[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

KR20140123369A - Question answering system using speech recognition and its application method thereof - Google Patents

Question answering system using speech recognition and its application method thereof Download PDF

Info

Publication number
KR20140123369A
KR20140123369A KR1020130040660A KR20130040660A KR20140123369A KR 20140123369 A KR20140123369 A KR 20140123369A KR 1020130040660 A KR1020130040660 A KR 1020130040660A KR 20130040660 A KR20130040660 A KR 20130040660A KR 20140123369 A KR20140123369 A KR 20140123369A
Authority
KR
South Korea
Prior art keywords
voice
answer
question
sentence
text
Prior art date
Application number
KR1020130040660A
Other languages
Korean (ko)
Inventor
윤재민
Original Assignee
얄리주식회사
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 얄리주식회사 filed Critical 얄리주식회사
Priority to KR1020130040660A priority Critical patent/KR20140123369A/en
Publication of KR20140123369A publication Critical patent/KR20140123369A/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
    • G06F16/63—Querying
    • G06F16/638—Presentation of query results
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/08—Speech classification or search
    • G10L15/18—Speech classification or search using natural language modelling
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/26—Speech to text systems
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
    • G10L2015/221—Announcement of recognition results

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

In particular, the present invention relates to a voice recognition query response system and method, and more particularly, to a voice recognition system that recognizes a voice of a question and an answer from a voice of a user and converts the voice into a question and answer sentence. And a method of operating the same.
To this end, the present invention recognizes a voice of a question and an answer from a voice of a user, converts the voice into a question and an answer sentence, stores a text file of the question and answer, indexes and stores the question and answer sentence, And a terminal for outputting a response to the sentence inputted by the question and answer by voice and text when the user inputs a question by voice, And provides a recognition query response system.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a speech recognition system and a speech recognition system,

The present invention relates to a voice recognition query response system and a method thereof, and more particularly, to a speech recognition system which recognizes a voice of a question and an answer from a voice of a user and converts the voice into a question and answer sentence, To a voice recognition query response system for performing a posterior query response and a method of operating the same.

The Q & A system queries the system to obtain the knowledge desired by the user, and the system analyzes the Q & A system and outputs the related answers. So far, the Q & A system has been implemented in various ways. However, existing systems have limitations in that questions and answers are stored and expressed in text form.

The present invention has been proposed in order to solve the problems of the related art as described above, and it is an object of the present invention to provide a system and method for storing a question and an answer sentence by voice, and a system and a method for voice conversation.

According to an aspect of the present invention, there is provided a voice recognition question answering system for recognizing a voice of a question and an answer from a voice of a user and converting the voice into a question and answer sentence, storing a text file of the question and answer, When a user inputs a question by voice, the speech and the sentence are converted into text, and a question and answer is performed, and the answer to the sentence inputted by the question and answer is outputted as voice and text The terminal is configured to be a terminal.

Meanwhile, the speech recognition question answering system of the present invention recognizes a voice of a question and an answer from a voice of a user, converts the voice into a question and an answer sentence, stores a text file of the question and an answer, Indexes and stores the texts. When a user inputs a question by voice, the speech is recognized and converted into text, a query response is performed, and a response to the sentence input by the query response is output as voice and text .

Here, the voice file for the question and answer is stored, and the question and answer voice file is indexed and stored.

A voice input device for inputting voice; A voice input unit for converting the analog voice transmitted through the voice input device into a digital signal; A voice recognition unit for performing voice recognition from the voice information received by the voice input unit; A natural language processing unit for performing indexing and querying based on information converted from speech to text by the speech recognition unit; A screen output unit for outputting a reply sent from the natural language processing unit as text; A voice output unit for converting the voice into a digital signal to an analog signal; And a voice output device for outputting the voice.

The speech recognition unit recognizes speech by a speech recognition algorithm and converts the speech into text, and stores the text as a text file.

The natural language processing unit performs an indexing process on the basis of the question-and-answer sentence information converted from the speech to the text by the speech recognition unit, analyzes the morpheme based on the question and answer sentence information, And a query response process is performed.

The screen output unit may output a response sentence sent from the natural language processing unit as text on a screen.

The voice output unit may output a voice file corresponding to a response sentence sent from the natural language processing unit to a speaker or an earphone.

In addition, a voice file for the question and answer is stored, and the voice file for question and answer is indexed and stored.

Further, the present invention is characterized by further comprising a text portion for converting the answer sentence into speech.

Meanwhile, in the method of operating the voice recognition question answering system of the present invention, a method of storing a question and an answer sentence by voice includes a step 1) of inputting a question and an answer as a voice; A second step of recognizing speech from the speech; And indexing the speech-recognized speech and the text generated after the speech recognition.

Here, the method may further include a step 2a of storing the voice recognized as a voice file.

The voice file corresponding to the question sentence and the answer sentence is stored in association with the question sentence and the answer sentence, respectively.

At this time, in the step 1 of inputting the question and the answer by voice, a question input button is provided to the user to check whether the voice input button is activated, and when the voice is inputted, the completion of the question input is displayed, A button is provided to check whether a voice input button is activated, and if the voice is inputted, the completion of answer input is displayed, and the inputted question and answer are respectively transmitted to the voice recognition step.

In the second step of voice recognition from the voice, the question input voice and the answer input voice are received, respectively, and the voice is converted into text and displayed to the user as a question sentence and a reply sentence.

The third step of indexing the speech recognized speech and the text generated after the speech recognition includes extracting a keyword list displayed in the question sentence and an answer sentence, And stores it in the indexing DB.

Alternatively, the speech recognition apparatus may further comprise a step 2b of storing the sentence in which the speech is recognized as the question and answer sentence.

According to another aspect of the present invention, there is provided a method of operating a voice recognition system, the method comprising the steps of: receiving a question voice; A second step of recognizing speech from the speech; A third step of analyzing a sentence with text generated after the speech recognition; And a fourth step of performing a query response after analyzing the sentence; And a fifth step of outputting the answers extracted from the query response DB or generated through the query response DB as voice and text after the query response.

According to another aspect of the present invention, there is provided a method of operating a voice recognition system, the method comprising the steps of: Two steps of speech recognition; A third step of performing a query response processing on the sentence information generated after the speech recognition; And outputting the answers extracted or generated by the query response as answer voices and answer texts.

The voice recognition query response system and its operation method constructed as described above have a useful effect of storing a question and answering sentence by voice or conversing with a voice.

1 is a block diagram of a voice recognition query response system according to an embodiment of the present invention;
FIG. 2 illustrates a procedure for storing questions and answers from a voice in a voice recognition query response system according to an embodiment of the present invention; FIG.
3 is a diagram illustrating an operation method procedure for storing a question and an answer from a voice of a voice recognition question and answer system according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating a procedure for a voice-response-based conversation of a voice-recognition question-and-answer system according to an embodiment of the present invention; FIG.
5 is a diagram illustrating an operation method procedure of a voice-response-based conversation in a voice-recognition question-and-answer system according to an embodiment of the present invention;
FIG. 6 is a diagram illustrating speech input and speech recognition results of a speech recognition query response system according to an embodiment of the present invention; FIG.
7 is a diagram illustrating an internal configuration of a voice recognition question answering system according to an embodiment of the present invention;
FIG. 8 is a flowchart illustrating a method for storing a question and an answer from a voice in a voice recognition query response system according to an embodiment of the present invention; FIG.
9 is a diagram illustrating a method for storing questions and answers from a voice in a voice recognition query response system according to an embodiment of the present invention;
FIG. 10 is a flowchart illustrating a method for voice-based query-response conversation in a speech recognition query response system according to an exemplary embodiment of the present invention; FIG.
FIG. 11 is a diagram illustrating a method for voice-based query-response conversation in a voice-recognition query response system according to an embodiment of the present invention; FIG.
12 is a screen for voice conversation in a voice recognition question answering system according to an embodiment of the present invention;
13 is a screen for voice conversation in a voice recognition question answering system according to an embodiment of the present invention;
14 and 15 are a screen for displaying a question and an answer sentence after inputting a question and answer voice in a voice recognition question answering system according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily carry out the present invention.

The present invention can be embodied in various different forms, and thus the present invention is not limited to the embodiments described herein.

1 is a block diagram of a voice recognition query response system according to an embodiment of the present invention.

As shown in FIG. 1, the present invention may include a terminal 200, and may include a voice input device 100 and a voice output device 300.

The terminal 200 recognizes a voice of a question and an answer from a voice of a user and converts the voice into a question and an answer sentence to store a voice file for the question and answer and a text file for the question and answer, An answer sentence, and the question and answer voice file are indexed and stored.

Here, the terminal 200 may store a voice file for the question and answer, and index and store the voice file of the question and answer.

Then, when the user inputs a question by voice, the terminal 200 converts it into text after speech recognition, performs a query response after text processing, and outputs the answer to the sentence inputted by the query response as voice and text do.

The voice input apparatus 100 inputs voice and the voice output apparatus 300 outputs voice to be applied to the terminal.

Specifically, the terminal 200 includes an audio input unit 210, a voice recognition unit 220, a natural language processing unit 230, a screen output unit 250, and a sound output unit 260.

The voice input unit 210 decodes the analog voice into a digital signal and converts the analog signal transmitted from the external microphone or the internal microphone of the terminal into a digital signal.

The speech recognition unit 220 performs speech recognition from the speech information received from the speech input unit 210, recognizes speech by a speech recognition algorithm, converts the speech into text, and stores the text as a text file.

Here, the speech recognition unit 220 may recognize speech by a speech recognition algorithm and convert the speech into text, and store the speech as a speech file.

The natural language processing unit 230 performs indexing, sentence analysis (morpheme analysis, syntax analysis, semantic analysis), and query response based on the speech-to-text converted information and the speech voice file by the speech recognition unit 220 .

That is, the natural language processing unit 230 performs an indexing process on the basis of the question-and-answer sentence information converted from speech to text by the speech recognition unit 220, performs an indexing process, (Sentence analysis, semantic analysis, statistical analysis), query response (response sentence extraction algorithm (similarity search, pattern search), or response sentence generation algorithm).

The speech output unit 260 outputs a speech file corresponding to the response sentence sent from the natural language processing unit 230 to the speech output unit 250. [ Or the earphone.

In addition, the terminal 200 provides a question input unit and an answer input unit, provides the user with the question input unit to receive a question as a voice, provides the answer input unit and receives a response as a voice, And stores the question and sentence where the specific word (keyword) is generated and the position information (sentence number) of the answer sentence by indexing the question sentence and the answer sentence.

Here, the terminal 200 may store the question sentence and the voice file path information of the answer sentence in the DB.

In addition, upon receiving the voice of the user, the terminal 200 converts the voice into text, recognizes the voice, performs a query response after analyzing the sentence (word extraction in sentence by morphological analysis) And retrieves the answer to the sentence from the indexing DB and the question and answer DB, and outputs it as voice and text.

Here, the terminal 200 may extract the answer to the sentence input by the query response from the voice DB and output it as a voice.

2 is a diagram illustrating a procedure for storing a question and an answer from a voice in a voice recognition question answering system according to an embodiment of the present invention.

A method for storing a question and an answer sentence by voice in a method of operating a voice recognition question answering system according to the present invention includes a first step S100a of inputting a question and an answer by voice, a second step S200a of recognizing a voice from the voice, (S300a) of storing the speech-recognized speech as a speech file (S300a); and 4th step S400a of indexing the speech-recognized speech and the text generated after the speech recognition.

Specifically, in the first step S100a of inputting the question and the answer by voice, a question input button is provided to the user to check whether or not the voice input button is activated. When all the voice is input, A question input voice is stored in a memory and an answer input button is provided to the user to check whether a voice input button is activated so as to indicate completion of answer input when voice is inputted and store answer input voice in a memory, To the speech recognition step.

In the second step S200a of speech recognition from the speech, the question input speech and the answer input speech are received, and the speech is converted into text and displayed to the user as a question sentence and a reply sentence.

Next, in step S300a of storing the voice recognized as a voice file, voice files corresponding to the question sentence and the answer sentence are stored in association with the question sentence and the answer sentence, respectively.

Finally, in step S400a of indexing the speech-recognized speech and the text generated after the speech recognition, a word (keyword) list shown in the question sentence and the answer sentence is extracted, (Sentence number) of a sentence, voice file path information of a question sentence and an answer sentence into a word list, and stores it in the indexing DB 122. [

3 is a diagram illustrating an operation method procedure for storing a question and an answer from a voice in a voice recognition question answering system according to an embodiment of the present invention.

First, when a voice of a user is input, voice recognition is performed by checking whether a question and an answer are inputted, voice is stored in a voice DB 240 in a voice file, and text is divided into a question and answer DB 243). At this time, the question and answer sentence included in the question and answer DB 243 stores the voice file path information of the sentence in the voice query information DB 241.

Thereafter, the conventional indexing process, which is frequently used in the information search field, is performed based on the query response information DB 243 and the voice query information DB 241, and is stored in the indexing DB 242.

The indexing DB 242 stores word and keyword list information extracted from a question and answer sentence in the question and answer DB 243, position information (a sentence number) of a question and an answer sentence including the corresponding word, Includes voice file path information for sentences.

4 is a diagram illustrating a procedure for a voice-response-based conversation in a voice recognition query response system according to an embodiment of the present invention.

A method for performing a query-response dialogue by voice in a method for operating a speech recognition query response system of the present invention includes a first step (S100b) of receiving a question voice, a second step (S200b) of recognizing speech from the speech, (S300b), a fourth step (S400b) of querying and responding after the sentence analysis; And a fifth step (S500b) of outputting the answers extracted from the question and answer DB 243 or the answers generated through the question and answer DB 243 as voice and text after the question and answer.

In step 1 (S100b) of receiving the question voice, a separate voice input device 100 is used to attach a question voice of a user in real time using a voice input device (microphone) 100 attached or built in the outside of the terminal 200 Receive input.

The second step S200b of recognizing the voice from the voice receives the voice of the user, performs voice recognition, and converts the voice of the user into text.

Next, in a third step S300b of analyzing a sentence with text generated after the speech recognition, a preprocessing process is performed to extract a word from a sentence by morphological analysis of the text to perform a query response.

After analyzing the sentence, the fourth step (S400b) of querying and responding is a process of analyzing a sentence (analysis of semantics, statistical analysis), response sentence extraction algorithm (similarity search, pattern search) .

Finally, in step 5 (S500b) of outputting the answers extracted from the query response DB 243 or generated through the query response DB 243 as voice and text after the query response, When a reply sentence is extracted, the existing answer sentence is outputted as a voice through a voice file and displayed as text. When a new answer sentence is generated through the question and answer DB 243, And displays the corresponding answer text as text.

Here, TTS (TTS) is a text-to-speech automatic conversion technology, short for Text to speech

FIG. 5 is a diagram illustrating an operation method procedure of a voice-response-based conversation in a voice-recognition question-and-answer system according to an embodiment of the present invention.

According to the present invention, when a voice of a user is input, a voice recognition process is performed from the voice, and a sentence analysis is performed based on the extracted text.

In the sentence analysis, basic keyword combinations can be extracted from the inputted question text through morphological analysis, so that preparation for basic natural language processing is completed. Then, the user's intention is grasped through a separate analysis and semantic analysis process.

Thereafter, it is possible to perform various procedures for the already-known query response using the keyword information, sentence information, and semantic information extracted through the sentence sentence, and to obtain the answer to the question from the question and answer DB 243 . The extracted answer is retrieved from the previously stored voice file path information and outputted as voice or displayed as text.

6 is a diagram illustrating speech input and speech recognition results of the speech recognition question answering system according to an embodiment of the present invention.

In order to store a question and an answer from a voice, a voice of a user is input by pressing a question voice input start button before voice input. After receiving the input, if the speech recognition is performed, the speech recognition sentence (for example, I love you) is displayed.

Also, the answer voice input start button is pressed to receive the voice of the user. After receiving the input, when the speech recognition is performed, the speech recognition sentence (for example, I love you) is displayed.

When the input completion button is pressed, the voice corresponding to the question and answer inputted from the voice is stored as the voice file, and the voice recognition result is stored as the text, respectively.

Meanwhile, FIG. 7 is a diagram showing an internal configuration of a voice recognition question and answer system (using a voice recognition system) according to an embodiment of the present invention.

7, the present invention includes a voice input device 1100, a voice input unit 1110, a voice recognition unit 1120, a natural language processing unit 1130, a text output unit 1140, a TTI unit 1161, An audio output unit 1160, and a sound output apparatus 1170. [

The voice input unit 1100 receives the voice, the voice input unit 1110 converts the analog voice transmitted through the voice input device to a digital signal, and the voice recognition unit 1120 receives the voice from the voice input unit 1110 And performs voice recognition from the voice information.

The natural language processing unit 1130 performs indexing or querying based on the information converted from speech to text by the speech recognizing unit 1120 and the text output unit 1140 outputs the answer transmitted from the natural language processing unit 1130 And outputs it to the monitor 1150 screen as text.

The TTI unit 1161 converts the answer sentence into a voice, the voice output unit 1160 converts the voice into a digital signal to an analog signal, and the voice output device 1170 outputs the voice to the earphone or speaker .

According to this configuration, the voice of the question and answer is recognized from the voice of the user, converted into the question and answer sentence, the text file for the question and answer is stored, and the question and answer sentence is indexed and stored.

Then, when the user inputs a question by voice, the speech is recognized and converted into text, a query response is performed, and a reply to the sentence input by the query response can be outputted as speech and text.

The voice recognition query response system (using the TTS) according to an embodiment of the present invention provides two-way voice and data communication such as a personal computer (PC), a notebook, a smart phone (iPhone, Android phone, Lt; RTI ID = 0.0 > media.

Specifically, the voice input unit 1110 provides a question input unit and an answer input unit. When the question input unit receives a voice as a voice from a user, the voice input unit 1110 displays a question voice after a voice recognition as a question sentence. When the answer is inputted by voice, the answer voice is displayed as a response sentence after speech recognition. When the voice input of the question and answer is completed and the user clicks the input completion button, the question sentence and the answer sentence are indexed, The query sentence in which the word (keyword) occurs and the location information (sentence number) of the answer sentence are stored in the DB.

In addition, the voice recognition question answering system (using T-TES) according to an embodiment of the present invention detects a voice of a user, displays a voice recognition result on a question input window, and displays a response sentence , Displays the answer sentence in the answer input window, and outputs the answer voice using the text message.

In addition, when outputting answers using Titles, users can choose from a variety of voice-overs by voice, age, and gender.

In addition, when the voice of the user is sensed and voice data sensed with a meaningful voice is recognized, if there is no voice recognition result, a message prompting the user to input voice again is displayed, thereby prompting the user to input the voice accurately.

Here, the voice input and output method converts a question voice, which is an analog signal transmitted to the voice input device 1100, which is an external microphone or a terminal internal microphone, from the voice input unit 1110 to a digital signal and transmits the voice to the voice output unit 1160 And outputs the converted answer voice to an analog signal through an audio output device 1170 including an earphone or a speaker. At this time, the text output unit 1140 displays the text information on the terminal screen.

The sentence text information, which is the result of speech recognition after speech recognition in the speech recognition unit 1120, is stored in the query response DB 1121, and the query response DB 1121 receives information composed of pairs of question and answer sentences And stores the result in the indexing DB 1122. [

Further, the speech recognition unit 1120 recognizes speech by a speech recognition algorithm and converts the speech into sentence text, and stores the sentence text as text information.

The natural language processing unit 1130 searches for an answer by a question and answer module 1132 that finds an answer to a specific question based on the question and answer sentence information converted from speech to text by the speech recognition unit 1120 Create an answer.

The question-and-answer module 1132 analyzes sentences from a question sentence to determine the exact intent of a question. A question requesting an accurate answer includes an answer from a pre-established answer DB. When requesting specific information, An answer is generated based on the information.

The inquiry response module 1132 may generate an answer by fetching the information through the wired / wireless communication network when the question sentence requests specific information such as time, news, and weather.

8 is a flow diagram illustrating a method for storing questions and answers from a voice in a voice recognition query response system (using Titles) in accordance with an embodiment of the present invention.

As shown in FIG. 8, a method of storing a question and an answer sentence by voice in a method of operating a voice recognition question answering system (using T-TES) according to the present invention includes a first step S100c Step S200c of recognizing the speech, step S300c of storing the sentence as a question and answer sentence, and step S400c of indexing the question and answer sentence.

Specifically, the user's voice about the question and the answer is inputted (S100c), the speech for the question and answer is recognized (S200c), converted into the question and answer sentence, the text file for the question and the answer is extracted, Stored in the response DB 1132 (S300c), indexes the question and answer sentence, and stores it in the indexing DB 1122 (S400c).

The indexing DB 1122 stores the morpheme information list of the words in the question and answer sentences, the question sentences in which the morpheme is generated, and the location information (sentence numbers) of the answer sentences in the DB.

9 is a diagram illustrating a method for storing questions and answers from a voice in a voice recognition query response system (using Titles) according to an embodiment of the present invention.

In the present invention, a procedure for receiving a question and an answer with the voice may include providing a predetermined question input unit and an answer input unit, providing the user with the question input unit and receiving a voice as a voice, When the answer input unit is provided to the user and the answer is inputted by voice, the answer voice is received and the answer voice is displayed as the answer text, and the voice of the question and answer is inputted and input from the user When the completion button is clicked, the question sentence and the answer sentence are indexed, and the question sentence where the specific word (keyword) is generated and the location information (sentence number) of the answer sentence are stored in the DB.

In the case of recognizing and storing the speech, the question input speech and the answer input speech are respectively received, and the speech is converted into the question sentence and the answer sentence, and is stored in the DB. The speech is then subjected to morphological analysis and indexed for each keyword. Record the position of the question sentence and the answer sentence (sentence number).

10 is a flow diagram illustrating a method for voice-based query-response conversation in a voice recognition query response system (using Titles) according to an embodiment of the present invention.

The method of operating the voice recognition query response system of the present invention comprises the steps of: receiving a question by voice (S100d); performing a voice recognition step (S200d); generating sentence information (S300d), and outputting the answers extracted or generated by the query response as answer voice and answer text (S400d).

Here, in the first step (S100d) of receiving a question by voice, the voice of the user is sensed and the voice recognition result is received and displayed on the question input window, and the answer sentence for the question is sent to the answer input window Displays the sentence, and outputs the answer voice to the TTI.

At this time, a separate voice input device 1100 may be attached to the outside of the terminal or receive a voice of the user in real time using a built-in voice input device (microphone). Here, if the user does not receive the voice, the user may further receive the text.

In the second step S200d of speech recognition, speech can be recognized by a predetermined speech recognition algorithm and converted into text (sentence).

Next, in a third step S300d of performing a question and answer process on the text (sentence) generated after the speech recognition, a question and answer module 1132 that finds an answer to a specific question based on the question information converted from a voice to a text (sentence) To find an answer or to generate an answer.

At this time, the question answering module 1132 performs a sentence analysis process (morpheme analysis, syntax analysis, semantic analysis, and speech analysis) from the question sentences to grasp the precise intent of the question, When a specific information is requested, an answer is generated based on the information. In answering daily life or common sense, a reply sentence is searched using a similarity search method.

In addition, when the question sentence requests specific information such as time, news, weather, etc., the question and answer module 1132 can generate the answer by fetching the information through the wire / wireless wire / wireless communication network.

Finally, in the fourth step S400d of outputting the answer extracted or generated by the query response as the answer voice and the answer text, the answer sentence (text) extracted or generated by the query response is transmitted, And displays the corresponding answer sentence as text (sentence).

At this time, when a voice is output through the voice recognition system by receiving the answer sentence (text) extracted or generated by the query response, the user can select various voices by voice type, age and gender.

In addition, after the first step S100d, it may further include confirming whether or not the result of speech recognition is correctly inputted.

FIG. 11 is a diagram illustrating a method for voice-based query-response conversation in a voice recognition query response system (using Titles) according to an embodiment of the present invention.

First, when a user's voice is detected and a question is input, the voice analog signal is converted into a digital signal to recognize a voice for a question (S400), converted into a question sentence, a query response process is performed (S410) And outputs the text information of the answer in the form of voice and text.

The query response S410 is a sentence analysis process (morpheme analysis, syntax analysis, semantic analysis, transcription analysis) from the question sentence to grasp the precise intent of the question, and a question requiring an accurate answer (S430) When an answer is requested from the DB (S431) and specific information is requested (S440), an answer is generated based on the information, and an answer requesting daily life or common sense is transmitted to the indexing DB S421 ) And a query response dictionary DB (S422).

That is, the morpheme (word) information included in the question sentence is searched in the indexing DB S421, the question and answer sentence number including the morpheme information is searched in the question and answer dictionary DB (S422) Finds the most frequently asked questions or answers in the question and answer dictionary DB (S422), extracts answers from the question and answer pairs, and outputs them in voice and text form.

FIG. 12 is a screen for voice conversation in the voice recognition question and answer system using Titles according to an embodiment of the present invention.

When talking with a voice, a question voice input start button (S500) is clicked to receive a voice of a user. When the speech recognition is performed after receiving the input, a sentence (for example, who you are?) Is displayed in the question speech input window S510.

When the send (S520) is clicked, the answer sentence text is returned by the question and answer function and the answer sentence is displayed in the answer display window S540 (for example, I am a robot). In addition, the answer sentence is output to a speaker or earphone using a TTS.

In addition, the send button can be pressed and the answer sentence can be received by the question and answer function as soon as the speech recognition sentence is displayed on the question voice input window without setting it as the default.

FIG. 13 is a screen for voice conversation in a voice recognition query response system (using a TTIS) according to an embodiment of the present invention.

When talking with a voice, the user's voice can be automatically input (S500_1). In addition, if the automatic voice input is detected, the automatic answer may be outputted in voice and text form by the question and answer function (S510_1).

FIG. 14 is a screen for displaying a question and an answer sentence after inputting a question and answer voice in a voice recognition question and answer system (using T-TES) according to an embodiment of the present invention.

In order to store a question and an answer from a voice, a voice of a user is input by pressing a question voice input start button (S600) before voice input. After receiving the input, if the speech recognition is performed, a sentence (for example, I love you) that is recognized as a speech is displayed in the question speech input window S610.

Also, the answer voice input start button S630 is pressed to receive the voice of the user. When the speech recognition is performed after receiving the input, a sentence (I love you) which is recognized as a speech is displayed in the answer input window S620.

When the input completion button S660 is pressed, the voice inputted from the voice by the voice response function is voice recognized and stored as the question and answer sentence text, respectively.

If the initialization button S620 or S650 is pressed, the sentence entered in the voice input window S610 and the answer input window S620 can be deleted.

FIG. 15 is a screen for displaying a question and an answer sentence after inputting a question and an answer voice in a voice recognition question and answer system (using T-TES) according to an embodiment of the present invention.

In order to store a question and an answer from a voice, a question voice is input first, and then an answer voice is input. When the input completion button S660_1 is pressed, the voice inputted from the voice by the voice response function is voice recognized and stored as the question and answer sentence text, respectively.

100: voice input device 200:
210: voice input unit 220: voice recognition unit
221: speech recognition 230: natural language processing unit
231: Indexing 232: Statement Analysis
233: Q & A 240: Voice DB
241: voice query information DB 242: indexing DB
243: query response DB 250:
251: Text output 260: Audio output unit
300: audio output device 1100: audio input device
1110: voice input unit 1120: voice recognition unit
1130: Natural language processing unit 1140: Text output unit
1150: Monitor 1160: Audio output unit
1161: TITLE SUB 1170: AUDIO OUTPUT DEVICE

Claims (50)

Recognizes a voice of a question and an answer from a voice of a user and converts the voice into a question and an answer sentence, stores a text file of the question and answer, indexes and stores the question and answer sentence,
And a terminal for outputting a response to the sentence inputted by the question and answer by voice and text when the user inputs a question by voice, Recognition query response system.
Recognizes a voice of a question and an answer from a voice of a user and converts the voice into a question and an answer sentence, stores a text file of the question and answer, indexes and stores the question and answer sentence,
Wherein when the user inputs a question by voice, the speech recognition unit converts the speech into text, performs a query response, and outputs a response to the sentence input by the query response as voice and text. .
3. The method according to claim 1 or 2,
Storing the voice file for the question and answer, and indexing the voice file for the question and answer and storing the voice file.
4. The method according to any one of claims 1 to 3,
A voice input device for inputting voice;
A voice input unit for converting the analog voice transmitted through the voice input device into a digital signal;
A voice recognition unit for performing voice recognition from the voice information received by the voice input unit;
A natural language processing unit for performing indexing and querying based on information converted from speech to text by the speech recognition unit;
A screen output unit for outputting a reply sent from the natural language processing unit as text;
A voice output unit for converting the voice into a digital signal to an analog signal; And
And a voice output device for outputting the voice.
5. The method of claim 4,
Wherein the voice input unit of the voice input unit is an external microphone or an internal microphone of the terminal.
5. The method of claim 4,
Wherein the speech recognition unit recognizes speech by a speech recognition algorithm and converts the speech into text, and stores the text as a text file.
The method according to claim 6,
The sentence text information, which is a result of speech recognition after speech recognition in the speech recognition unit, is stored in a query response DB, and an indexing process is performed based on information of a question and an answer sentence constructed in pairs in the query response DB, Wherein the voice recognition system comprises a voice recognition system.
The method according to claim 6,
Wherein the voice recognition unit stores the recognized voice as a voice file.
5. The method of claim 4,
The natural language processing unit performs an indexing process on the basis of the question and answer sentence information converted from the speech to the text by the speech recognition unit, and then performs an indexing process. In order to inquire an answer to a specific question, And a voice recognition unit for performing a voice recognition process.
10. The method of claim 9,
And a question and answer module for finding answers to the specific questions. The question and answer module analyzes a sentence from a question sentence and grasps the intent of the correct question. And when a specific information is requested, an answer is generated based on the information.
5. The method of claim 4,
Wherein the screen output unit outputs a response sentence transmitted from the natural language processing unit as text on a screen.
5. The method of claim 4,
Wherein the voice output unit outputs a voice file corresponding to a response sentence transmitted from the natural language processing unit to a speaker or an earphone.
4. The method according to any one of claims 1 to 3,
A question input unit and an answer input unit are provided, a question input unit is provided to a user to input a question as a voice, and when the answer input unit is provided and a response is inputted as a voice, And a response sentence, and indexes the question sentence and the answer sentence, and stores the question sentence in which the specific keyword occurs and the location information of the answer sentence in the DB.
14. The method of claim 13,
And stores the question sentence and the voice file path information of the answer sentence in the DB.
4. The method according to any one of claims 1 to 3,
The user's voice is inputted, the voice is converted into a text, the sentence is analyzed and a question and answer is performed, and the answer to the sentence inputted by the question and answer is fetched from the indexing DB and the question and answer DB, And outputting the result as text.
16. The method of claim 15,
And the answer to the sentence inputted by the query response is fetched from the speech DB and outputted as speech and text.
4. The method according to any one of claims 1 to 3,
And a voice input unit for inputting a voice to the user when the voice data detected by the voice is sensed after the voice data is sensed by the user, Recognition query response system.
4. The method according to any one of claims 1 to 3,
Wherein when the question sentence requests specific information such as time, news, weather, etc., the information is fetched through the wired / wireless communication network to generate a response.
4. The method according to any one of claims 1 to 3,
Storing the voice file for the question and answer, and indexing the voice file for the question and answer and storing the voice file.
4. The method according to any one of claims 1 to 3,
Further comprising a text-to-speech unit for converting the response sentence into speech.
21. The method of claim 20,
The user's voice is detected, the voice recognition result is displayed on the question input window, the answer sentence for the question is found by inquiry response, the answer sentence is displayed on the answer input window, and the answer voice is output using the titles Wherein the voice recognition system comprises:
21. The method of claim 20,
Wherein when the answer voice is output using the voice recognition method, the voice recognition voice response system can be user-selectable in various voices, age, sex, and the like.
To save your question and answer sentences by voice,
A step 1 of voice inputting a question and an answer;
A second step of recognizing speech from the speech;
And a third step of indexing the speech recognized speech and the text generated after the speech recognition.
24. The method of claim 23,
The method of claim 1, further comprising the step of storing the voice recognized as a voice file.
25. The method according to claim 23 or 24,
Wherein the voice file corresponding to the question sentence and the answer sentence is stored in association with the question sentence and the answer sentence, respectively.
25. The method according to claim 23 or 24,
In the first step of inputting the question and the answer by voice,
A question input button is provided to the user to check whether or not the voice input button is activated,
An answer input button is provided to the user to check whether or not the voice input button is activated,
And transmits the inputted question and answer to the voice recognition step, respectively.
27. The method of claim 26,
Wherein the question input speech and the answer input speech are stored in a memory.
25. The method according to claim 23 or 24,
In the second step of speech recognition from the speech,
Wherein the voice input unit receives the question input voice and the answer input voice, converts the voice into text, and displays the voice as a question and a reply to the user.
The method according to claim 23 or 24,
The third step of indexing the speech recognized speech and the text generated after the speech recognition,
Extracting a keyword list appearing in the question sentence and an answer sentence and writing the position information of another question sentence and an answer sentence in which the keyword is displayed in a word list and storing the same in an indexing DB. Way.
30. The method of claim 29,
Wherein the query sentence and the voice file path information of the answer sentence are written into the word list and stored in the indexing DB.
24. The method of claim 23,
And storing the sentence as a question and an answer sentence in step 2b.
32. The method of claim 31,
The procedure for inputting the question and the answer with the voice includes:
A question input unit and an answer input unit are provided and the question input unit is provided to the user to input a question as a voice, the speech recognition result is returned, the question voice is displayed as a question text,
When the answer input unit is provided to the user and the answer is inputted as a voice, the answer voice is received and the answer voice is displayed as the reply text,
When the voice input of the question and answer is completed and the click of the input completion button is detected by the user, the question sentence and the answer sentence are indexed and the position information of the question sentence and the answer sentence in which the specific keyword occurs is stored in the DB Wherein the speech recognition system comprises:
32. The method of claim 31,
When recognizing and storing the speech,
A voice input unit for receiving a question input voice and an answer input voice, converting the voice into a question sentence and a response sentence, storing the sentence in a DB, and performing a morphological analysis process for each keyword to index the question sentence, Is recorded in the voice recognition system.
How to communicate with the voice by voice,
A first step of receiving a question voice;
A second step of recognizing speech from the speech;
A third step of analyzing a sentence with text generated after the speech recognition;
After the sentence analysis, a fourth step of querying and responding; And
And outputting, as the voice and text, a reply extracted from the query response DB or generated through the query response DB after the query response.
35. The method of claim 34,
In the first step of receiving the question voice,
Wherein a separate voice input device is attached to the outside of the terminal or a user's voice is input in real time using the built-in voice input device.
35. The method of claim 34,
In the second step of speech recognition from the speech,
A method of operating a voice recognition question answering system, the method comprising: receiving voice of a user and performing voice recognition to convert the voice of the user into text.
35. The method of claim 34,
In the third step of analyzing the sentence with the text generated after the speech recognition,
And a preprocessing step of performing a query response by extracting words in a sentence by morpheme analysis of the text.
35. The method of claim 34,
After analyzing the sentence, the fourth step of query response is:
A sentence analysis, a response sentence extraction algorithm, or a response sentence generation algorithm.
35. The method of claim 34,
After the query response, the fifth step of outputting the answers extracted from the query response DB or generated through the query response DB as voice and text,
When the existing answer sentence is extracted through the question and answer DB, the existing answer sentence is outputted as a voice through the voice file and displayed as text,
And when a new answer sentence is generated through the question and answer DB, the answer sentence is output to the corresponding answer sentence through a voice message, and the corresponding answer sentence is displayed as text.
How to communicate with the voice by voice,
A step 1 for inputting a question by voice;
Two steps of speech recognition;
A third step of performing a query response processing on the sentence information generated after the speech recognition; And
And outputting the answers extracted or generated by the query response as an answer voice and an answer text.
41. The method of claim 40,
In the first step of inputting the question by the voice,
The user's voice is detected and the voice recognition result is returned and displayed on the question input window. After the question and answer of the question, the answer sentence is displayed on the answer input window and the answer voice is output to the TTI A method of operating a voice recognition query response system. 41. The method of claim 40,
Wherein a separate voice input device is attached to the outside of the terminal or a user's voice is input in real time using the built-in voice input device.
41. The method of claim 40,
Further comprising the step of receiving a text if the voice input is not received.
41. The method of claim 40,
In the second step of speech recognition,
And recognizing the voice by the voice recognition algorithm and converting the voice into text.
41. The method of claim 40,
In the third step of performing the query response processing with the text generated after the speech recognition,
Wherein the answer is found by a question and answer module that finds an answer to a specific question based on the question information converted from voice to text, or an answer is generated.
41. The method of claim 40,
In the third step of performing the query response processing with the text generated after the speech recognition,
The question and answer module analyzes a sentence analysis process from a question sentence to grasp an accurate question intention. A question requesting an accurate answer takes an answer from a pre-established answer DB. When requesting specific information, And a response sentence is searched by using a similarity search method to an answer requesting daily life or common sense.
41. The method of claim 40,
In the third step of performing the query response processing with the text generated after the speech recognition,
Wherein the query response module generates the response by fetching the information through the wire / wireless communication network when the question sentence requests specific information such as time, news, and weather.
41. The method of claim 40,
The fourth step of outputting the answer extracted or generated by the query response as the answer voice and the answer text,
Receiving a response sentence extracted or generated by a query response, outputting a voice through a text message, and displaying the response sentence in text form.
41. The method of claim 40,
The fourth step of outputting the answer extracted or generated by the query response as the answer voice and the answer text,
The voice recognition system according to claim 1 or 2, wherein when the speech sent out or generated by the query response is received and the voice is outputted through the voice recognition system, How to operate the Q & A system.
41. The method of claim 40,
After the first step,
Further comprising the step of confirming whether the result of speech recognition is correctly inputted.


KR1020130040660A 2013-04-12 2013-04-12 Question answering system using speech recognition and its application method thereof KR20140123369A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
KR1020130040660A KR20140123369A (en) 2013-04-12 2013-04-12 Question answering system using speech recognition and its application method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
KR1020130040660A KR20140123369A (en) 2013-04-12 2013-04-12 Question answering system using speech recognition and its application method thereof

Publications (1)

Publication Number Publication Date
KR20140123369A true KR20140123369A (en) 2014-10-22

Family

ID=51994100

Family Applications (1)

Application Number Title Priority Date Filing Date
KR1020130040660A KR20140123369A (en) 2013-04-12 2013-04-12 Question answering system using speech recognition and its application method thereof

Country Status (1)

Country Link
KR (1) KR20140123369A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017051949A1 (en) * 2015-09-23 2017-03-30 (주)에프에스알엔티 Human care device
KR20190079791A (en) * 2017-12-28 2019-07-08 네이버 주식회사 Method for providing service using plurality wake up word in artificial intelligence device, and system thereof
US10389873B2 (en) 2015-06-01 2019-08-20 Samsung Electronics Co., Ltd. Electronic device for outputting message and method for controlling the same
WO2019168253A1 (en) * 2018-02-27 2019-09-06 주식회사 와이즈넛 Interactive counseling chatbot device and method for hierarchically understanding user's expression and generating answer
US10446145B2 (en) 2015-11-27 2019-10-15 Samsung Electronics Co., Ltd. Question and answer processing method and electronic device for supporting the same
CN111966840A (en) * 2020-08-18 2020-11-20 北京猿力未来科技有限公司 Man-machine interaction management method and management system for language teaching
KR20220073350A (en) * 2020-11-26 2022-06-03 주식회사 포켓메모리 A method and apparatus for providing conversation service through external data linkage

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10389873B2 (en) 2015-06-01 2019-08-20 Samsung Electronics Co., Ltd. Electronic device for outputting message and method for controlling the same
WO2017051949A1 (en) * 2015-09-23 2017-03-30 (주)에프에스알엔티 Human care device
US10446145B2 (en) 2015-11-27 2019-10-15 Samsung Electronics Co., Ltd. Question and answer processing method and electronic device for supporting the same
KR20190079791A (en) * 2017-12-28 2019-07-08 네이버 주식회사 Method for providing service using plurality wake up word in artificial intelligence device, and system thereof
WO2019168253A1 (en) * 2018-02-27 2019-09-06 주식회사 와이즈넛 Interactive counseling chatbot device and method for hierarchically understanding user's expression and generating answer
CN111966840A (en) * 2020-08-18 2020-11-20 北京猿力未来科技有限公司 Man-machine interaction management method and management system for language teaching
KR20220073350A (en) * 2020-11-26 2022-06-03 주식회사 포켓메모리 A method and apparatus for providing conversation service through external data linkage

Similar Documents

Publication Publication Date Title
CN109493850B (en) Growing type dialogue device
WO2021232725A1 (en) Voice interaction-based information verification method and apparatus, and device and computer storage medium
KR102191425B1 (en) Apparatus and method for learning foreign language based on interactive character
CN104078044B (en) The method and apparatus of mobile terminal and recording search thereof
KR20140123369A (en) Question answering system using speech recognition and its application method thereof
US11494434B2 (en) Systems and methods for managing voice queries using pronunciation information
KR20130086971A (en) Question answering system using speech recognition and its application method thereof
US20100217591A1 (en) Vowel recognition system and method in speech to text applictions
KR20180064504A (en) Personalized entity pronunciation learning
KR20130108173A (en) Question answering system using speech recognition by radio wire communication and its application method thereof
CN101158947A (en) Method and apparatus for machine translation
CN107844470B (en) Voice data processing method and equipment thereof
CN105210147B (en) Method, apparatus and computer-readable recording medium for improving at least one semantic unit set
CN109543021B (en) Intelligent robot-oriented story data processing method and system
CN106713111B (en) Processing method for adding friends, terminal and server
CN110427455A (en) A kind of customer service method, apparatus and storage medium
WO2021051564A1 (en) Speech recognition method, apparatus, computing device and storage medium
CN112669842A (en) Man-machine conversation control method, device, computer equipment and storage medium
WO2021179703A1 (en) Sign language interpretation method and apparatus, computer device, and storage medium
US20210034662A1 (en) Systems and methods for managing voice queries using pronunciation information
JP2012168349A (en) Speech recognition system and retrieval system using the same
KR102536944B1 (en) Method and apparatus for speech signal processing
US11410656B2 (en) Systems and methods for managing voice queries using pronunciation information
KR20160104243A (en) Method, apparatus and computer-readable recording medium for improving a set of at least one semantic units by using phonetic sound
KR20130116128A (en) Question answering system using speech recognition by tts, its application method thereof

Legal Events

Date Code Title Description
WITN Withdrawal due to no request for examination