JPS6258329A - Specialist system for voice question and answer - Google Patents

Abstract

PURPOSE:To confirm the information given from a specialist system with no reference to a terminal screen by adding a function to perform the voice input and output of the character information and therefore supplying the information to the system with no contact secured with a keyboard. CONSTITUTION:A specialist system comprises a keyboard input device 5, a CRT screen display device 8, a speech recognizing device 6 which recognizes and codes voices and a speech synthesizing device 7 serving as a terminal equipment which delivers an output code in voices. Thus the voice input/output is possible with this system. Thus, even a beginner can perform simultaneously plural jobs requiring hands and feet without having any troublesome operation for questions and answers.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention relates to an expert system that makes decisions similar to those of experts by transplanting the knowledge and judgments of experts to a computer. , a voice question answering expert system that uses voice as a means of transmitting information during question answering between the expert system and system users.

(Prior Art) Conventional expert systems serve as a means of transmitting information between the system and system users, when the system users input information into the system.

Keyboards, light pens, and mice were used, and when the system outputted information to system users, it used terminal screen displays and teletypewriter printing.

(Problems to be Solved by the Invention) In this conventional expert system, due to the nature of the device used as the above-mentioned information input/output means, the system user has to interact with the system from beginning to end using hands and eyes. must be carried out. However, in reality, when we consider cases such as diagnosing machine failures using measuring instruments,
Situations such as being unable to take one's eyes off the machine or leaving the measuring instrument unattended (Means for solving the problem) The present invention eliminates the drawbacks of the prior art described above, and is capable of constantly communicating with expert systems. The purpose is to enable mutual information transfer even when system users are separated from the system. Therefore, in the expert system of the present invention, a voice recognition device and a voice synthesis device are added, and in addition to the conventional keyboard input and terminal screen display, the expert system of the present invention has the ability to directly input and output character information by voice. You can provide information to the system without touching the terminal or looking at the terminal screen.
You can check information from the system.

(Example) An example of the present invention will be described with reference to the drawings. 1st
The figure shows an overview of copying machine failure diagnosis as one embodiment of the voice question answering expert system of the present invention.

In FIG. 1, 1 is a reasoning device, 2 is a working memory device, and 3 is a reasoning device.
4 is a knowledge storage device, 4 is a knowledge management device, 5 is a keyboard input device, 6 is a speech recognition device, 7 is a speech synthesis device, 8 is a screen display device, and 9 is an input confirmation device.

The inference device 1 controls the entire system.

The details of the inference device 1 are shown in FIG. 2. The inference device 1 is composed of a comparator 1a, a rule selector ib, and an executor IC1 stopper 1d. The comparator 1a compares the data in the knowledge storage device 3 and the working memory device 2, and outputs the rule that matches the condition. If the number of rules that match the condition is zero, a stop signal is sent to the stop device 1d. , the inference device 1 stops. The rule selector 1b counts the frequency of matching conditions when comparing among a plurality of rules with matching conditions,
The rule with the largest number is selected as the most likely rule. However, if the number of selected rules is multiple, the rule that has been selected the most times in the past will be selected. One younger rule is finally selected.

The execution part of the selected rule is sent to the executor 12 and executed. The execution results are stored in the working memory device 2,
The working memory device 2, which activates the comparator 1a again, temporarily stores various data in the process of inference using the knowledge of the expert system. The knowledge storage device 3 stores the expert's knowledge in the form of rules in the form of character strings containing the failure situation and the cause of the failure as shown in Figure 2, and the failure situation as shown in Figure 3. For example, the defect status of an image is stored as a set of character strings, image data, and audio data. The knowledge management device 4 performs editing operations such as reading, adding, and deleting knowledge of experts, and checks whether there are any inconsistencies in the stored knowledge. The keyboard input device 5 and the image display device 8 are standard input/output devices provided in a normal computer system, and in the present invention, a voice recognition device 6 and a voice synthesis device 7 are newly added to these devices, so that information can also be transmitted by voice. It enables input/output.

The configuration of the speech recognition device is shown in FIG. The speech recognition device 6 includes a microphone 4a, a feature extraction section 4b, and an A/D conversion section 4.
C1 input cover storage section 4d.

It is composed of a standard button storage section 4e and a button matching section 4f. The audio input to the microphone 4a is converted into an electrical signal and has a frequency of 200 to 6000H2.
The feature extractor 4b, which is composed of a group of band-pass filters divided into 30 channels, extracts features such as power signals, formant signals, and the number of zero crossings, and then
Each e c is sampled and quantized by the A/D converter 4C and stored in the input button storage section 4d. The baton matching section 4e matches the data in the input baton storage section with the data in the standard baton storage section 4f prepared in advance, and temporarily sends the most probable one as a recognition result to the input confirmation device 9. Sent.

In order to confirm the input information, the input confirmation device 9 displays the information on the screen display device 8, and outputs the information as a voice by the speech synthesis device 7, which will be described next. On the other hand, if it is incorrectly recognized and input, it will be corrected by re-occurring within 5 seconds. Also,
It is also possible to input again from the keyboard input device 5. Otherwise, it is determined that the input information is correct, and the input information is sent to the inference device l.

Figure 5 shows a simple block diagram of a speech synthesis device.6 There are various speech synthesis methods, such as analysis and synthesis methods, but the system of the present invention can be applied to speech synthesis devices of any speech synthesis method. In the present invention, it is possible to
A speech synthesizer 5 using a versatile regular speech synthesis method is installed. This rule speech synthesis device 5 includes a speech output string storage section 5a, a monosyllabic string conversion section 5b, a monosyllabic speech data storage section 5c, a monosyllabic parameter conversion section 5d, a speech parameter generation section 5e, and a D/A converter. 5f, and a speaker 5g. The character string data sent from the inference device 1 or the input confirmation device 9 that you want to output as voice is
It is stored in the audio output character string storage section 5a, and decomposed into monosyllables by the monosyllabic character string conversion section 5b. On the other hand, the monosyllabic speech data storage section 5c stores speech parameters for each monosyllable in advance, and the speech parameters for each previously decomposed monosyllable are transferred to the monosyllabic speech parameter conversion section 5c.
Extracted by d. The speech parameters of each monosyllable extracted here are determined by the speech parameter generation unit 5e, which determines the intermittent time, power, pitch value, etc. of each monosyllable, and performs articulatory combination between monosyllables. Generate a set of audio parameters. These voice parameters are D/A converted by the D/A converter 5f, and the voice is outputted through the speaker 5g.The voice synthesizer 5 also processes pauses and accents, so more natural vocalization is realized. ing.

'In the present invention, by using the voice recognition device 6 and the voice synthesis device 7, questions about the failure status and responses to the questions can be performed entirely by voice.

The operation of the embodiment configured as described above will be explained.

FIG. 6 shows a specific example of the input/output of the fault diagnosis expert system including voice question and answering. When the system starts, rSTAR is stored in the working memory 2.
The letters TJ are stored as shown in 21. The inference device 1 executes the execution part to match the condition part of the rule 01 shown in the knowledge storage device 3, and the question "Is the image thin?" is asked as shown on the screen 81 of the screen display device 8. At the same time, using this character string as a key, the image data and audio character string data "Gazo LILIUS" shown in FIG. 3 are generated. '' is accessed, the image data, characters and images are output to the image display device 8 as shown in screen 81, and the voice character string data is sent to the voice synthesizer 7 as shown in 71 and converted into voice. Conventionally, characters and images were only displayed on the screen display device 8, so the system user had to view or read them from a close distance of the single-screen display device 8.
By outputting audio as in the present invention, the same information as the character string from the system displayed on the screen display device 8 can be obtained even when the user is away from the screen display device 8 or even without looking at the screen. In response to a question from the system, the system user can manually press "Yes" from the keyboard input device 5, or by saying "Yes" even from a distance from the keyboard input device 5, and the voice recognition device 6 can respond by saying "Yes". °°
If this is recognized, the contents of the 1 working memory device 2 will become "image pale" as shown at 22. Furthermore, the reasoning device 1 compares the condition part of the rule 02 shown in FIG. 2 in the knowledge storage device 3 with the content 22 in the working memory device 2, and selects the rule 02,
By executing the execution part, the cause of the failure "the wire position of the charger is low" as shown in 82 is displayed on the single screen display device 8, and the voice synthesizer 7 displays "TIDEN" The audio output is "hiku". The inference device 1 compares the condition part of the rule set shown in FIG. 2 in the knowledge storage device 3 with the contents of the working memory device 2, and if there is no match, the inference device 1 stops.

(Effects of the Invention) As is clear from the above description, the present invention uses voice information in addition to text information and image information as a means of transmitting information between the expert system and the system user. It is characterized by Information input speed by voice input is
We were able to confirm the effectiveness of voice input faster than keyboard input.For example, when inputting "Hi", approx.
I recognized it in 3 seconds. By enabling voice input, even beginners who are not very familiar with using computers can perform other tasks that use their hands and feet at the same time without having to perform tedious operations to answer questions. In addition, even for those who are accustomed to using computers, there is no need to look at one screen each time to confirm the input/output results of information, so the eyes do not get tired. Another advantage is that even at a location away from the system, it is possible to answer questions with the system without relying on keyboard input or screen display.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing an overview of the voice question answering expert system of the present invention, Fig. 2 is a diagram showing an example of rule knowledge stored in the knowledge storage device, and Fig. 3 is a diagram showing an example of rule knowledge stored in the knowledge storage device. A diagram explaining image data and audio data as stored output data from the system, Figure 4 is a diagram showing a speech recognition device, and Figure $5 is an example of speech synthesis using a regular speech synthesis method. FIG. 6 is a diagram showing the speech synthesis device; FIG. 6 is a diagram explaining the flow of diagnosis by the voice question answering expert system of the present invention, taking copying machine failure diagnosis as an example; FIG. 7 is a diagram explaining the configuration of the reasoning device. It is a diagram. 1-11 Reasoning device, 2--1 Working memory device 3
--- Knowledge storage device, 4 --- Knowledge management device 5 ---
--Keyboard input device

Claims (1)

[Claims] A knowledge storage device that stores the knowledge of experts; a knowledge management device that can read out that knowledge, add new information, delete it, and check the consistency of knowledge; an inference device that can derive facts, a working memory device that stores data from the inference process, a voice recognition device that recognizes the voice generated by the system user in real time, and a voice recognition device that confirms the voice input results. In an expert system that is equipped with an input confirmation device and a speech synthesis device for generating any text that the system conveys to the system user in voice, the system is used as an information transmission means for answering questions using voice. A voice question answering expert system characterized by: