Identifier 1 Cheonggukjang Identifier 1 Kingfront Seafood Pizza Identifier 1 Smartphone Identifier 2 Selfie stick Identifier 3 ... ...

The sentence database 112 stores a plurality of preset sentences including one or more words of the plurality of entity names in the sentence.

The plurality of sentences stored in the sentence database 112 means sentences that have already been configured for use as a corpus for a speech recognizer.

For example, a plurality of sentences may be stored in the sentence database 112 as shown in Table 2 below.

Multiple entity words Multiple sentences Black Tiger Shrimp Pizza 1. It is recommended to try Hancom Pizza's new product Black Tiger Shrimp Pizza.
2. This time, the new Black Tiger Shrimp Pizza is coming out.
... hamburger 1. Eat a delicious hamburger.
2. Give me a hamburger.
... Cheonggukjang 1. The smell of Cheonggukjang stimulates the nose.
2. I miss Cheonggukjang made in my hometown.
... Kingfront Seafood Pizza 1. Hancom Pizza sells a special set of summer new King's Seafood Pizza.
... Smartphone 1. You can enjoy the game using your smartphone.
... Selfie stick 1. Use selfie sticks when taking pictures.
2. You can take a selfie using the selfie stick.
... ... ...

The entity name word extracting unit 113 inputs a new entity name word, a plurality of example sentences in which the new entity name word is inserted, and a first category identifier indicating a category of the new entity name word, while the new entity name word extraction unit 113 is input. When a sentence automatic generation command for generating a corpus associated with an entity name word is applied, a plurality of first entity name words recorded corresponding to the first category identifier is extracted from the word table.

For example, when the new entity name word input from the user is called'Hulala Pizza', the plurality of example sentences are as shown in Table 3 below, and when the first category identifier is'identifier 1', the entity name word The extracting unit 113 recalls the words of the'Black Tiger Shrimp Pizza, Hamburger, Cheonggukjang, Kingpron Seafood Pizza, ...' recorded corresponding to'Identifier 1'from the word table shown in Table 1 above. It may be extracted as a plurality of first entity name words.

Multiple example sentences We recommend Hancom Pizza's new product, Hulala Pizza.
Hancom Pizza sells new hula pizza.
You can order hula pizzas only at the store.

When the plurality of first entity name words are extracted, the sentence extraction unit 114 extracts a plurality of first sentences including one or more of the plurality of first entity name words from the sentence database 112.

In relation to this, as in the above-described example, if the'black tiger shrimp pizza, hamburger, chungkukjang, king pron seafood pizza, ...' is said to be extracted with the plurality of first object name words, the sentence extraction unit 114 ) From the sentence database 112 shown in Table 2 above, the plurality of agents containing one or more words in the sentence'Black Tiger Shrimp Pizza, Hamburger, Cheonggukjang, King Front Seafood Pizza, ...' One sentence can be extracted, and the results can be expressed as shown in Table 4 below.

Multiple first sentences 1. It is recommended to try Hancom Pizza's new product Black Tiger Shrimp Pizza .
2. This time, the new black tiger shrimp pizza was released, so I waited for the release date and ate it.
... 1. Eat a delicious hamburger .
2. Give me a hamburger .
... 1. The smell of Cheonggukjang stimulates the nose.
2. I miss Cheonggukjang made in my hometown.
... 1. Hancom Pizza sells a special set of summer new King's Seafood Pizza .
... ...

When the plurality of first sentences are extracted, the candidate word selector 115 selects sentences among the plurality of first sentences having a sentence similarity with the plurality of example sentences exceeding a preset threshold, and the plurality of first sentences Among the first individual word words, words included in the selected sentences more than once are selected as a plurality of candidate individual name words.

At this time, according to an embodiment of the present invention, the candidate word selector 115 may include a corresponding example sentence selector 117, a sentence selector 118, and a candidate selector processor 119.

When the plurality of first sentences are extracted, the corresponding example sentence selecting unit 117 extracts the example sentences having the maximum sentence similarity with each of the plurality of first sentences among the plurality of example sentences, and the plurality of first sentences. Select the corresponding example sentences for each of them.

The sentence selection unit 118 compares the sentence similarity between the corresponding example sentences for each of the plurality of first sentences and each of the plurality of first sentences, and then compares the sentence similarity with the corresponding example sentence among the plurality of first sentences Selects sentences that exceed the threshold.

The candidate selection processing unit 119 selects words included in the selected sentences one or more of the plurality of first entity name words as the plurality of candidate entity name words.

In this case, according to an embodiment of the present invention, the sentence similarity used in the candidate word selector 115 may be the length of the longest common subsequence (LCS) of the strings constituting the sentences to be compared. have.

In relation to this, when the plurality of example sentences are as shown in Table 3 and the plurality of first sentences are as shown in Table 4, the operation of the candidate word selector 115 is described in detail as follows. same.

When the plurality of first sentences, such as Table 4, are extracted through the sentence extraction unit 114, the corresponding example sentence selection unit 117 includes a plurality of examples, as shown in Table 3, for each of the plurality of first sentences. You can calculate the sentence similarity with each of the sentences.

At this time, the corresponding example sentence selector 117 may calculate the length of the LCS between each sentence and the plurality of example sentences for each of the plurality of first sentences with sentence similarity.

LCS means a continuous string that matches two strings. When there are'ACAYKP' and'CAPCAK', the LCS between the two strings is'ACAK', and the length of the LCS is '4'.

In relation to this, the corresponding example sentence selector 117 can calculate the length of the LCS between each of the sentences,'I recommend eating and recommending the new Hancom Pizza Black Tiger Shrimp Pizza.' The length of the LCS between each of the sentences,'I waited for the release date and ate it, because the new Black Tiger Shrimp Pizza is coming out' and the above example sentences can be calculated with sentence similarity. In this way, the corresponding example sentence selector 117 may calculate sentence similarity with each of the plurality of example sentences for each of the plurality of first sentences.

In this way, when sentence similarity with the plurality of example sentences for each of the plurality of first sentences is calculated, the corresponding example sentence selector 117 each of the plurality of first sentences among the plurality of example sentences An example sentence having the maximum sentence similarity to and may be selected as a corresponding example sentence for each of the plurality of first sentences.

For example, in the case of the sentence,'Hancom Pizza's new Black Tiger Shrimp Pizza is recommended.' If the example sentence with the maximum sentence similarity is'Hancom Pizza's new Hula Pizza,' the corresponding example sentence selector (117) can select the example sentence'Hancom Pizza's new hula pizza' as the corresponding example sentence for the sentence'Hancom Pizza's new Black Tiger Shrimp Pizza.

As such, when the selection of the corresponding example sentences for each of the plurality of first sentences is completed, the sentence selection unit 118 may respectively correspond to the corresponding example sentences for each of the plurality of first sentences and the plurality of first sentences. The sentence similarity between the first sentence among the plurality of first sentences may be selected by comparing the sentence similarity between the sentences.

For example,'I recommend eating and recommending Hancom Pizza's new Black Tiger Shrimp Pizza' and the corresponding example sentence'Hancom Pizza's new Hula Pizza' is recommended. If the length of the LCS of the liver is said to exceed the threshold, the sentence selection unit 118 selects the sentence'Hancom Pizza New Black Tiger Shrimp Pizza directly and recommends it' as a sentence having sentence similarity that exceeds the threshold. Can.

In this way, the sentence selection unit 118 may select sentences among the plurality of first sentences whose sentence similarity with the corresponding example sentence exceeds the threshold. At this time, in this embodiment, it is assumed that the result of selecting sentences in which the sentence similarity with the corresponding example sentence among the plurality of first sentences exceeds the threshold is as shown in Table 5 below.

Selected sentences It is recommended to try Hancom Pizza's new Black Tiger Shrimp Pizza. Hancom Pizza sells a limited edition of the new Summer King's Seafood Pizza special set. ...

When sentences are selected in the sentence selection unit 118, as shown in Table 5, the candidate selection processing unit 119 is the first plurality of'black tiger shrimp pizza, hamburger, cheonggukjang, king pron seafood pizza, ...' Words included in the selected sentences more than once among the object name words may be selected as the plurality of candidate object name words.

If, among the plurality of first individual words words of'Black Tiger Shrimp Pizza, Hamburger, Cheonggukjang, King Front Seafood Pizza, ...', the words included in the selected sentences more than once are'Black Tiger Shrimp. In the case of'Pizza, KingFront Seafood Pizza', the candidate selection processing unit 119 may select the entity names words'Black Tiger Shrimp Pizza, KingFront Seafood Pizza' as the plurality of candidate entity name words.

When the plurality of candidate entity name words are selected through the candidate word selector 115 according to the embodiment described so far, the new sentence generator 116 generates one of the plurality of candidate entity name words from the sentence database 112. The new object name word is included by extracting the sentences containing the above into a plurality of candidate sentences and replacing the plurality of candidate object name words included in the plurality of candidate sentences with the new object name word Create multiple new sentences.

At this time, according to an embodiment of the present invention, the new sentence generation unit 116 may include a candidate sentence table generation unit 120 and a new sentence generation processing unit 121.

The candidate sentence table generator 120 extracts, as the plurality of candidate sentences, sentences containing one or more of the plurality of candidate object name words from the sentence database 112 when the plurality of candidate object name words are selected. A candidate sentence table in which each of the plurality of candidate sentences is recorded is generated.

For example, when the plurality of candidate object names words are called'Black Tiger Shrimp Pizza, King Front Seafood Pizza', the candidate sentence table generating unit 120 may generate a'Black Tiger Shrimp Pizza, King Front' from the sentence database 112. After extracting the sentences containing one or more food pizzas into the plurality of candidate sentences, a candidate sentence table in which each of the plurality of candidate sentences is recorded may be generated, wherein the candidate sentence table is shown in Table 6 below. Can be represented together.

Multiple candidate entity names words Multiple candidate sentences Black Tiger Shrimp Pizza It is recommended to try Hancom Pizza's new Black Tiger Shrimp Pizza. This time, the new Black Tiger Shrimp Pizza was released, so I waited for the release date and tried it. This new Black Tiger Supreme pizza has no fishy taste. ... Kingfront Seafood Pizza Hancom Pizza sells a limited edition of the new Summer King's Seafood Pizza special set. The pizza I ordered is KingFront Seafood Pizza, which is large in the original dough. Still, I was curious about the taste of KingFron Seafood Pizza, which goes well with new menus these days. ...

The new sentence generation processing unit 121 replaces the plurality of candidate entity name words included in the plurality of candidate sentences in the candidate sentence table with the new entity name word, so that the plurality of new entity name words are included. New sentences can be created.

In relation to this, when it is said that the candidate sentence table is generated as shown in Table 6, the new sentence generation processing unit 121 recalls the'Black Tiger Shrimp Pizza, King Front Seafood Pizza' included in the plurality of candidate sentences. By substituting the new entity name word'Hulala Pizza', new sentences as shown in Table 7 below can be generated.

New sentences It is recommended to try Hancom Pizza's new Hula Pizza. This time, I was waiting for the release date and ate it. This new hula pizza has no fishy taste. ... Hancom Pizza will be selling a special set of summer new hula pizza. The pizza I ordered is a large size with hula pizza and original dough. Still, I was wondering how the hula pizza, which goes well with the new menu, tastes good. ...

In this case, according to an embodiment of the present invention, the automatic sentence generation device 110 for generating a corpus associated with a new entity name word may further include a new storage unit 122.

When the plurality of new sentences are generated, the new storage unit 122 records the new entity name word and the first category in the word table in correspondence with each other, and stores the plurality of new sentences in the sentence database 112. do.

After all, the automatic sentence generation device 110 for generating a corpus associated with a new object name word according to the present invention is input by the user when certain example sentences including the new object name word and the new object name word are input. By comparing the sentence similarity between the existing sentences containing the object name word and the object word of the same category and the example sentences, the sentence similarity with the example sentences among the existing sentences exceeds a predetermined threshold. After selecting, extracting the word of the object included in the selected sentence, and re-selecting the existing sentences containing one or more of the extracted word of the object, the word of the object included in the selected existing sentence as the new object word By substituting, it is possible to automatically generate new sentences for generating a corpus associated with the new entity name word.

2 is a flowchart illustrating an operating method of an automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention.

In step S210, a word table in which a predetermined plurality of entity name words and a predetermined category identifier for identifying each category of the plurality of entity name words are correspondingly recorded is stored and maintained.

In step S220, a sentence database in which a plurality of preset sentences including one or more words of the plurality of entity names are stored is maintained.

In step S230, a new entity name word, a plurality of example sentences in which the new entity name word is inserted, and a first category identifier indicating a category of the new entity name word are input from the user, and the new entity name word is input to the new entity name word. When a command for automatically generating a sentence for generating an associated corpus is applied, a plurality of first entity word words recorded corresponding to the first category identifier is extracted from the word table.

In step S240, when the plurality of first entity name words are extracted, a plurality of first sentences including one or more of the plurality of first entity name words in the sentence are extracted from the sentence database.

In step S250, when the plurality of first sentences are extracted, among the plurality of first sentences, sentences having similarity with the plurality of example sentences exceeding a preset threshold, and selecting the plurality of first sentences Among the words of the object name, words included in the selected sentences more than once are selected as a plurality of candidate object names.

In step S260, when the plurality of candidate entity name words are selected, sentences including one or more of the plurality of candidate entity name words are extracted from the sentence database as a plurality of candidate sentences, and the plurality of candidate sentence sentences By replacing the plurality of candidate entity name words included in the new entity name word, a plurality of new sentences including the new entity name word are generated.

At this time, according to an embodiment of the present invention, in step S250, when the plurality of first sentences are extracted, among the plurality of example sentences, the example sentences having the maximum sentence similarity with each of the plurality of first sentences Selecting a corresponding example sentence for each of the plurality of first sentences, comparing the similarity of sentences between the corresponding example sentence for each of the plurality of first sentences and each of the plurality of first sentences, and comparing the sentence similarities Selecting sentences in which sentence similarity with a corresponding example sentence among the first sentences exceeds the threshold value, and words included in the selected sentences one or more times among the plurality of first object name words; And selecting candidate candidate name words.

At this time, according to an embodiment of the present invention, the sentence similarity used in step S250 may be the length of the longest common subsequence (LCS) for the strings constituting the sentences to be compared.

Further, according to an embodiment of the present invention, when the plurality of candidate entity name words are selected in step S260, the plurality of candidates may include sentences including one or more of the plurality of candidate entity name words from the sentence database. Generating a candidate sentence table in which each of the plurality of candidate sentences is recorded after extracting them into sentences and the plurality of candidate object names words included in the plurality of candidate sentences in the candidate sentence table, the new object name And generating the plurality of new sentences including the new entity name word by replacing the word with a word.

Further, according to an embodiment of the present invention, an operation method of an automatic sentence generation device for generating a corpus associated with the new object name word, when the plurality of new sentences are generated, the new object name word and the word in the word table The method may further include storing the first categories in correspondence with each other and storing the plurality of new sentences in the sentence database.

The operation method of the automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention has been described above with reference to FIG. 2. Here, an operation method of an automatic sentence generation device for generating a corpus associated with a new individual name word according to an embodiment of the present invention is an automatic sentence generation device 110 for generating a corpus associated with a new individual name word described with reference to FIG. 1. ), so a detailed description thereof will be omitted.

An operation method of an automatic sentence generation device for generating a corpus associated with a new object name word according to an embodiment of the present invention may be implemented as a computer program stored in a storage medium for execution through combination with a computer.

In addition, an operation method of an automatic sentence generation device for generating a corpus associated with a new entity name word according to an embodiment of the present invention is implemented in the form of program instructions that can be executed through various computer means to be recorded in a computer-readable medium. Can. 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 usable by those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs, DVDs, and magnetic media such as floptical disks. Includes hardware devices specifically configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language code that can be executed by a computer using an interpreter, etc., as well as machine language codes produced by a compiler.

As described above, in the present invention, specific matters such as specific components and the like have been described by limited embodiments and drawings, but they are provided only to help the overall understanding of the present invention, and the present invention is not limited to the above embodiments , Anyone who has ordinary knowledge in the field to which the present invention pertains can make various modifications and variations from these descriptions.

Accordingly, the spirit of the present invention should not be limited to the described embodiments, and should not be determined, and all claims that are equivalent or equivalent to the scope of the claims as well as the claims described below belong to the scope of the spirit of the invention. .

110: automatic sentence generation device for generating a corpus associated with a new entity name word
111: word table maintenance unit 112: sentence database
113: object name word extraction unit 114: sentence extraction unit
115: candidate word selection unit 116: new sentence generation unit
117: corresponding example sentence selector 118: sentence selector
119: candidate selection processing unit 120: candidate sentence table generation unit
121: new sentence generation processing unit 122: new storage unit

Claims (12)

A word table holding unit which stores and maintains a word table in which a plurality of predefined object name words and a predetermined category identifier for identifying each category of the plurality of object name words are recorded;
A sentence database storing a plurality of preset sentences including one or more words of the plurality of entity names in the sentence;
A new entity name word, a plurality of example sentences in which the new entity name word is inserted, and a first category identifier indicating a category of the new entity name word are input from a user, for generating a corpus associated with the new entity name word. An object name word extraction unit extracting a plurality of first object name words recorded corresponding to the first category identifier from the word table when the sentence automatic generation command is applied;
A sentence extracting unit that extracts a plurality of first sentences including one or more words of the plurality of first object names from the sentence database when the plurality of first object name words are extracted;
When the plurality of first sentences are extracted, among sentences among the plurality of first sentences, sentences with similarity to the plurality of example sentences exceed a preset threshold, and among the plurality of first entity name words A candidate word selector for selecting words included in the selected sentences more than once as a plurality of candidate entity name words; And
When the plurality of candidate entity name words are selected, sentences including one or more of the plurality of candidate entity name words are extracted from the sentence database as a plurality of candidate sentences, and included in the plurality of candidate sentences A new sentence generation unit that generates a plurality of new sentences including the new object name word by replacing a plurality of candidate object name words with the new object name word
Automatic sentence generation device for generating a corpus associated with a new object name word comprising a. According to claim 1,
The candidate word selector
When the plurality of first sentences are extracted, among the plurality of example sentences, an example sentence having a maximum sentence similarity with each of the plurality of first sentences is selected as a corresponding example sentence for each of the plurality of first sentences Corresponding example sentence selector;
A sentence similarity between a corresponding example sentence for each of the plurality of first sentences and a corresponding example sentence among the plurality of first sentences by comparing a sentence similarity between each of the plurality of first sentences Sentence selection unit for selecting them; And
A candidate selection processing unit for selecting words included in the selected sentences one or more of the plurality of first entity name words as the plurality of candidate entity name words
Automatic sentence generation device for generating a corpus associated with a new object name word comprising a. According to claim 2,
The sentence similarity used in the candidate word selection unit is a sentence automatic generation device for generating a corpus associated with a new entity name word, which is the length of the longest common subsequence (LCS) for a string constituting sentences to be compared. . According to claim 1,
The new sentence generation unit
When the plurality of candidate entity name words are selected, the sentences including one or more of the plurality of candidate entity name words are extracted from the sentence database as the plurality of candidate sentences, and then each of the plurality of candidate sentences is recorded. A candidate sentence table generator for generating a sentence table; And
In the candidate sentence table, a new sentence generating the plurality of new sentences including the new object name word by replacing the plurality of candidate object name words included in the plurality of candidate sentences with the new object name word Processing
Automatic sentence generation device for generating a corpus associated with a new object name word comprising a. According to claim 1,
When the plurality of new sentences are generated, a new storage unit that records the new entity name word and the first category in the word table in correspondence with each other, and stores the plurality of new sentences in the sentence database.
Automatic sentence generation device for generating a corpus associated with a new object name word further comprising. Storing and maintaining a word table in which a plurality of predefined object name words and a predetermined category identifier for identifying each category of the plurality of object name words are recorded in correspondence;
Maintaining a sentence database in which a plurality of preset sentences including one or more words of the plurality of entity names are stored;
A new entity name word, a plurality of example sentences in which the new entity name word is inserted, and a first category identifier indicating a category of the new entity name word are input from a user, for generating a corpus associated with the new entity name word. Extracting a plurality of first entity name words recorded corresponding to the first category identifier from the word table when an automatic sentence generation command is applied;
Extracting a plurality of first sentences including one or more of the plurality of first object name words from the sentence database when the plurality of first object name words are extracted;
When the plurality of first sentences are extracted, among sentences among the plurality of first sentences, sentences with similarity to the plurality of example sentences exceed a preset threshold, and among the plurality of first entity name words Selecting words included in the selected sentences one or more times as a plurality of candidate entity name words; And
When the plurality of candidate entity name words are selected, sentences including one or more of the plurality of candidate entity name words are extracted from the sentence database as a plurality of candidate sentences, and included in the plurality of candidate sentences Generating a plurality of new sentences including the new entity name word by replacing a plurality of candidate entity name words with the new entity name word
A method of operating a sentence automatic generation device for generating a corpus associated with a new entity name word comprising a. The method of claim 6,
The step of selecting the plurality of candidate entity name words
When the plurality of first sentences are extracted, among the plurality of example sentences, an example sentence having a maximum sentence similarity with each of the plurality of first sentences is selected as a corresponding example sentence for each of the plurality of first sentences To do;
A sentence similarity between a corresponding example sentence for each of the plurality of first sentences and a corresponding example sentence among the plurality of first sentences by comparing a sentence similarity between each of the plurality of first sentences Selecting them; And
Selecting words included in the selected sentences at least once among the plurality of first entity word words as the plurality of candidate entity name words.
A method of operating a sentence automatic generation device for generating a corpus associated with a new entity name word comprising a. The method of claim 7,
The sentence similarity used in the step of selecting the plurality of candidate entity name words is a corpus associated with a new entity name word that is the length of the Longest Common Subsequence (LCS) for the strings that constitute the sentences to be compared. Method of operation of the automatic sentence generation device for generation. The method of claim 6,
The step of generating the plurality of new sentences
When the plurality of candidate entity name words are selected, the sentences including one or more of the plurality of candidate entity name words are extracted from the sentence database as the plurality of candidate sentences, and then each of the plurality of candidate sentences is recorded. Generating a sentence table; And
Generating the plurality of new sentences including the new object name word by replacing the plurality of candidate object name words included in the plurality of candidate sentences in the candidate sentence table with the new object name word.
A method of operating a sentence automatic generation device for generating a corpus associated with a new entity name word comprising a. The method of claim 6,
When the plurality of new sentences are generated, recording the new entity name word and the first category in the word table corresponding to each other, and storing the plurality of new sentences in the sentence database
A method of automatically generating a sentence for generating a corpus associated with a new entity name word further comprising a. A computer-readable recording medium recording a computer program for executing the method of any one of claims 6 to 10 through coupling with a computer. A computer program stored in a storage medium for executing the method of any one of claims 6 to 10 through combination with a computer.

Images