Processing

Please wait...

PATENTSCOPE will be unavailable a few hours for maintenance reason on Saturday 31.10.2020 at 7:00 AM CET
Settings

Settings

Goto Application

1. WO2016151699 - LEARNING APPARATUS, METHOD, AND PROGRAM

Publication Number WO/2016/151699
Publication Date 29.09.2016
International Application No. PCT/JP2015/058564
International Filing Date 20.03.2015
IPC
G10L 15/22 2006.01
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
22Procedures used during a speech recognition process, e.g. man-machine dialog
G10L 13/00 2006.01
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
13Speech synthesis; Text to speech systems
G10L 15/10 2006.01
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
08Speech classification or search
10using distance or distortion measures between unknown speech and reference templates
G10L 15/28 2013.01
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
28Constructional details of speech recognition systems
CPC
G06F 40/30
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
40Handling natural language data
30Semantic analysis
G10L 15/063
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
06Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
063Training
G10L 15/10
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
08Speech classification or search
10using distance or distortion measures between unknown speech and reference templates
G10L 15/1815
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
08Speech classification or search
18using natural language modelling
1815Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
G10L 15/22
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
22Procedures used during a speech recognition process, e.g. man-machine dialogue
G10L 15/30
GPHYSICS
10MUSICAL INSTRUMENTS; ACOUSTICS
LSPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
15Speech recognition
28Constructional details of speech recognition systems
30Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
Applicants
  • 株式会社 東芝 KABUSHIKI KAISHA TOSHIBA [JP]/[JP]
Inventors
  • 藤井 寛子 FUJII, Hiroko
Agents
  • 蔵田 昌俊 KURATA, Masatoshi
Priority Data
Publication Language Japanese (JA)
Filing Language Japanese (JA)
Designated States
Title
(EN) LEARNING APPARATUS, METHOD, AND PROGRAM
(FR) PROGRAMME, PROCÉDÉ ET APPAREIL D'APPRENTISSAGE
(JA) 学習装置、方法およびプログラム
Abstract
(EN)
The present invention makes it possible to reduce the data production costs for intention estimation. A learning apparatus according to an embodiment of the present invention comprises a first storage unit, a detection unit, and a correction unit. The learning apparatus uses a spoken intention of a user inferred from a first text, which is the result of speech recognition of the user's speech. The first storage unit stores similarity information including at least a second text, which is the result of voice recognition of similar speech indicating a series of similar speech in a dialogue history, intention candidates inferred from speech that was determined as having a successful dialogue in the similar speech, and a certainty factor indicating the degree to which the intention candidate becomes the intention of the second text. The detection unit, from the similarity information, detects corresponding similarity information including the second text matching the first text. The correction unit corrects the spoken intention to the intention candidates included in the corresponding similarity information in cases where the certainty factor included in the corresponding similarity information exceeds or is equal to a threshold value.
(FR)
La présente invention concerne la possibilité de réduire les coûts de production de données pour l'estimation d'intention. Un appareil d'apprentissage selon un mode de réalisation de la présente invention comprend une première unité de mémorisation, une unité de détection et une unité de correction. L'appareil d'apprentissage utilise une intention parlée d'un utilisateur déduite à partir d'un premier texte, qui est le résultat d'une reconnaissance de parole de la parole d'un utilisateur. La première unité de mémorisation mémorise des informations de similitude comprenant au moins un second texte, qui est le résultat d'une reconnaissance vocale de parole similaire indiquant une série de paroles similaires dans un historique de dialogue, des candidats d'intention déduits à partir de la parole déterminée comme ayant un dialogue réussi dans la parole similaire et un facteur de certitude indiquant le degré selon lequel le candidat d'intention devient l'intention du second texte. L'unité de détection détecte, à partir des informations de similitude, des informations de similitude correspondantes comprenant le second texte correspondant au premier texte. L'unité de correction corrige l'intention parlée dans les candidats d'intention compris dans les informations de similitude correspondantes, dans des cas dans lesquels le facteur de certitude compris dans les informations de similitude correspondantes est égal ou supérieur à une valeur seuil.
(JA)
意図推定のためのデータ作成コストを低減できる。 本実施形態に係る学習装置は、第1格納部、検出部および修正部を含む。学習装置は、ユーザの発話を音声認識した結果である第1テキストから推定された該ユーザの発話意図を用いる。第1格納部は、対話履歴の中で類似する一連の発話を示す類似発話を音声認識した結果である第2テキストと、該類似発話の中で対話が成功したと判定された発話から推定される意図候補と、該意図候補が該第2テキストの意図となる度合いを示す確信度とを少なくとも含む類似情報を格納する。検出部は、前記類似情報から、前記第1テキストと一致する前記第2テキストを含む対応類似情報を検出する。修正部は、前記対応類似情報に含まれる確信度が閾値以上である場合、前記発話意図を、該対応類似情報に含まれる意図候補に修正する。
Also published as
Latest bibliographic data on file with the International Bureau