JPH06133039A - Automatic guidance device for weather information - Google Patents

Abstract

PURPOSE:To obtain a weather information guidance device which prevents the rise of labor costs, improves the rapidity of service and fitted to the subdivision of a service area. CONSTITUTION:The continuous voice recognition part of a voice recognition part 3 recognizes natural talk in the sentence unit of a user. When the recognition of continuous voice is failed, a word voice recognition part recognizes voice in a word unit. An interaction management part 4 analyzes the meaning of the talk of the user, and generates a response sentence based on the talk. A voice synthesis part 5 synthesizes response voice, recognizes an important matter for the user and supplies information. Thus, weather information of a place which the user intends to know can directly be guided by using voice recognition technology, and the subdivision of the service area can be realized. Since latest weather information can be supplied through a data communication part 7, it is advantageous for shortening the time interval of information update, for making the device unmanned and for continuous operation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to automatic speech recognition technology,
The present invention relates to a weather information automatic guidance device in which text-to-speech synthesis technology is applied to an information service.

[0002]

2. Description of the Related Art Conventionally, as a weather information automatic guidance service, there is "weather forecast" of NTT No. 177. This "weather forecast" is based on the announcement of the Meteorological Agency forecast department.
The weather information is provided to the user by repeatedly playing back the tape on which the weather information such as the sky within a certain time period and the forecast of precipitation is recorded.

However, this service can only provide a wide range of weather information (usually one prefecture is divided into 1 to 5 areas to provide the service), and there is a limit to providing more detailed local weather information. There is.

Further, in the conventional method, since the weather information is blown into the recording tape by the human, the cost of the human operation cannot be ignored in order to subdivide the service area or shorten the service time interval. In addition, not only a time loss such as a reading error and an increase in the check work accompanied therewith, but also an increase in cost.

On the other hand, in order to subdivide the service area according to the conventional method, there is no choice but to identify the area by the area code, but the user always knows the area code of the place where the user wants to receive the service. Not necessarily. For example, the request "I want to know the weather in Tokyo Disneyland" cannot be answered. Further, the service area division of the area code depends on the government, and since there are many geographically irregular divisions, it is not necessarily suitable for the weather information service.

There is also a service for providing weather information to a specific target person such as a personal computer communication user or a user who has a contract with an information providing service company without using a telephone, but it is not suitable for the general public, so it is popular. It is hard to be done.

[0007]

As described above, the conventional weather information guidance system has the following problems. (A) Since human beings read out weather information and record it on recording tapes, it is inevitable that costs such as personnel expenses increase due to subdivision of service areas and shortening of service time intervals. (B) The increase in recording check work time due to a reading error hinders the speeding up of services. (C) The division of service areas by area code is not always convenient for users.

The present invention solves the above-mentioned problems by using a weather information automatic guiding apparatus, that is, preventing an increase in costs such as personnel costs,
It is an object of the present invention to provide a weather information guide device that improves the speed of service and is suitable for subdividing a service area.

[0009]

In order to solve the above-mentioned problems, the present invention comprises a continuous voice recognition unit for recognizing a user's natural utterance, a word voice recognition unit for recognizing word voice,
A dialogue management unit for generating an appropriate response sentence based on the voice recognition result by the continuous voice recognition unit or the word voice recognition unit;
A text-to-speech synthesis unit that converts a response sentence into voice and outputs it, a storage unit that stores weather information, a data communication unit that communicates with the outside to update the weather information, and controls each of these parts. It is characterized by having a control unit.

[0010]

According to the present invention, the continuous speech recognition unit recognizes the natural utterance of the user in sentence units, and when the continuous speech recognition fails, the word speech recognition unit performs word unit speech recognition. The dialogue management unit analyzes the meaning of the user's utterance and creates a response sentence based on it. The text-to-speech synthesizer synthesizes the response voice, confirms the message and provides the information to the user.

As described above, according to the present invention, by using the voice recognition technology, it is possible to directly guide the user to the weather information of the place that the user wants to know, and it is also suitable for subdividing the service area. In addition, by using the text-to-speech synthesis technology, the data sent from the weather information provider can be instantly converted into voice, which not only saves humans the trouble of recording but also eliminates recording mistakes. You can save the trouble of checking the contents.

Further, since the latest weather information can be instantaneously provided by using the data communication technique, it is advantageous for shortening the time interval for updating information, unmanning the apparatus, continuous operation and the like.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram showing a schematic structure of a weather information automatic guide device of an embodiment of the present invention. This weather information automatic guide device is, for example, an individual circuit such as an integrated circuit,
Alternatively, it is configured by program control using a processor and is connected to the outside by a telephone line 2 or the like,
A control unit 1 for controlling the entire apparatus is provided, and a voice recognition unit 3 for recognizing a continuous voice or a word voice uttered by a user and a voice synthesizing unit 5 for synthesizing a response sentence are connected to the control unit 1. Further, the control unit 1 includes a dialogue management unit 4 that analyzes the meaning of the user's utterance with respect to the output of the voice recognition unit 3 and generates an appropriate response sentence or question sentence, a voice standard pattern, a word dictionary, a grammar, and the like. A storage unit 6 that stores the data and a data communication unit 7 that communicates with the outside to capture the latest weather information are connected. The data communication unit 7 includes a communication network 8
Are connected.

FIG. 2 is a diagram showing a processing procedure of an embodiment of the present invention, FIG. 3 is a diagram showing a processing procedure when a continuous speech recognition technique is used in the processing procedure of FIG. 2, and FIG. Shows the processing procedure when the word voice recognition technique is used in the processing procedure of FIG. The meteorological information guide device of this embodiment will be described below with reference to these drawings.

In FIG. 2, when the information guidance is started (step 11), the voice recognition unit 3 recognizes the input user's utterance in sentence units (step 12). In the present embodiment, for example, reference 1 (Seiichi Nakagawa, “Speech recognition by probabilistic model”, published by The Institute of Electronics, Information and Communication Engineers, 1988 pp. 1)
36-137, 139-141). That is, continuous speech recognition is performed using a sentence recognition algorithm described by a probabilistic context-free grammar.

When the continuous speech recognition is not successful in step 13, the control unit 1 switches the speech recognition unit 3 to the word speech recognition, and for example, reference 2 (Sadahiro Furui, "Digital Speech Processing", Tokai University Press Publishing). , 1985, pp. 185.
The word voice recognition algorithm described in 188) is used to recognize the user's word voice (step 14). That is, a plurality of standard patterns are prepared for each word in order to correspond to the unspecified speaker, and the standard pattern closest to the word voice uttered by the user is used as the recognition result.

Based on the result of the continuous voice recognition or the result of the word voice recognition, the dialogue management unit 4 generates an appropriate response sentence (step 15), and the voice synthesis unit 5 uses, for example, Document 3 (Sadahiro Furui, "Digital Speech processing ", Tokai University Press, 1985, pp. 144-146) to perform text-to-speech synthesis (step 16). (1) Detection of word boundaries. (2) Conversion of words into phoneme symbol strings. (3) Addition of phoneme prosody information. (4) Generation of a voice synthesizer control signal according to a synthesis unit and a rule. Perform text-to-speech synthesis according to the procedure. In this way, the user is informed of the information and the operation ends (step 17).

FIG. 3 shows a processing procedure when the speech recognition unit 3 performs continuous speech recognition. When the process is started (step 101), the control unit 1 first sets the variable RL indicating the voice recognition situation and the variable NC indicating the number of recognitions to 0 (step 102), and waits for input. Here, since no voicing restriction is applied to the user, the user can naturally utter in sentence units. When an inquiry voice is received from the user (step 10
3), the voice recognition unit 3 recognizes the continuous utterance of the user using the algorithm described in Document 1 (step 104). The dialogue management unit 4 analyzes the meaning of the user's utterance based on the recognition result of the voice recognition unit 3, and extracts a message necessary for weather information guidance, such as a date or a place name.

Next, while collating the value of the variable RL, the process branches into three directions (step 106). First, RL =
In the case of 0, that is, the first utterance or when the message cannot be extracted from the previous recognition result, the control unit 1 confirms whether the message is extracted from the current recognition result (step 11).
5). If the message extraction is successful, it is checked whether the extracted message is sufficient (step 118). If there is no missing message, the dialogue management unit 4 generates a confirmation sentence as shown in (a) of Table 1, and recognizes the recognition status variable RL.
Is set to 2 (step 120), and the voice synthesizer 5
In step 121, a text-to-speech synthesis is performed to announce to the user, and the process returns to step 103 to wait for the user's next utterance.

In step 118, if it is determined that the message is missing, the dialogue management unit 4 generates the confirmation & recurring speech prompting sentence shown in (b) of Table 1 and sets the value of the recognition status variable RL to 1
(Step 119), and text-to-speech synthesis is performed (step 121).

In step 115, when it is determined that the message is not extracted, 1 is added to the variable NC indicating the number of misrecognitions (step 108), and NC is a predetermined threshold value K.
It is confirmed whether or not it exceeds (step 112). N
If C does not exceed the threshold value, the dialogue management unit 4 generates a recurrence story promoting sentence shown in (c) of Table 1 and recognizes the recognition status variable RL.
Is set to 0 (step 116), and text-to-speech synthesis is performed (step 121). If the value of NC becomes larger than the threshold value K, it is switched to the word voice recognition under the control of the control unit 1, and the processing shifts to B in FIG.

In step 106, if RL = 1,
That is, if there is a missing message in the previous process, the control unit 1 confirms from the recognition result of this time whether or not the message is extracted (step 107), and if the message is not extracted, the step 108 is performed. Proceed to. When the message is extracted, the message extracted this time is compared with the message extracted last time, and it is checked whether or not they match (step 111). If there is no match, the process proceeds to step 108, and if there is a match, the process proceeds to step 118.

If RL = 2 in step 106, that is, if the message extraction is successful in the previous process, the dialogue management unit 4 analyzes the meaning of the user's reply to the confirmation sentence in Table 1 (a) in step 105. Then, it is checked whether or not the message confirmation is successful (step 110). Here, for example, if a positive answer such as “Yes, yes” is received from the user, the dialogue management unit 4 creates a response sentence as shown in Table 1 (d) according to the extracted message. (Step 11
4) Then, the voice synthesizing unit 5 voice-synthesizes the response sentence text (step 117), guides the weather information to the user, and then returns to A in FIG. 3 to wait for the next user to speak.

FIG. 4 is a flow chart showing the processing procedure of word voice recognition. In step 112 or step 113 of FIG. 3, when the variable NC indicating the number of erroneous recognitions exceeds a predetermined threshold value K, the control unit 1 stops continuous voice recognition and switches to word voice recognition.

When switching to word speech recognition, the control unit 1 first resets the value of the variable RL (step 20).
1), the guidance sentence as shown in (e) of Table 1 is reproduced (step 202). Next, the question sentence as shown in (f) of Table 1 is reproduced sentence by sentence (step 203), and the answer from the user is taken in (step 204). The voice recognition unit 3 recognizes the voice uttered by the user using the word voice recognition technique (step 205).

In the next step, the process branches into two branches according to the value of the variable RL (step 206). First, RL =
In the case of 0, as a result of word recognition, it is checked whether or not the message extraction is successful (step 208). If the message extraction fails, the process returns to step 203, the same question sentence is reproduced, and if the message extraction succeeds, the next step 210 is performed.
Check to see if the message is missing. If there is a gap, the process returns to step 203 and the next question sentence is reproduced. If there is no omission, a confirmation sentence as shown in (g) of Table 1 is generated and the value of the variable RL is set to 2 (step 21).
2). The voice synthesizing unit 5 synthesizes the previously generated confirmation sentence with the voice (step 213) and returns to step 204.

If the variable RL is 2 in step 206, it is checked whether or not the message confirmation to the user has succeeded according to the result of word recognition (step 207). If the user's answer is “yes”, go back to step 203,
Play the question from the beginning. User response is "high"
If it is, a response sentence as shown in (d) of Table 1 is generated (step 209), the response sentence text is voice-synthesized by the voice synthesizing unit 5 (step 211), and the weather information is presented to the user. Returning to FIG. 3A, the next user is waiting for a speech.

[0028]

[Table 1] ○ ○ = Place name, facility name, etc. △ △ = Today, tomorrow, Sunday, July 2nd ...

[0029]

According to the present invention, by using the voice recognition technology, it is possible to directly guide the weather information of the place that the user wants to know, so that it is possible to provide a user-friendly weather information guidance system. In addition, by using the text-to-speech synthesis technology, the data sent from the weather information provider can be instantly converted into voice, which not only saves the human recording but also eliminates recording mistakes. There is an advantage that you can. Furthermore, the latest weather information can be provided instantaneously by using the data communication technology. Therefore, according to the present invention, labor saving and cost reduction can be achieved, and it is also suitable for shortening the information update time interval, unmanned apparatus, continuous operation, and subdivision of service area. By changing a part of the system, not only the weather information but also other information can be guided.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a weather information guidance device according to an embodiment.

FIG. 2 is a diagram showing a processing procedure of an embodiment.

FIG. 3 is a diagram showing a processing procedure of continuous speech recognition.

FIG. 4 is a diagram showing a processing procedure of word voice recognition.

[Explanation of symbols]

 1 control unit 2 telephone line 3 voice recognition unit 4 dialogue management unit 5 voice synthesis unit 6 storage unit 7 data communication unit 8 communication network

Claims (1)

[Claims] 1. A continuous voice recognition unit that recognizes a user's natural utterance, a word voice recognition unit that recognizes a word voice, and an appropriate voice recognition unit based on the voice recognition result by the continuous voice recognition unit or the word voice recognition unit. A dialogue management unit that generates response text, a text-to-speech synthesis unit that converts the response text into voice and outputs it, a storage unit that stores weather information, and a data communication that communicates with the outside to update the weather information. An automatic weather information guide device comprising a control section and a control section for controlling each of these sections.