KR20200072005A - Method for correcting speech recognized sentence - Google Patents

Abstract

Disclosed is a method for correcting a speech recognized sentence. Operations of a device for correcting a speech recognized sentence comprise the steps of: obtaining a speech recognized sentence; obtaining a profile of a user; selecting one or more of a plurality of corpuses stored in a database based on the profile of the user; correcting the sentence based on the one or more corpuses; and providing the corrected sentence.

Description

How to correct speech-recognized sentences{METHOD FOR CORRECTING SPEECH RECOGNIZED SENTENCE}

The following embodiments relate to a method of correcting a speech recognized sentence, for example, a technique related to speech recognition.

As computer computing power has developed and network bandwidth has expanded, data analysis near real-time is possible. As artificial intelligence (AI) technologies such as real-time data analysis and deep learning are combined, voice recognition technology is rapidly developing, especially in the mobile market.

The general speech recognition may be performed by extracting the characteristics of the input speech and converting the recognition result into text through a pre-built speech recognition model (for example, a probability model or a neural network model). Since each person has a different voice, pronunciation, and intonation, data related to the voices of multiple users can be collected, and a reference model can be generated from the collected data.

A method of correcting a speech recognized sentence according to an embodiment includes obtaining a speech recognized sentence; Obtaining a user's profile; Selecting one or more corpuses among a plurality of corpuses stored in a database based on the user's profile; Correcting the sentence based on the one or more corpus; And providing the corrected sentence.

According to one embodiment, the step of correcting the sentence may include selecting one or more words that need correction from among words included in the sentence, based on a probability distribution associated with one or more words included in the sentence; Determining one or more candidate words corresponding to the selected one or more words based on the selected one or more corpus; And substituting the selected one or more words with the determined one or more candidate words.

According to an embodiment, the method of correcting a speech recognized sentence further includes storing a result of correcting the sentence in the database, and the result of correcting the sentence includes the selected one or more words and the selected one. The candidate word may include at least one of the candidate word, the number of substitutions, and the user's profile.

According to an embodiment, the step of selecting the one or more corpuses may include: scoring a profile of the user based on a preset weight in relation to a plurality of detailed categories included in the plurality of categories; And selecting the one or more corpuses based on the scored profile, and the plurality of categories may include at least one of gender, age, region, utterance history, and service use area.

According to an embodiment, the plurality of corpus may correspond to the plurality of detailed categories included in the plurality of categories. Each of the plurality of corpuses includes a plurality of words related to a corresponding detailed category; And a plurality of candidate words for substituting the plurality of words.

According to an embodiment, the step of acquiring the sentence may include receiving a speech of the user; And converting the received speech into text, and receiving the speech of the user comprises receiving the user's voice in real time; And receiving a file in which the user's voice is recorded.

According to an embodiment, the step of acquiring the sentence may include determining a probability distribution related to one or more words included in the sentence, based on the word sequence included in the sentence.

According to one embodiment, the step of obtaining the user's profile includes fetching the profile from the database, the profile includes information related to one or more categories, and the one or more categories are the gender of the user, Age, region, history of utterances, and areas of service use.

The apparatus for correcting a speech recognized sentence according to an embodiment acquires a speech recognized sentence, acquires a user's profile, and based on the user's profile, one or more of a plurality of corpuses stored in the database And a processor for selecting a corpus, correcting the sentence based on the one or more corpus, and providing the corrected sentence.

1 is an operation flowchart illustrating a method of correcting a speech recognized sentence according to an embodiment.
2 is a view for explaining a method of obtaining a user's profile and scoring a user's profile according to an embodiment.
3 is a diagram for explaining a method of generating a corpus and storing it in a database according to an embodiment.
4 is an exemplary view of a result of selecting one or more corpus according to an embodiment.
5 is a view for explaining a method of correcting a sentence according to an embodiment.
6 is an exemplary view of a result of storing a result of correcting a sentence in a database according to an embodiment.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only, and may be implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosure form, and the scope of the present specification includes modifications, equivalents, or substitutes included in the technical spirit.

Terms such as first or second may be used to describe various components, but these terms should be interpreted only for the purpose of distinguishing one component from other components. For example, 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.

When an element is said to be "connected" to another element, it should be understood that other elements may be present, either directly connected to or connected to the other element.

Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as “include” or “have” are intended to designate the existence of a described feature, number, step, operation, component, part, or combination thereof, one or more other features or numbers, It should be understood that the existence or addition possibilities of steps, actions, components, parts or combinations thereof are not excluded in advance.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person skilled in the art. Terms such as those defined in a commonly used dictionary should be interpreted as having meanings consistent with meanings in the context of related technologies, and should not be interpreted as ideal or excessively formal meanings unless explicitly defined herein. Does not.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. The same reference numerals in each drawing denote the same members.

1 is an operation flowchart illustrating a method of correcting a speech recognized sentence according to an embodiment. According to an embodiment, all or part of a method of correcting a speech recognized sentence may be performed by a server. According to an embodiment, all or part of a method of correcting a speech recognized sentence may be performed by a terminal. Hereinafter, a server and/or a terminal that performs a method of correcting a speech-recognized sentence will be referred to as a'device for correcting a speech-recognized sentence'.

Referring to FIG. 1, an apparatus for correcting a speech recognized sentence acquires a speech recognized sentence (110). Speech-recognized sentences can be obtained by receiving a user's speech and converting the received speech into text . According to an embodiment, the apparatus for correcting a speech recognized sentence may receive a user's speech by receiving the user's voice in real time. For example, the user may utter a speech recognition unit associated with a device for correcting a speech-recognized sentence at a distance capable of clearly recognizing speech, and the device for correcting the speech-recognized sentence may recognize a user's speech and text Can be converted to According to an embodiment, the apparatus for correcting a speech-recognized sentence may receive a user's speech by receiving a file in which the user's voice is recorded. For example, a device for correcting a voice-recognized sentence receives a file containing a voice recorded by a user using a voice recorder or a terminal capable of performing a recording function, and recognizes the user's voice included in the file to recognize text Can be converted to Alternatively, the apparatus for correcting the speech-recognized sentence may receive a speech-completed sentence from an external device.

The step of acquiring a speech-recognized sentence may include determining a probability distribution related to one or more words included in the sentence based on the word sequence included in the sentence. The word sequence may include a plurality of consecutive words in a sentence. According to an embodiment, the apparatus for correcting a speech recognized sentence may determine a probability distribution related to one or more words included in the sentence using a language model. The probability of a word may be an index for linguistically evaluating a corresponding word in a sentence. For example, the language model may be a model having a context-dependent character. In this case, it may be determined whether an individual word included in the sentence is an appropriate word in relation to other words included in the sentence, and the determined result may be represented as a probability distribution. For example, in relation to the sentence "I went to Honkono with horseback," the probability distribution related to the words included in the sentence would be determined as'horse: 20%, stock: 80%, Honkono: 10%, went: 80%'. Can be. The word'horse' and the word'honkono' may be judged to be not appropriate words in relation to other words.

The apparatus for correcting the speech-recognized sentence acquires the user's profile (120 ). The apparatus for correcting speech-recognized sentences can retrieve a user's profile from a database. According to an embodiment, the database may be included in a device for correcting speech recognized sentences. According to an embodiment, the database may be implemented as a separate device that is distinct from a device for correcting speech-recognized sentences.

The user's profile may include information related to one or more categories. According to an embodiment, the one or more categories may include at least one of a user's gender, age, region, utterance history, and service use area.

More details related to the user's profile will be described later with reference to FIG. 2.

The apparatus for correcting the speech-recognized sentence selects one or more corpuses among a plurality of corpuses stored in the database based on the user's profile (130 ). The plurality of corpuses may correspond to a plurality of detailed categories included in the plurality of categories. Each of the corpus may include a plurality of words related to a corresponding subcategory and candidate words for substituting the corresponding words.

According to an embodiment, a plurality of corpuses may be stored in a database based on a plurality of categories and a plurality of detailed categories included in the plurality of categories. According to an embodiment, the apparatus for correcting a speech-recognized sentence may score a user's profile based on a preset weight in relation to a plurality of detailed categories. Based on the scored profile, the apparatus for correcting speech recognized sentences may select one or more corpus.

More details related to scoring the user's profile will be described later with reference to FIG. 2. More details related to selecting one or more corpus will be described later with reference to FIGS. 3 and 4.

The apparatus for correcting a speech-recognized sentence corrects the sentence based on one or more corpuses (140 ). The apparatus for correcting the speech-recognized sentence may select one or more words that need correction from among words included in the obtained sentence, based on a probability distribution related to one or more words included in the obtained sentence determined in step 110. According to an embodiment, it may be determined whether an individual word needs correction according to whether an individual probability value included in a probability distribution related to one or more words included in the obtained sentence is less than a predetermined value.

The apparatus for correcting the speech recognized sentence may determine one or more candidate words for substituting the selected one or more words based on the selected one or more corpus. The candidate word may be a candidate for replacing the selected word. The apparatus for correcting the speech-recognized sentence may replace the selected one or more words with one or more candidate words corresponding to the selected one or more words.

The apparatus for correcting the speech-recognized sentence may store a result of correcting the sentence in a database. According to an embodiment, the result of correcting a sentence may include at least one of the selected one or more words, the selected one or more candidate words, the number of substitutions, and the user's profile.

More details related to correcting the sentence will be described later with reference to FIGS. 5 and 6.

The apparatus for correcting the speech-recognized sentence provides the corrected sentence (150). According to an embodiment, the apparatus for correcting a sentence may output the corrected sentence. The corrected sentence can be output explicitly or implicitly. For example, the corrected sentence may be explicitly output in the form of a text message or the like and provided to the user. Alternatively, the corrected sentence may be used for implicit output, such as being used to improve service of a customer center or to improve NLU (Natural Language Understanding) technology instead of being output explicitly.

2 is a view for explaining a method of obtaining a user's profile and scoring a user's profile according to an embodiment.

Referring to FIG. 2, an apparatus for correcting a speech recognized sentence may retrieve a user's profile from a database. The user's profile may include information related to one or more categories. According to an embodiment, one or more categories included in the user's profile may include at least one of gender, age, region, utterance history, and service use region. For example, the user's profile may include information such as'ID: abcd1234','gender: male','age: 10','region: Seoul','service use area: game', etc. (210 ).

According to an embodiment, a plurality of corpuses may be stored in a database based on a plurality of categories and a plurality of detailed categories included in the plurality of categories. According to an embodiment, the apparatus for correcting a speech-recognized sentence may score a user's profile based on a preset weight in relation to a plurality of detailed categories included in a plurality of categories. A plurality of categories according to an embodiment may include gender, age, region, utterance history, and service use area. According to an embodiment, the more prominent a category can be to a user's utterance, the greater the weight associated with the subcategory. For example, considering that the terminology used by people of teenagers and the dialect of Jeju Island used by people residing in Jeju Island contain many distinct features from the standard language, the detailed category included in category'Age' A larger weight may be set in relation to the detailed category'Jeju-do' included by the'teens' and the category'regions'.

Hereinafter, embodiments related to preset weights with respect to various categories and detailed categories are presented. According to an embodiment, the category'gender' may include a detailed category'male' and a detailed category'female'. According to an embodiment, the subcategory'male' and the subcategory'female' may give characteristics to the user's speech. However, since the feature is not prominent, the weight can be set to 2 in relation to the subcategory'male', and the weight can be set to 2 in the subcategory'female'.

According to an embodiment, the category'age' may include a detailed category '10', a detailed category '20', and a detailed category '30'. According to an embodiment, the subcategory '10s' may give a more prominent characteristic to the utterance than the subcategory '20s', and the subcategory '20s' to the utterance compared to the subcategory '30s' You can add more prominent features. According to one embodiment, the weight may be set to 7 in relation to the detailed category '10', the weight may be set to 5 in relation to the detailed category '20', and the weight in relation to the detailed category '30' Can be set to 3.

According to an embodiment, the category'region' may include a detailed category'Seoul', a detailed category'Busan', and a detailed category'Game'. According to an embodiment, the sub-category'Busan' may give a more prominent characteristic to the utterance than the sub-category'Gyeonggi', and the sub-category'Match' is more prominent to the utterance than the sub-category'Seoul' Can be given. According to an embodiment, the weight may be set to 1 in relation to the detailed category'Seoul', the weight may be set to 7 in relation to the detailed category'Busan', and the weight is set to 2 in relation to the detailed category'Game' Can be set.

According to an embodiment, in the example of the profile of the user presented above,'ID: abcd1234','gender: male','age: 10','region: Seoul','service use area: game','gender : The pre-set weight with respect to'M' may be 2, the pre-set weight with respect to'Age: 10' may be 7, and the pre-set weight with respect to'Region: Seoul' may be 1, The preset weight in relation to the'service use area: game' may be 3 (220). In this case, the user's profile will be scored as'ID: abcd1234','Gender: Male/2','Age: Teenagers/7','Area: Seoul/1','Service Usage Area: Game/3'. Can be.

Based on the scored profile, the apparatus for correcting speech recognized sentences may select one or more corpus. According to an embodiment, the apparatus for correcting the speech-recognized sentence may select one corpus corresponding to the detailed category having the highest score. For example, the example of the scored user profile presented above is'ID: abcd1234','Gender: Male/2','Age: Teenagers/7','Region: Seoul/1','Service use area: Game With respect to /3', one corpus corresponding to the detailed category'age: teenager' with the highest score may be selected (230). In this case, the selected corpus is a subcategory related to the sub-category'Age: Teenager', and includes a plurality of words, such as ‘Aeopsi’ and ‘Death’, and ‘Adagi except corpse’, ‘A wound of the heart’, etc. It may include a plurality of candidate words.

More details related to selecting one or more corpus will be described later with reference to FIGS. 3 and 4.

3 is a diagram for explaining a method of generating a corpus and storing it in a database according to an embodiment.

Referring to FIG. 3, a plurality of corpuses according to an embodiment may be generated based on a plurality of categories and a plurality of detailed categories included in the plurality of categories. According to one embodiment, various characteristics associated with the user may characterize the user's speech. According to an embodiment, a plurality of categories may be determined to include various characteristics related to a user who can characterize a user's speech. For example, the plurality of categories may include characteristics such as gender, age, region, utterance history, and service use area (310).

According to an embodiment, a plurality of detailed categories may be determined in relation to a plurality of predetermined categories (320). For example, detailed categories such as'teens', '20s', and '30s' may be determined in relation to the category'age'. According to an embodiment, corpuses may be stored in response to a plurality of detailed categories (330 ). For example, different corpuses may be stored in relation to the subcategory '10', subcategory '20', and subcategory '30'.

4 is an exemplary view of a result of selecting one or more corpus according to an embodiment.

Referring to FIG. 4, based on the scored profile, the apparatus for correcting speech recognized sentences may select one or more corpuses. According to an embodiment, the apparatus for correcting the speech-recognized sentence may select one corpus corresponding to the detailed category having the highest score. For example, if the user's profile is scored as'first subcategory/7' 410,'second subcategory/3' 420, and'third subcategory/5' 430, voice The apparatus for correcting the recognized sentence may select a corpus corresponding to the first detailed category, which is the detailed category with the highest score (a).

According to an embodiment, the apparatus for correcting a speech recognized sentence may refer to all corpuses corresponding to detailed categories of the profile. According to an embodiment, based on all corpuses corresponding to detailed categories of the profile, the apparatus for correcting a speech-recognized sentence may include one or more candidates corresponding to one or more words that need correction among words included in the obtained sentence You can decide the word. According to an embodiment, when the determined candidate word is singular, the apparatus for correcting the speech-recognized sentence may select one corpus corresponding to the detailed category having the highest score among corpuses including the candidate word. According to an embodiment, when there are a plurality of determined candidate words, the apparatus for correcting the speech-recognized sentence may sum the scores of one or more corpuses including individual candidate words. According to an embodiment, the apparatus for correcting a speech-recognized sentence compares a result of summing scores for each candidate word, selects one candidate word having the highest sum, and one or more corpuses including the candidate word You can choose For example, a user's profile may have been scored as'first subcategory/7' 410,'second subcategory/3' 420, and'third subcategory/5' 430, Among words included in the obtained sentence, a word requiring correction may be A. The corpus corresponding to the first sub-category may include candidate word B corresponding to the word A, and the corpus corresponding to the second sub-category and the corpus corresponding to the third sub-category may select the candidate word C corresponding to the word A. It can contain. In this case, the score corresponding to the first detailed category may be 7 points, and the result of summing the scores corresponding to the second detailed category and the third detailed category may be 8 points. Accordingly, the apparatus for correcting the speech-recognized sentence may select the candidate word as C, and may select a corpus corresponding to the second sub-category and a corpus corresponding to the third sub-category (b).

5 is a view for explaining a method of correcting a sentence according to an embodiment. 5 may correspond to step 140 of FIG. 1.

Referring to FIG. 5, the apparatus for correcting a voice-recognized sentence may select one or more words that need correction from among words included in the obtained sentence, based on a probability distribution related to one or more words included in the obtained sentence have. The probability distribution associated with one or more words may be determined at step 110 of FIG. 1. According to an embodiment, it may be determined whether an individual word needs correction according to whether an individual probability value included in a probability distribution related to one or more words included in the obtained sentence is less than a predetermined value. For example, the predetermined value for determining whether correction is required is 30%, and the probability distribution in relation to the obtained sentence "I went to Honkono wearing horseback" is'Wearing: 20%, wearing: 80%, Honkono: 10 %, went: 80%, the device that corrects the speech-recognized sentence selects the words'horse' and the word'Honkono' whose individual probability values are less than 30% included in the probability distribution as words that need correction. It can be (510).

The apparatus for correcting the speech recognized sentence may determine one or more candidate words corresponding to the selected one or more words based on the selected one or more corpus. For example, when the selected corpus is a corpus corresponding to'age, teenagers', the apparatus for correcting the speech-recognized sentence is a candidate word corresponding to the word'horse' based on the selected corpus as the'heart wound' It is possible to determine, and the candidate word corresponding to the word'Honkono' may be determined as'single coin karaoke' (520).

The apparatus for correcting the speech-recognized sentence may replace the selected one or more words with the determined one or more candidate words. For example, the apparatus for correcting a voice-recognized sentence is a word'horse' of the obtained sentence "I went to Honkono with a horseman" as a candidate word "wound of the heart", and the word "Honkono" as a candidate word "alone coin karaoke" '. That is, the obtained sentence may be replaced with "Wounded heart and went to coin karaoke alone" (530).

6 is an exemplary view of a result of storing a result of correcting a sentence in a database according to an embodiment.

The apparatus for correcting the speech-recognized sentence may store a result of correcting the sentence in a database. According to an embodiment, the result of correcting a sentence may include at least one of the selected one or more words, the selected one or more candidate words, the number of substitutions, and the user's profile. According to an embodiment, the apparatus for correcting a speech-recognized sentence may store the selected one or more words and the selected one or more candidate words in a corresponding database format. For example, in the obtained sentence "I went to Honkono with a horseman", the word'horseman' was replaced with the candidate word "Wound of the Mind", and the word "Honkono" was replaced with the candidate word "Single Coin Karaoke". The form in which the above word and the selected one or more candidate words are matched may be the same as (horseback, wound of the heart), (Honkono, karaoke alone coin), and the number of substitutions may be two times. In this case, the form in which the result of correcting the sentence is stored in the database may be the same as {(horse, wound of heart), (Honkono, karaoke alone coin), 2, male, teenage, Seoul} (610).

According to one embodiment, all or part of the above-described methods through FIGS. 1 to 6 may be performed by a server. According to an embodiment, all or part of the above-described methods through FIGS. 1 to 6 may be performed by a terminal. According to an embodiment, when some of the above-described methods are performed by the server through FIGS. 1 to 6 and the other part is performed by the terminal, the server and the terminal mutually communicate information necessary to perform each step. Can be exchanged for.

The embodiments described above may be implemented by hardware components, software components, and/or combinations of hardware components and software components. For example, the devices, methods, and components described in the embodiments include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors (micro signal processors), microcomputers, and field programmable gates (FPGAs). It can be implemented using one or more general purpose computers or special purpose computers, such as arrays, programmable logic units (PLUs), microprocessors, or any other device capable of executing and responding to instructions. The processing device may run an operating system (OS) and one or more software applications running on the operating system. In addition, the processing device may access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of understanding, a processing device may be described as one being used, but a person having ordinary skill in the art, the processing device may include a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that may include. For example, the processing device may include a plurality of processors or a processor and a controller. In addition, other processing configurations, such as parallel processors, are possible.

The software may include a computer program, code, instruction, or a combination of one or more of these, and configure the processing device to operate as desired, or process independently or collectively You can command the device. Software and/or data may be interpreted by a processing device, or to provide instructions or data to a processing device, of any type of machine, component, physical device, virtual equipment, computer storage medium or device. , Or may be permanently or temporarily embodied in the transmitted signal wave. The software may be distributed over networked computer systems, and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on 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 in the medium may be specially designed and configured for the embodiments 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. -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 codes that can be executed by a computer using an interpreter, etc., as well as machine language codes produced by a compiler. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited drawings, those skilled in the art may apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order than the described method, and/or the components of the described system, structure, device, circuit, etc. are combined or combined in a different form from the described method, or other components Alternatively, even if replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (17)

Obtaining a speech-recognized sentence;
Obtaining a user's profile;
Selecting one or more corpuses among a plurality of corpuses stored in a database based on the user's profile;
Correcting the sentence based on the one or more corpus; And
Providing the corrected sentence
A sentence correction method comprising a.
According to claim 1,
The step of correcting the sentence is
Selecting one or more words to be corrected among words included in the sentence based on a probability distribution associated with one or more words included in the sentence;
Determining one or more candidate words corresponding to the selected one or more words based on the selected one or more corpus; And
Substituting the selected one or more words with the determined one or more candidate words
A sentence correction method comprising a.
According to claim 2,
Storing the result of correcting the sentence in the database
Further comprising,
The result of correcting the sentence
The selected one or more words, the selected one or more candidate words, the number of substitutions, and the user's profile
Sentence correction method comprising at least one of.
According to claim 1,
The step of selecting the one or more corpuses
Scoring the user's profile based on a preset weight in relation to a plurality of detailed categories included in the plurality of categories; And
Selecting the one or more corpus based on the scored profile
Including,
The plurality of categories
Gender, age, geography, utterance history, and area of service use
Sentence correction method comprising at least one of.
According to claim 1,
The plurality of corpus
Corresponding to a plurality of detailed categories included in the plurality of categories,
Each of the plurality of corpus
A plurality of words related to the corresponding subcategory; And
A plurality of candidate words for substituting the plurality of words
A sentence correction method comprising a.
According to claim 1,
The step of obtaining the sentence
Receiving the user's speech; And
Converting the received speech into text
Including,
The step of receiving the user's speech is
Receiving the user's voice in real time; And
Receiving a file in which the user's voice is recorded
Sentence correction method comprising at least one of.
According to claim 1,
The step of obtaining the sentence
Determining a probability distribution related to one or more words included in the sentence based on the word sequence included in the sentence
A sentence correction method comprising a.
According to claim 1,
The step of obtaining the user's profile is
Loading the profile from the database
Including,
The profile is
Contains information related to one or more categories,
The one or more categories
The user's gender, age, region, utterance history, and service usage area
Sentence correction method comprising at least one of.
A computer program stored in a medium to execute the method of any one of claims 1 to 8 in combination with hardware.
Acquire the speech recognition sentence,
Get the user's profile,
Based on the user's profile, one or more corpuses stored in a database are selected,
Correct the sentence based on the one or more corpus,
Processor providing the corrected sentence
Included, sentence correction device.
The method of claim 10,
The processor to correct the sentence
Based on the probability distribution associated with one or more words included in the sentence, one or more words needing correction are selected from the words included in the sentence,
Determine one or more candidate words corresponding to the selected one or more words based on the selected one or more corpus,
A processor that replaces the selected one or more words with the determined one or more candidate words
Included, sentence correction device.
The method of claim 11,
Processor that stores the result of correcting the sentence in the database
Further comprising,
The result of correcting the sentence
The selected one or more words, the selected one or more candidate words, the number of substitutions, and the user's profile
Sentence correction device comprising at least one of.
The method of claim 10,
The processor for selecting the one or more corpus
Scoring the user's profile based on a preset weight associated with a plurality of detailed categories included in the plurality of categories,
A processor that selects the one or more corpus based on the scored profile
Including,
The plurality of categories
Gender, age, geography, utterance history, and area of service use
Sentence correction device comprising at least one of.
The method of claim 10,
The plurality of corpus
Corresponding to a plurality of detailed categories included in the plurality of categories,
Each of the plurality of corpus
A plurality of words related to the corresponding subcategory; And
A plurality of candidate words for substituting the plurality of words
Included, sentence correction device.
The method of claim 10,
The processor that acquires the sentence
Receiving the user's speech,
Processor for converting the received speech into text
Including,
The processor that receives the user's speech is
A processor that receives the user's voice in real time; And
Processor for receiving a file in which the user's voice is recorded
Sentence correction device comprising at least one of.
The method of claim 10,
The processor that acquires the sentence
A processor that determines a probability distribution related to one or more words included in the sentence based on the word sequence included in the sentence
Included, sentence correction device.
The method of claim 10,
The processor to obtain the user's profile
Processor to retrieve the profile from the database
Including,
The profile is
Contains information related to one or more categories,
The one or more categories
The user's gender, age, region, utterance history, and service usage area
Sentence correction device comprising at least one of.

Images