JP3525999B2 - Language understanding method and language understanding device - Google Patents

Language understanding method and language understanding device

Info

Publication number
JP3525999B2
JP3525999B2 JP03783798A JP3783798A JP3525999B2 JP 3525999 B2 JP3525999 B2 JP 3525999B2 JP 03783798 A JP03783798 A JP 03783798A JP 3783798 A JP3783798 A JP 3783798A JP 3525999 B2 JP3525999 B2 JP 3525999B2
Authority
JP
Japan
Prior art keywords
dialogue
understanding
candidate
context
word
Prior art date
Legal status (The legal status 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 status listed.)
Expired - Fee Related
Application number
JP03783798A
Other languages
Japanese (ja)
Other versions
JPH11237894A (en
Inventor
幹生 中野
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nippon Telegraph and Telephone Corp
Original Assignee
Nippon Telegraph and Telephone Corp
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 Nippon Telegraph and Telephone Corp filed Critical Nippon Telegraph and Telephone Corp
Priority to JP03783798A priority Critical patent/JP3525999B2/en
Publication of JPH11237894A publication Critical patent/JPH11237894A/en
Application granted granted Critical
Publication of JP3525999B2 publication Critical patent/JP3525999B2/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Landscapes

  • Machine Translation (AREA)

Description

【発明の詳細な説明】Detailed Description of the Invention

【0001】[0001]

【発明の属する技術分野】本発明は、入力発語の言語理
解を行い対話文脈候補を出力する言語理解方法及び言語
理解装置に関し、詳しくは、たとえば人間が日常使って
いる言葉を、音声認識装置を介して文字列として計算機
に入力し、その文字列が意味する内容を計算機処理する
音声対話システム等に利用される言語理解方法及び言語
理解装置に関するものである。
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a language comprehension method and a language comprehension apparatus for comprehending the language of an input utterance and outputting a dialogue context candidate. The present invention relates to a language comprehension method and a language comprehension apparatus used in a voice dialogue system or the like for inputting a character string into a computer via a computer and processing the content of the character string by computer.

【0002】[0002]

【従来の技術】従来の言語理解技術は、文または発話の
切れ目が明らかなものを対象にしていた。書き言葉の解
析であれば句点を文の区切りとし、句点から句点までの
間のテキストを言語理解規則を用いて解析し、その意味
内容を計算し、それと、それより前の文から計算された
文脈情報とあわせて、新しい文脈情報を計算する方法が
用いられていた。話し言葉の解析の場合は、ユーザに発
話時にボタン等を押させることによって発話区間を調べ
るか、または、ポーズをもって発話の区切りとし、発話
終了後に直前の発話区間の言語解析を行い、その意味内
容を計算し、文脈情報を計算する方法が用いられてい
た。
2. Description of the Related Art Conventional language comprehension techniques have been aimed at sentences or utterances with clear breaks. In the case of analysis of written words, punctuation marks are used as sentence delimiters, the text between punctuation marks is analyzed using language understanding rules, the meaning content is calculated, and the context calculated from the sentence before it. Along with the information, a method of calculating new contextual information was used. In the case of analysis of spoken language, the user can press the button etc. at the time of utterance to check the utterance section, or use the pause as a utterance segment, and after the utterance ends, perform a language analysis of the immediately preceding utterance section and determine the meaning. The method of calculating and calculating contextual information was used.

【0003】[0003]

【発明が解決しようとする課題】話し言葉では文という
単位が明確ではない。単語1つで意図を表すこともある
し、長い発話を話すこともある。ポーズなどの区切りを
用いるとすると、ポーズが現れるまで入力の単位が決定
されないため、発話の意味の把握が行えず高速な反応が
できない。また、ユーザが多くの事柄を切目なく話す場
合もあり、その途中で意味を把握し、相槌をうったり問
い返しを行うなど、適切に反応しなくてはならない場合
もある。連続した発声の中から、言語理解規則を用いて
発話区切りを見つけることを試みても、規則の適用の仕
方は曖昧で様々な意味理解が可能であるため、どれを用
いればよいのか不明であるという問題がある。
[Problems to be Solved by the Invention] The unit of sentence is not clear in spoken language. Sometimes a single word expresses intent and sometimes speaks long utterances. If a pause is used, the unit of input is not determined until the pose appears, so the meaning of the utterance cannot be grasped and a fast reaction cannot be performed. In addition, there are cases where the user speaks a lot of things without interruption, and in the middle of the process, the user may have to grasp the meaning, and respond appropriately, such as humming a question or asking a question. Even if an attempt is made to find a speech break from a continuous utterance using a language understanding rule, the method of applying the rule is ambiguous and various meanings can be understood, so it is unclear which one should be used. There is a problem.

【0004】本発明の目的は、発語の区切りが明確にわ
からなくても、音声の理解を実時間で行うことを可能と
する言語理解方法及び装置を提供することにある。
An object of the present invention is to provide a language comprehension method and apparatus that enable the understanding of speech in real time even if the boundaries between utterances are not clearly understood.

【0005】[0005]

【課題を解決するための手段】本発明の言語理解方法
は、対話文脈候補を優先度と対話の理解の状態を記録し
た理解記録とで構成し、前の時点までに得られている対
話文脈候補を保持しておくようにし、新たに単語が入力
されたときに辞書を参照して、該単語の意味記述を求め
るステップと、前記単語が入力された時点までに得られ
ている各対話文脈候補の理解記録に前記単語の意味記述
を追加してそれぞれ新たな対話文脈候補を得るステップ
と、前記新たに得られた各対話文脈候補の理解記録に言
語理解規則を適用し、該適用に成功した場合に、該適用
結果を理解記録に反映させて更に新たな対話文脈候補を
得るステップと、前記言語理解規則の適用に成功して得
られた対話文脈候補と適用前の各対話文脈候補の各々の
優先度を、適用された言語理解規則と適用される前の理
解記録とに依存した値に変えるステップと、前記言語理
解規則の適用に成功して得られた対話文脈候補と適用前
の各対話文脈候補のうちで、優先度の最も高い対話文脈
候補を出力するステップとを有して、これら各ステップ
を単語が入力される度に繰り返すことを特徴とする。
A method for understanding a language according to the present invention
Records the context and the state of comprehension of dialogue context candidates
Comprehension records and the pairs that have been obtained up to the previous time.
Input a new word while keeping the story context candidate
When you are asked, refer to the dictionary to find the meaning description of the word
And the steps obtained by the time the word is entered
Meaning description of the word in the understanding record of each dialogue context candidate
To get new dialogue context candidates
And the newly obtained understanding record of each dialogue context candidate.
Apply the word understanding rule, and if the application is successful, the application
By reflecting the results in the understanding record, we can create new dialogue context candidates.
And the success of applying the language understanding rules.
Of each of the proposed dialogue context candidates and the respective dialogue context candidates before application.
Priorities are defined by the applied language understanding rules and the
The step of changing the value depending on the solution record and the linguistic theory
Dialog context candidates obtained by successful application of solution rules and before application
Among the dialogue context candidates of, the dialogue context with the highest priority
And outputting each of these steps, and each of these steps
Is repeated every time a word is input.

【0006】また、本発明の言語理解装置は、対話文脈
候補を優先度と対話の理解の状態を記録した理解記録と
で構成し、前の時点までに得られている対話文脈候補を
保持する手段と、新たに単語が入力されるときに辞書を
参照して、該単語の意味記述を求め、前記単語が入力さ
れた時点までに得られている各対話文脈候補の理解記
に前記単語の意味記述を追加してそれぞれ新たな対話文
脈候補を得る手段と、前記新たに得られた各対話文脈候
補の理解記録に言語理解規則を適用し、該適用に成功し
た場合に、該適用結果を理解記録に反映させて更に新た
な対話文脈候補を得る手段と、前記言語理解規則の適用
に成功して得られた対話文脈候補と適用前の各対話文脈
候補の各々の優先度を、適用された言語理解規則と適用
される前の理解記録とに依存した値に変える手段と、前
記言語理解規則の適用に成功して得られた対話文脈候補
と適用前の各対話文脈候補のうちで、優先度の最も高い
対話文脈候補を出力する手段とを具備し、各手段の処理
を単語が入力される度に繰り返すことを特徴とする。
Further, the language understanding device of the present invention has a dialogue context.
The candidate is an understanding record that records the priority and the state of understanding of the dialogue.
And the dialogue context candidates obtained by the previous time.
A means to hold and a dictionary when a new word is input
Referring to the meaning description of the word, the word is input.
Understanding Symbol record of each interaction context candidates that are obtained by the time that
Adding the semantic description of the above words to each new dialogue sentence
Means for obtaining pulse candidates and the newly obtained dialogue context indicators
Applied the language understanding rule to the comprehension record and succeeded in applying it.
In the case of
For obtaining various dialogue context candidates and application of the language understanding rule
Dialogue Candidates Successfully Obtained and Each Dialogue Context Before Application
The priority of each candidate is applied to the applied language understanding rules and applied
The means to change the value depending on the understanding record before
Dialog context candidates obtained by successful application of the language understanding rules
Of the dialogue context candidates before and applied, the highest priority
And means for outputting the dialogue context candidates, and processing of each means
Is repeated every time a word is input.

【0007】[0007]

【発明の実施の形態】図1は、本発明にかかる言語理解
方法の一実施例の処理フロー図を示す。なお、これは所
謂計算機の支援のもとに実行される。単語が入力される
と(ステップ110)、まず、辞書(単語辞書)を参照
して、入力された単語の意味記述(品詞、意味分類等)
を求める(ステップ120)。次に、この意味記述を、
その前の単語入力の時点までに得られている対話文脈候
補の各々とあわせることにより、すなわち、対話文脈候
補の理解記録に該意味記述を追加することにより、新た
な対話文脈候補の組み合わせを求める(ステップ13
0)。その後、これらの対話文脈候補について言語理解
規則を適用し、適用に成功した場合、その結果につい
て、更に新たな対話文脈候補の組み合せを求める(ステ
ップ140)。すなわち、適用結果を理解記録に反映さ
せた対話文脈候補を得る。そして、これら新たな対話文
脈候補と適用前の元の各対話文脈候補との優先度を計算
し(ステップ150)、最も優先度の高い対話文脈候補
を出力する(ステップ160)。以下、単語が入力され
る度に、ステップ110〜160の処理を繰り返し実行
する。
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a processing flow chart of an embodiment of a language understanding method according to the present invention . It should be noted that this is executed with the support of a so-called computer. When a word is input (step 110), first, referring to a dictionary (word dictionary), the meaning description of the input word (part of speech, meaning classification, etc.)
Is calculated (step 120). Next, this semantic description
By combining with each of the dialog context candidates obtained up to the time of the previous word input, that is, the dialog context index
A new combination of dialogue context candidates is obtained by adding the meaning description to the complementary understanding record (step 13).
0). Thereafter, the language understanding rule is applied to these dialogue context candidates, and if the application is successful, a new combination of dialogue context candidates is obtained for the result (step 140). That is, the application result is reflected in the understanding record.
Get the proposed dialogue context candidate. Then, the priorities of these new dialogue context candidates and the original dialogue context candidates before application are calculated (step 150), and the dialogue context candidate with the highest priority is output (step 160). Hereinafter, every time a word is input, the processes of steps 110 to 160 are repeatedly executed.

【0008】以上のように、本発明の言語理解方法にお
いては、入力発話に対する言語理解規則の適用の仕方が
複数ある場合に、その各々の方法で適用した結果得られ
る複数の対話文脈の候補を保持し、それらの優先度を計
算した上で、最も優先度の高いものを調べて出力する。
これにより、発話の区切りが明確にわからなくても、各
時点で最も優先度の高い文脈を得ることが可能であるた
め、実時間で出力することが可能である。
As described above, in the language understanding method of the present invention, when there are a plurality of ways of applying the language understanding rule to the input utterance, a plurality of candidates for the dialogue context obtained as a result of applying the respective methods are selected. It retains them, calculates their priorities, and then checks and outputs the one with the highest priority.
With this, it is possible to obtain the context with the highest priority at each point in time even if the utterance segment is not clearly known, and therefore it is possible to output in real time.

【0009】次に、図2に、本発明にかかる言語理解装
置の一実施例の構成図を示す。本言語理解装置200は
対話文脈管理部210、規則適用部220、優先度計算
部230、対話文脈候補保持部240、辞書(単語辞
書)250及び言語理解規則260よりなる。対話文脈
候補保持部240は、前の時点までに得られた対話文脈
候補(優先度と理解記録とで構成される)を保持してい
る。対話文脈管理部210は、単語が入力されると、辞
書250を参照して、該入力された単語の意味記述を求
め、対話文脈候補保持部240に保持されている前の単
語入力の時点までに得られた対話文脈候補の各々の理解
記録と合わせることにより、新たな対話文脈候補の組み
合わせを求め、規則適用部220に送る。規則適用部2
20は、対話文脈管理部210から送られた対話文脈候
補の各々に(理解記録に)言語理解規則260の適用を
試み、それが成功した場合、その結果として、理解記録
に適用結果が反映された新たな対話文脈候補を得、これ
ら得られた対話文脈候補を元の対話文脈候補とあわせて
優先度計算部230に送る。優先度計算部230は、規
則適用部220から送られた対話文脈候補の各々の優先
度を計算し、対話文脈管理部210に戻す。対話文脈管
理部210は、優先度度計算部230から戻ってきた優
先度の付された各対話文脈候補を対話文脈候補保持部2
40に書き込むとともに、その時点で最も優先度の高い
対話文脈候補を出力する。
Next, FIG. 2 shows a block diagram of an embodiment of a language understanding device according to the present invention. The language understanding device 200 includes a dialogue context management unit 210, a rule application unit 220, a priority calculation unit 230, a dialogue context candidate holding unit 240, a dictionary (word dictionary) 250, and a language understanding rule 260. The dialogue context candidate holding unit 240 holds the dialogue context candidates (composed of priority and understanding record) obtained up to the previous time. When a word is input, the dialog context management unit 210 refers to the dictionary 250 to obtain the meaning description of the input word, and holds the dialog context candidate holding unit 240 until the previous word input time. Of each of the dialogue context candidates obtained in
By combining with the record , a new combination of dialogue context candidates is obtained and sent to the rule application unit 220. Rule application part 2
20 tries to apply the language understanding rule 260 (to the understanding record) to each of the dialog context candidates sent from the dialog context management unit 210, and if it succeeds, as a result , the understanding record.
A new dialogue context candidate in which the application result is reflected is obtained, and the obtained dialogue context candidates are sent to the priority calculation unit 230 together with the original dialogue context candidates. The priority calculation unit 230 calculates the priority of each dialogue context candidate sent from the rule application unit 220, and returns it to the dialogue context management unit 210. The dialogue context management unit 210 stores each dialogue context candidate with a priority returned from the priority calculation unit 230 in the dialogue context candidate holding unit 2.
While writing in 40, the dialogue context candidate having the highest priority at that time is output.

【0010】図3に、一例として本言語理解装置200
を適用した音声対話システムの構成図を示す。図3にお
いて、処理対象となる音声が音声認識装置300に入力
されると、音声中の単語が認識される毎に、言語理解装
置200に送られる。言語理解装置200では、図2に
示したように、単語が入力される度に、辞書250と言
語理解規則260を用いて対話文脈を得て逐次出力し、
応答計画装置400に送る。応答計画装置400は、文
脈情報に応じて、応答が必要かどうかを判断し、必要な
らば適当な応答を計画し、応答メッセージを音声合成装
置500を用いて生成して出力する。
FIG. 3 shows, as an example, the language understanding device 200.
The block diagram of the spoken dialogue system which applied is shown. In FIG. 3, when a voice to be processed is input to the voice recognition device 300, it is sent to the language understanding device 200 every time a word in the voice is recognized. In the language understanding device 200, as shown in FIG. 2, every time a word is input, the dictionary 250 and the language understanding rule 260 are used to obtain a dialogue context and sequentially output the dialogue context.
It is sent to the response planning device 400. The response planning device 400 determines whether or not a response is necessary according to the context information, plans an appropriate response if necessary, and generates and outputs a response message using the voice synthesizer 500.

【0011】ここで、具体例として、会議室を予約する
対話中の「第一会議室、2時30分」(以下発話1)と
いう入力の理解を考える。音声認識装置300は音声入
力のなかから、「第一会議室、2時、30分」という3
つの単語を認識する度に、逐次的に言語理解装置200
に送る。言語理解装置200内の辞書250、言語理解
規則260は、それぞれ図4、図5のようになっている
とする。図5の一行ごとが言語理解規則に対応する。た
とえば、 発話(会議室設定)←名詞(会議室) は、意味分類が「会議室」であるような名詞が、「会議
室設定」を命令する発話であると解釈できることを示
す。また、 名詞(時間)←名詞(時間)名詞(分) は、意味分類が「時間」であるような名詞と、意味分類
が「分」であるような名詞の連続が、意味分類が「時
間」であるような名詞と解釈できることを示す。
Here, as a concrete example, let us consider an understanding of an input of "first meeting room, 2:30" (hereinafter, utterance 1) during a dialogue for reserving a meeting room. From the voice input, the voice recognition device 300 calls "1st conference room, 2:30, 3".
Every time one word is recognized, the language understanding device 200 is sequentially
Send to. It is assumed that the dictionary 250 and the language understanding rule 260 in the language understanding device 200 are as shown in FIGS. 4 and 5, respectively. Each line in FIG. 5 corresponds to the language understanding rule. For example, utterance (meeting room setting) ← noun (meeting room) indicates that a noun whose semantic classification is "meeting room" can be interpreted as an utterance instructing "meeting room setting". In addition, noun (time) ← noun (time) noun (minute) is a noun whose semantic classification is "time" and a sequence of nouns whose semantic classification is "minute" is "time". It shows that it can be interpreted as a noun such as.

【0012】まず、「第一会議室」という単語が入力さ
れると、対話文脈管理部210において、次のような対
話文脈候補(1)が形成される。なお、対話文脈候補保
持部240は、この時点では空とする。 (1)優先度:0 理解記録:名詞(会議室,第一会議室) 対話文脈候補は、優先度と理解記録からなり、優先度の
初期値は0であるとする。理解記録には今までの対話の
理解の状態が記録される。ここでは、図4の「第一会議
室」の辞書項目から、意味記述「名詞(会議室,第一会
議室)」が求められ、記録される。
First, when the word "first conference room" is input, the dialog context management unit 210 forms the following dialog context candidate (1). The dialogue context candidate holding unit 240 is empty at this point. (1) Priority: 0 Understanding record: noun (meeting room, first meeting room) Dialog context candidates are composed of priority and understanding record, and the initial value of the priority is 0. The understanding record records the state of understanding of the dialogue so far. Here, the meaning description "noun (meeting room, first meeting room)" is obtained and recorded from the dictionary item of "first meeting room" in FIG.

【0013】次に、対話文脈候補(1)が、規則適用部
220に送られる。規則適用部220では、図5の言語
理解規則を対話文脈候補(1)の理解記録に適用する。
この場合、図5の規則(A)が適用でき、次のような新
たな対話文脈候補(2),(3)ができる。
Next, the dialogue context candidate (1) is sent to the rule application section 220. The rule application unit 220 applies the language understanding rule of FIG. 5 to the understanding record of the dialogue context candidate (1).
In this case, the rule (A) of FIG. 5 can be applied, and the following new dialogue context candidates (2) and (3) can be obtained.

【0014】(2)優先度:0 理解記録:名詞(会議室,第一会議室) (3)優先度:0 理解記録:発話(会議室設定,第一会議室) 対話文脈候補(2)は、対話文脈候補(1)に規則を何
も適用しなかったものであり、対話文脈候補(3)が図
5の規則(A)の適用に成功したものである。ここで、
対話文脈候補(3)では、理解記録に規則(A)の適用
結果が反映されている。
(2) Priority: 0 Understanding record: noun (meeting room, first meeting room) (3) Priority: 0 Understanding record: utterance (meeting room setting, first meeting room) Dialogue context candidate (2) Shows that no rule was applied to the dialogue context candidate (1), and the dialogue context candidate (3) succeeded in applying the rule (A) of FIG. here,
Application of rule (A) to the understanding record in the dialogue context candidate (3)
The result is reflected.

【0015】これら二つの対話文脈候補(2),(3)
が優先度計算部230に送られる。ここで、規則を適用
した対話文脈候補(3)の方の優先度が増やされ、次の
(3)′のようになる。 (3)′優先度:10 理解記録:発話(会議室設定,第一会議室) 優先度は、使われる規則と、そのときの理解記録の状態
によるとする。本例では、規則が1つ適用される度に、
「10」増やされるとする。この対話文脈候補(3)′
が対話文脈管理部210より応答計画装置400に送ら
れる。
These two dialogue context candidates (2) and (3)
Is sent to the priority calculation unit 230. Here, the priority of the dialog context candidate (3) to which the rule is applied is increased to become the following (3) ′. (3) ′ Priority: 10 Understanding record: Utterance (meeting room setting, first meeting room) The priority depends on the rule used and the state of the understanding record at that time. In this example, each time a rule is applied,
It is supposed to be increased by "10". This dialogue context candidate (3) ′
Is sent from the dialogue context management unit 210 to the response planning device 400.

【0016】次に、「2時」が入力されたときの対話文
脈管理部210の対話文脈候補は、次のようになる。 (4)優先度:0 理解記録:名詞(会議室,第一会議室),名詞(時間,
2時) (5)優先度:10 理解記録:発話(会議室設定,第一会議室),名詞(時
間,2時)。
Next, the dialog context candidates of the dialog context management unit 210 when "2 o'clock" is input are as follows. (4) Priority: 0 Understanding record: noun (meeting room, first meeting room), noun (time,
2:00) (5) Priority: 10 Understanding record: Utterance (meeting room setting, first meeting room), noun (time, 2:00).

【0017】これら対話文脈候補(4),(5)は規則
適用部220、優先度計算部230を経て次のようにな
る。 (6)優先度:10 理解記録:名詞(会議室,第一会議室),発話(時間設
定,2時) (7)優先度:10 理解記録:発話(会議室設定,第一会議室),名詞(時
間,2時) (8)優先度:20 理解記録:発話(会議室設定,第一会議室),発話(時
間設定,2時) 対話文脈候補(6)は、対話文脈候補(4)に図5の規
則(B)を適用したものである。対話文脈候補(7)
は、対話文脈候補(5)に何も規則を適用しなかったも
のであり、対話文脈候補(8)は、対話文脈候補(5)
図5の規則(B)を適用したものである。対話文脈候
補(4)に何も規則を適用しなかったものは、優先度が
低いので捨てられる。これらの結果、対話文脈候補
(8)が対話文脈管理部210より応答計画装置400
に送られる。
The dialogue context candidates (4) and (5) are as follows via the rule application section 220 and the priority calculation section 230. (6) Priority: 10 Understanding record: noun (meeting room, first meeting room), utterance (time setting, 2:00) (7) Priority: 10 Understanding record: utterance (meeting room setting, first meeting room) , Noun (time, 2 o'clock) (8) Priority: 20 Understanding record: Utterance (meeting room setting, first meeting room), Utterance (time setting, 2 o'clock) Dialogue context candidate (6) is dialogue context candidate ( The rule (B) of FIG. 5 is applied to 4). Dialogue context candidates (7)
Means that no rule is applied to the dialogue context candidate (5), and the dialogue context candidate (8) is the dialogue context candidate (5).
In an application of the rule in FIG. 5 (B). Those that have not applied any rules to the dialogue context candidates (4) are discarded because they have low priority. As a result, the dialog context candidate (8) is determined by the dialog context management unit 210 as the response plan device 400.
Sent to.

【0018】次に、「30分」が入力された時、対話文
脈管理部210では次のような対話文脈候補(9),
(10),(11)が形成される。 (9)優先度:10 理解記録:発話(会議室設定,第一会議室),名詞(時
間,2時),名詞(分,30分) (10)優先度:10 理解記録:名詞(会議室,第一会議室),発話(時間設
定,2時),名詞(分,30分) (11)優先度:20 理解記録:発話(会議室設定,第一会議室),発話(時
間設定,2時),名詞(分,30分)。
Next, when "30 minutes" is input, the dialog context management unit 210 causes the following dialog context candidates (9),
(10) and (11) are formed. (9) Priority: 10 Understanding record: Utterance (meeting room setting, first meeting room), noun (time, 2:00), noun (minute, 30 minutes) (10) Priority: 10 Understanding record: noun (meeting) Room, first meeting room), utterance (time setting, 2:00), noun (minute, 30 minutes) (11) Priority: 20 Understanding record: utterance (meeting room setting, first meeting room), utterance (time setting) , 2:00), noun (min, 30 min).

【0019】これらの対話文脈候補(9),(10),
(11)は規則適用部220、優先度計算部230を経
て次のようになる。 (12)優先度:20 理解記録:発話(会議室設定,第一会議室),発話(時
間設定,2時),名詞(分,30分) (13)優先度:20 理解記録:発話(会議室設定,第一会議室),名詞(時
間,2時30分) (14)優先度:30 理解記録:発話(会議室設定,第一会議室),発話(時間
設定,2時30分)ここで、対話文脈候補(12)は、対
話文脈候補(11)に何も規則を適用しなかったもので
ある。対話文脈候補(13)は、対話文脈候補(9)に
図5の規則(C)を適用したものであり、対話文脈候補
(14)は、対話文脈候補(11)に、図5の規則
(B)を適用したものである。これらの結果、対話文脈
候補(14)が対話文脈管理部21より応答計画装置に
送られる。
These dialogue context candidates (9), (10),
(11) is as follows via the rule application unit 220 and the priority calculation unit 230. (12) Priority: 20 Understanding record: utterance (meeting room setting, first meeting room), utterance (time setting, 2:00), noun (minute, 30 minutes) (13) Priority: 20 Understanding record: utterance ( Meeting room setting, first meeting room), noun (time, 2:30) (14) Priority: 30 Understanding record: utterance (meeting room setting, first meeting room), utterance (time setting, 2:30) Here, the dialogue context candidate (12) is one in which no rule is applied to the dialogue context candidate (11). The conversation context candidate (13) is obtained by applying the rule (C) of FIG. 5 to the conversation context candidate (9), and the conversation context candidate (14) is applied to the conversation context candidate (11) of the rule (FIG. 5 ) . B) is applied. As a result, the dialog context candidate (14) is sent from the dialog context management unit 21 to the response planning device.

【0020】即ち、「2時」が入力されたあとは、時間
設定が「2時」である解釈が一番優先度が高いが、「3
0分」が入力されたあとは、時間設定が「2時30分」
である解釈が一番優先度が高くなる。このように逐次的
に優先度の高い解釈が変り得る。このようにして、各時
点で最も優先度の高い文脈が得られる。
That is, after "2 o'clock" is input, the interpretation that the time setting is "2 o'clock" has the highest priority, but "3 o'clock"
After "0 minutes" is entered, the time setting is "2:30"
Interpretation has the highest priority. In this way, the interpretation with high priority may change sequentially. In this way, the highest priority context is obtained at each point in time.

【0021】[0021]

【発明の効果】以上説明したように、本発明の言語理解
方法及び装置を用いることにより、発話の区切りが明確
ではない音声の理解を実時間で行うことができる。これ
により、自由なタイミングでユーザが話しても、応答が
必要な場合には、即座に応答ができる音声対話システム
等の作成が可能になる。
As described above, by using the language understanding method and apparatus of the present invention, it is possible to understand in real time a voice in which speech boundaries are not clear. As a result, even if the user speaks at any time, when a response is required, it is possible to create a voice dialogue system or the like that can immediately respond.

【図面の簡単な説明】[Brief description of drawings]

【図1】本発明の言語理解方法の一実施例の処理フロー
図である。
FIG. 1 is a processing flow chart of an embodiment of a language understanding method of the present invention.

【図2】本発明の言語理解装置の一実施例の構成図であ
る。
FIG. 2 is a configuration diagram of an embodiment of a language understanding device of the present invention.

【図3】本発明の言語理解装置の適用例のシステム構成
図である。
FIG. 3 is a system configuration diagram of an application example of the language understanding device of the present invention.

【図4】辞書の一例を示す図である。FIG. 4 is a diagram showing an example of a dictionary.

【図5】言語理解規則の一例を示す図である。FIG. 5 is a diagram showing an example of a language understanding rule.

【符号の説明】[Explanation of symbols]

210 対話文脈管理部 220 規則適用部 230 優先度計算部 240 対話文脈候補保持部 250 辞書 260 言語理解規則 210 Dialog Context Management Unit 220 Rule Application Department 230 Priority calculator 240 Dialog context candidate holding unit 250 dictionary 260 Language Understanding Rules

フロントページの続き (56)参考文献 特開 平8−263092(JP,A) 特開 平7−191687(JP,A) 特開 平8−146990(JP,A) 小暮, 島津, 中野,対話における 時間依存プラン認識に向けて,電子情報 通信学会技術研究報告[言語理解とコミ ュニケーション],日本,1995年12月15 日,NLC95−64,pages 39−46 (58)調査した分野(Int.Cl.7,DB名) G10L 15/18 G06F 17/27 - 17/28 JICSTファイル(JOIS)Continuation of front page (56) References JP-A-8-263092 (JP, A) JP-A-7-1916987 (JP, A) JP-A-8-146990 (JP, A) Kogure, Shimazu, Nakano, in dialogue Toward time-dependent plan recognition, IEICE Technical Report [Language Understanding and Communication], Japan, December 15, 1995, NLC95-64, pages 39-46 (58) Fields investigated (Int.Cl . 7 , DB name) G10L 15/18 G06F 17/27-17/28 JISST file (JOIS)

Claims (2)

(57)【特許請求の範囲】(57) [Claims] 【請求項1】 発語の言語理解を行い、逐次的に対話文
脈候補を出力する言語理解方法であって、 対話文脈候補を優先度と対話の理解の状態を記録した理
解記録とで構成し、前の時点までに得られている対話文
脈候補を保持しておき、 新たに単語が入力されたときに辞書を参照して、該単語
の意味記述を求めるステップと、 前記単語が入力された時点までに得られている各対話文
脈候補の理解記録に前記単語の意味記述を追加してそれ
ぞれ新たな対話文脈候補を得るステップと、 前記新たに得られた各対話文脈候補の理解記録に言語理
解規則を適用し、該適用に成功した場合に、該適用結果
を理解記録に反映させて更に新たな対話文脈候補を得る
ステップと、 前記言語理解規則の適用に成功して得られた対話文脈候
補と適用前の各対話文脈候補の各々の優先度を、適用さ
れた言語理解規則と適用される前の理解記録とに依存し
た値に変えるステップと、 前記言語理解規則の適用に成功して得られた対話文脈候
補と適用前の各対話文脈候補のうちで、優先度の最も高
い対話文脈候補を出力するステップと、を単語が入力さ
れる度に繰り返すことを特徴とする言語理解方法。
1. A language is used for understanding the language of spoken words, and a dialogue sentence is sequentially provided.
A language understanding method for outputting a pulse candidate physical recording the state of understanding of the priorities and dialogue interaction context candidate
Dialogue sentences composed up to the previous time, composed of solution records
The pulse candidate is held, and when a new word is input, the dictionary is referred to and the word
The step of obtaining the meaning description of each of the dialogue sentences obtained by the time the word is input.
Add the meaning description of the word to the understanding record of the pulse candidate
Each step of obtaining a new dialogue context candidate and the linguistic logic in the understanding record of each newly obtained dialogue context candidate.
If the solution rule is applied and the application is successful, the application result
Is reflected in the understanding record to obtain more new dialogue context candidates
Steps and dialogue contexts obtained by successfully applying the language understanding rules
The priority of each dialogue context candidate before and after application is applied.
Language understanding rules and records of understanding before being applied.
And the dialogue context sign obtained by successfully applying the language understanding rule.
Of the dialogue context candidates before and after application, the highest priority
The step of outputting the candidate dialogue context,
A language comprehension method characterized by repeating each time.
【請求項2】 発語の言語理解を行い、逐次的に対話文
脈候補を出力する言語理解装置であって、 対話文脈候補を優先度と対話の理解の状態を記録した理
解記録とで構成し、前の時点までに得られている対話文
脈候補を保持する手段と、 新たに単語が入力されたときに辞書を参照して、該単語
の意味記述を求め、前記単語が入力された時点までに得
られている各対話文脈候補の理解記録に前記単語の意味
記述を追加してそれぞれ新たな対話文脈候補を得る手段
と、 前記新たに得られた各対話文脈候補の理解記録に言語理
解規則を適用し、該適用に成功した場合に、該適用結果
を理解記録に反映させて更に新たな対話文脈候 補を得る
手段と、 前記言語理解規則の適用に成功して得られた対話文脈候
補と適用前の各対話文脈候補の各々の優先度を、適用さ
れた言語理解規則と適用される前の理解記録とに依存し
た値に変える手段と、 前記言語理解規則の適用に成功して得られた対話文脈候
補と適用前の各対話文脈候補のうちで、優先度の最も高
い対話文脈候補を出力する手段とを具備し、各手段の処
理を単語が入力される度に繰り返すことを特徴とする言
語理解装置。
2. The language of the utterance is understood, and the dialogue sentence is sequentially obtained.
A language understanding device for outputting a pulse candidate physical recording the state of understanding of the priorities and dialogue interaction context candidate
Dialogue sentences composed up to the previous time, composed of solution records
A means for holding a pulse candidate, and referring to a dictionary when a new word is input, the word
Sought the meaning description of and obtained by the time the word was entered
Meaning of the word in the understanding record of each dialogue context candidate
Means to add new description and get new dialogue context candidates
And the linguistic theory in the understanding record of each newly obtained dialogue context candidate.
If the solution rule is applied and the application is successful, the application result
The to reflect the understanding recording obtain further new dialogue context candidate
Means and dialogue contexts obtained by successfully applying the language understanding rules
The priority of each dialogue context candidate before and after application is applied.
Language understanding rules and records of understanding before being applied.
Means for changing the value to a different value and the dialogue context indication obtained by successfully applying the language understanding rule.
Of the dialogue context candidates before and after application, the highest priority
And a means for outputting candidate dialogue contexts.
A word characterized by repeating the process every time a word is input
Word understanding device.
JP03783798A 1998-02-19 1998-02-19 Language understanding method and language understanding device Expired - Fee Related JP3525999B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP03783798A JP3525999B2 (en) 1998-02-19 1998-02-19 Language understanding method and language understanding device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP03783798A JP3525999B2 (en) 1998-02-19 1998-02-19 Language understanding method and language understanding device

Publications (2)

Publication Number Publication Date
JPH11237894A JPH11237894A (en) 1999-08-31
JP3525999B2 true JP3525999B2 (en) 2004-05-10

Family

ID=12508653

Family Applications (1)

Application Number Title Priority Date Filing Date
JP03783798A Expired - Fee Related JP3525999B2 (en) 1998-02-19 1998-02-19 Language understanding method and language understanding device

Country Status (1)

Country Link
JP (1) JP3525999B2 (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006171096A (en) * 2004-12-13 2006-06-29 Ntt Docomo Inc Continuous input speech recognition device and continuous input speech recognizing method
WO2012023450A1 (en) * 2010-08-19 2012-02-23 日本電気株式会社 Text processing system, text processing method, and text processing program
CN106057200A (en) * 2016-06-23 2016-10-26 广州亿程交通信息有限公司 Semantic-based interaction system and interaction method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
小暮, 島津, 中野,対話における時間依存プラン認識に向けて,電子情報通信学会技術研究報告[言語理解とコミュニケーション],日本,1995年12月15日,NLC95−64,pages 39−46

Also Published As

Publication number Publication date
JPH11237894A (en) 1999-08-31

Similar Documents

Publication Publication Date Title
US7949536B2 (en) Intelligent speech recognition of incomplete phrases
EP1089193A2 (en) Translating apparatus and method, and recording medium used therewith
JPH03224055A (en) Method and device for input of translation text
US20020007275A1 (en) Speech complementing apparatus, method and recording medium
JP4729902B2 (en) Spoken dialogue system
US6763331B2 (en) Sentence recognition apparatus, sentence recognition method, program, and medium
JPH08106298A (en) Device and method for conversation processing
JP2002116796A (en) Voice processor and method for voice processing and storage medium
JP4634889B2 (en) Voice dialogue scenario creation method, apparatus, voice dialogue scenario creation program, recording medium
CN114360504A (en) Audio processing method, device, equipment, program product and storage medium
Fan et al. Just speak it: Minimize cognitive load for eyes-free text editing with a smart voice assistant
JP2000172293A (en) Device and method for recognizing continuous speech
JP3525999B2 (en) Language understanding method and language understanding device
WO2003017251A1 (en) Prosodic boundary markup mechanism
JP4220151B2 (en) Spoken dialogue device
JPH07261793A (en) Interactive processor
JP3162896B2 (en) Topic structure recognition method and apparatus
JP3691773B2 (en) Sentence analysis method and sentence analysis apparatus capable of using the method
JP7028203B2 (en) Speech recognition device, speech recognition method, program
CN113515586A (en) Data processing method and device
US20020143525A1 (en) Method of decoding telegraphic speech
CN111914560B (en) Text inclusion relation recognition method, device, equipment and storage medium
JP2001013992A (en) Voice understanding device
JP2867695B2 (en) Continuous speech recognition device
Bennett et al. Using acoustic models to choose pronunciation variations for synthetic voices.

Legal Events

Date Code Title Description
TRDD Decision of grant or rejection written
A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

Effective date: 20040210

A61 First payment of annual fees (during grant procedure)

Free format text: JAPANESE INTERMEDIATE CODE: A61

Effective date: 20040210

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20080227

Year of fee payment: 4

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20090227

Year of fee payment: 5

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20090227

Year of fee payment: 5

FPAY Renewal fee payment (event date is renewal date of database)

Free format text: PAYMENT UNTIL: 20100227

Year of fee payment: 6

LAPS Cancellation because of no payment of annual fees