JP2967688B2 - Continuous word speech recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a grammar definition and grammar display user interface of a continuous word speech recognition apparatus.

[0002]

2. Description of the Related Art In a discrete word speech recognition apparatus, in order to inform a user what kind of speech input the apparatus can accept, it is common to list words that can be spoken and display them on a display or the like as a word list. It is a target.

On the other hand, in a continuous word speech recognition apparatus, words can be inputted by one utterance as sentences connected according to rules. Regarding continuous word speech recognition, a document "2-stage DP matching for efficiently recognizing continuously uttered word speech, Nikkei Electronics, November 7, 1983"
Familiar with.

[0004]

However, in the above-described continuous word speech recognition apparatus, it is difficult to list all possible sentences in order to inform the user of the input that the apparatus can accept. That is, the number of combinations of words that can be accepted by the rules is generally huge, making it difficult to enumerate all of them, and a very unfriendly interface for the user. Therefore, conventionally, a method of not displaying the message at all, displaying only a limited number of sentences among the sentences that can be accepted according to the rule as an example, or displaying a separately created example sentence is used. Was.

As a method of defining a sentence that can be input in the continuous word speech recognition apparatus, a definition has been created using a grammatical rule such as a rewriting rule, and the definition has been compiled and validated. However, concepts such as grammar rules and their compilation are extremely esoteric interfaces for developers, making it difficult for ordinary users of recognizers to add words and sentence patterns.

According to the present invention, in a continuous word speech recognition apparatus, in order to define and change a sentence that can be accepted by the apparatus, a grammar is created in accordance with a grammar rule (rewriting rule), and the grammar is compiled and reflected on the apparatus. We propose a new continuous word speech recognition device that solves the problems that ordinary users needed difficult procedures. Also, the present invention
We propose a grammar display unit that informs the user of acceptable sentence types with a small display area.

[0007]

[Means for Solving the Problems]

(1) As a grammar definition part that defines an acceptable word string,
A column number for identifying a column in a word string is a first dimension, and a word number for identifying a word in each column of the word string is a second dimension.
A continuation word characterized by having a grammar definition part which can develop and display a grammar in a dimensional table format, define and edit a word specified by a row and a column of the table, and define and edit the grammar. Voice recognition device. (2) The continuous word speech recognition device according to claim 1, wherein the grammar definition section of the continuous word speech recognition device defines a name in each column of the table. (3) In the continuous word speech recognition device according to claim 2, as a grammar display unit for notifying a speaker of an acceptable word string,
A continuous word speech recognition apparatus, comprising: a grammar display unit that defines a grammar in a table format and displays the names of the columns in the table separated by delimiters and connected and displayed. (4) In the grammar definition part of the continuous word speech recognition apparatus according to claim 1, the meaning is defined for each word in the table, and when the input speech is recognized as a word string, the meaning of each recognized word is ordered. And outputting the result as a recognition result of the input speech.

[0008]

1 shows an embodiment of a continuous word speech recognition apparatus according to the present invention. The speech recognition unit 20 uses the grammar information 30 to recognize speech input from the microphone 10 and output a recognition result. The grammar information 30 includes word connection information such as which words can be connected to which words, and attribute information (notation, reading, meaning, etc.) of each word. The grammar definition unit 40 provides a user interface for defining and editing the grammar information 30. The grammar display unit 50 provides a user interface for notifying the user of the device of the types of sentences that the device can accept.

FIG. 2 shows an example of a definition of an acceptable sentence by a grammar definition part of the conventional method. In order to accept a date such as "September 15", it is defined in a rewrite rule format using a non-terminal symbol enclosed in parentheses and a terminal symbol enclosed in quotation marks (Example 1). The same applies to a grammar for receiving an input such as "From Tokyo to Kyoto" (Example 2).

According to the present invention, in order to define grammatical information of a sentence that can be accepted by the continuous word speech recognition apparatus, a column number for identifying a column in a word string is assigned a first dimension and a word is assigned to each word in each word string. A two-dimensional tabular user interface with the word number to be identified as the second dimension is used. A word is defined at a position (called a cell) specified by a row number and a column number in a table. A sentence formed by selecting words one by one from each column of the table and connecting the words in the order of the columns is a sentence that can be input by the user. By adopting the table format, the grammar can be visually grasped, and even a user who is not familiar with the rewriting rules can easily define and change the grammar.

FIG. 3 shows an example of a table used in the grammar definition section of the present invention. In the grammar definition section, a table as shown in FIG. 3 is displayed on the screen, and the user selects a word on the table, defines and edits the attribute of the word, and moves, copies, and moves the word on the table. And can be deleted. Attributes of words include the “notation” of words used for display on the screen,
There are "reading" of a word, "meaning" of a word used as a recognition result, and the like.

[0012] In the case of defining by a rewriting rule as in the conventional method, it is generally difficult to grasp an acceptable sentence at a glance, and it is impossible for a non-expert to add or change the rule. On the other hand, when the definition is made in the form of a table as in the present invention, even non-experts can easily understand the sentence which can be visually recognized at a glance, and the definition can be easily extended and changed.

FIG. 4 shows an example of a display of an acceptable sentence by a grammar display unit of the conventional method. Lists all acceptable statements. Although it depends on the size of the display screen, it is difficult to display all the sentences, and there is no choice but to limit the number to a certain number or prepare a separate example sentence and display it. In general, it is often difficult to infer an acceptable grammar from only sentence examples.

In the grammar definition section of the present invention, a name (column name) representing a word in the column can be defined for each column of the table. Utilizing this definition result, the grammar display unit of the present invention connects the column names of each column of the table when displaying a sentence that can be accepted by the apparatus, and forms <column name 1 | column name 2 | column name 3 It is displayed by separating with symbols like>. If the column name is not defined, the word notation in the first row of the table can be used instead. By inserting a delimiter, it can be expressed that each column has a plurality of options. FIG. 5 shows a display example of an acceptable sentence by the grammar display unit of the present invention. Example 1 in FIG.
<What | month | what | day> in accordance with the grammar of Example 2, and <from departure station | to | destination station |> in accordance with the grammar of Example 2. With this display, tabular grammars that can be received with only one line of display can be listed, and a combinatorial explosion of sentence examples to be displayed can be avoided.

As the recognition result of the continuous word recognition, the meaning information of the word selected as the recognition result from each column of the table is sequentially connected and output. Examples of the semantic information include a command to a controlled device that operates upon receiving the output of the speech recognition device. In the case of the continuous word speech recognition device, a plurality of commands can be sent to the controlled device continuously by one utterance.

[0016]

As described above, according to the present invention, compared with the conventional method, a non-expert has a user interface capable of easily grasping and defining / changing the grammar in a table format easily and visually, and has a small number. It has become possible to realize a continuous word speech recognition device that can effectively display the types of sentences that can be accepted by the device in the display area.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention.

FIG. 2 is an example of a grammar definition by a grammar definition unit of a conventional method.

FIG. 3 is a grammar definition example by a grammar definition unit according to the present invention.

FIG. 4 is a display example of a sentence that can be accepted by a grammar display unit according to a conventional method.

FIG. 5 is a display example of an acceptable sentence by a grammar display unit according to the present invention.

[Explanation of symbols]

 Reference Signs List 10 microphone 20 voice recognition unit 30 grammar information 40 grammar definition unit 50 grammar display unit

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-62-293472 (JP, A) JP-A-4-289899 (JP, A) JP-A-6-95828 (JP, A) Papers of the Institute of Electronics, Information and Communication Engineers Magazine Vol. 75 -D-IIII, No. 6, June 1992, “Continuous Speech Recognition Using Phoneme Context-Dependent Model and High-Speed Search Method”, p. 1023-1030 (58) Fields investigated (Int. Cl. ⁶ , DB name) G10L 3/00 537 G10L 3/00 571 G06F 17/22 G06F 17/27 JICST file (JOIS) Practical file (PATOLIS) Patent file (PATOLIS)

Claims (4)

(57) [Claims] In a continuous word speech recognition device, a grammar definition unit for defining an acceptable word string is a column number for identifying a column in a word string in a first dimension and a word in each column of the word string. The grammar can be defined and edited by developing and displaying the grammar in a two-dimensional table format with the word number to be identified as the second dimension, and defining and editing the words specified by the rows and columns of the table. A continuous word speech recognition device having a grammar definition unit. 2. The continuous word speech recognition apparatus according to claim 1, wherein the grammar definition section of the continuous word speech recognition apparatus defines a name in each column of the table. 3. The continuous word speech recognition device according to claim 2, wherein the grammar is defined in a table format as a grammar display unit for notifying a speaker of an acceptable word sequence, and the name of each column in the table is represented by a delimiter. A continuous word speech recognition device having a grammar display unit for displaying divided and connected. 4. The grammar definition section of the continuous word speech recognition apparatus according to claim 1, wherein the meaning of each word in the table is defined, and when the input speech is recognized as a word string, the meaning of each recognized word is determined. Are sequentially connected and output as a recognition result of the input speech.