JP3683504B2 - Voice utilization type information retrieval apparatus, voice utilization type information retrieval program, and recording medium recording the program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a voice-use type information search device and a voice-use type information search program for searching for corresponding information based on a search keyword inputted by voice from a user terminal and outputting the information to the user terminal, and a recording medium storing the program. .
[0002]
[Prior art]
Among various services on the Internet, for example, a Web content search service is based on the input of text. Therefore, character input with a keyboard is indispensable.
[0003]
The search service is a service that outputs a URL of a document including the keyword, particularly a document including the keyword in a Web search when a keyword to be searched is input, and software that provides the service is referred to as a search engine. For details on the search service and search engine, refer to, for example, “InfoBee text information search technology, NTT R & D, Vol. 46, No. 10, 1997, pp. 93-98”.
[0004]
As a search service, a service called a voice portal in which an Internet search engine has a voice interface has been started. In this service, in order to perform speech recognition, it is necessary to register words and their readings in advance in the speech recognition dictionary. In the speech recognition system, the word having the highest degree of coincidence with the speech input to the system is selected from the words registered in the speech recognition dictionary, and this is output as a recognition result. Therefore, since all of the input speech and the words registered in the speech recognition dictionary are compared to determine whether or not they match, the recognition processing increases as the number of words registered in the dictionary increases. Takes a long time, increasing the probability of selecting the wrong word, and lowering the recognition accuracy.
[0005]
At the same time, in the voice portal, a search service for web contents on the Internet is provided with a voice interface. However, in the voice interface by the voice recognition system, a word that is not registered in the recognition dictionary is spoken by the user. Even it is not recognized.
[0006]
Therefore, in order to recognize as many words input by the search service as possible, it is necessary to register as many words as possible in the recognition dictionary and to sufficiently increase the word coverage. The word coverage is the number of words registered in the dictionary in the speech input to the system, and is ideally 100%.
[0007]
Conventionally, since both the recognition accuracy and the word coverage cannot be satisfied, the service target is limited and only the word coverage is realized. For example, when you call a phone number and say a place name, there is a service that announces the weather forecast for the area of this place name, but this service limits the input words to "place name of a place" All the place names in the area are registered in the recognition dictionary, and a sufficiently large coverage is achieved.
[0008]
However, for example, the number of words included in a certain 100,000 URL is in the order of hundreds of thousands to millions, and in a voice portal that searches more URLs, the words to be registered as in the above example Can not be limited.
[0009]
[Problems to be solved by the invention]
As described above, since the conventional search service cannot satisfy both the word coverage and the recognition accuracy, the service coverage is limited and only the word coverage within the limited range is increased. When many URLs are to be searched, it is not possible to limit the words to be registered, and there is a problem that the recognition accuracy decreases if an attempt is made to increase the word coverage.
[0010]
The present invention has been made in view of the above, and an object of the present invention is to use a ubiquitous search keyword and register a frequently used word in a recognition dictionary while suppressing a decrease in recognition accuracy. It is an object of the present invention to provide a voice-based information search device, a voice-based information search program, and a recording medium on which the program is recorded, which can realize a word coverage sufficient for a search service.
[0017]
[Means for Solving the Problems]
The present invention according to claim 1 is a voice-utilization type information retrieval device that retrieves relevant information based on a search keyword inputted by voice from a user terminal and outputs the information to the user terminal, which is inputted by voice from the user terminal. A speech recognition unit that receives a search keyword, recognizes the keyword by referring to a speech recognition dictionary, and outputs a keyword that is a text; a keyword that is a text recognized by the speech recognition unit; and an Internet terminal Search means for searching for information corresponding to the keyword using a keyword that is sent text as a search request, search result output means for outputting the searched information to a user terminal or an Internet terminal, and used for the search Search keyword storage means for storing keywords that are text, and the search key Keyword group creation means for calculating the relationship between keywords stored in the word storage means and creating a keyword group that summarizes the keywords with large relations, and the frequency of use of multiple keywords in each created keyword group Total use frequency calculation means for calculating the total of the above, and keyword group selection means for selecting a keyword group based on the calculated total use frequency and registering each keyword of the selected keyword group in the speech recognition dictionary This is the gist.
[0018]
In the present invention described in claim 1, a speech search keyword is received from the user terminal, the speech keyword is speech-recognized by referring to the speech recognition dictionary, and the speech-recognized text keyword and the Internet terminal are used. The keyword that is the text that is sent is retrieved as a search request, information corresponding to the keyword is retrieved and output to the user terminal or text terminal, the keyword that is the text used for the search is saved, and this saved Calculate the relationship between keywords, create a keyword group that summarizes highly related keywords, calculate the total frequency of use of multiple keywords in each keyword group, and determine the keyword group based on this total usage frequency. Select each keyword in this selected keyword group To register with the voice recognition dictionary, it is possible to improve the recognition accuracy while realizing coverage sufficient word search service.
[0019]
According to a second aspect of the present invention, there is provided a voice utilization type information retrieval apparatus for retrieving relevant information based on a search keyword inputted by voice from a user terminal and outputting the information to the user terminal. A speech recognition unit that receives an input search keyword, speech-recognizes the keyword with reference to a speech recognition dictionary, and outputs a keyword that is text, a keyword that is text recognized by the speech recognition unit, and the Internet A search means for searching for information corresponding to the keyword using a keyword as a text sent from the terminal as a search request, a search result output means for outputting the searched information to a user terminal or an Internet terminal, and used for the search Search keyword storage that saves keywords that are stored text along with the search time And a keyword stored within a predetermined period based on a search time stored in the search keyword storage unit, and calculating a relationship between the selected keywords to collect a large related keyword A keyword group creation means for creating a keyword group, a total usage frequency calculation means for calculating the total usage frequency of a plurality of keywords in each of the created keyword groups, and a keyword group based on the calculated total usage frequency The gist of the invention is to include keyword group selection means for selecting and registering each keyword of the selected keyword group in the speech recognition dictionary.
[0020]
In the present invention described in claim 2, a speech search keyword is received from the user terminal, the speech keyword is speech-recognized by referring to the speech recognition dictionary, and the speech-recognized text keyword and the Internet terminal are used. The keyword that is the text that is sent is retrieved as a search request, information corresponding to the keyword is retrieved and output to the user terminal or text terminal, and the keyword that is the text that was used for the search is stored, and together with this keyword Select keywords saved within a specified period based on the saved search time, calculate the relationship between the selected keywords, create a keyword group that summarizes the relevant keywords, and create a keyword group within each keyword group. To calculate the total usage frequency of multiple keywords Select a keyword group based on the frequency of use, and register each keyword in this keyword group in the speech recognition dictionary as an effective word for search, improving recognition accuracy while realizing sufficient word coverage for search services can do.
[0027]
According to a third aspect of the present invention, there is provided a speech utilization type information retrieval program that is executed by a speech utilization type information retrieval device that retrieves corresponding information based on a retrieval keyword inputted by speech from a user terminal and outputs the information to the user terminal. A voice recognition means for receiving a search keyword inputted by voice from a user terminal, recognizing the keyword by referring to a voice recognition dictionary, and outputting a keyword as text. Search means for searching for information corresponding to the keyword using the keyword that is the text recognized by the voice recognition means and the keyword that is the text sent from the Internet terminal as a search request, and the searched information Search result output means for outputting to a user terminal or Internet terminal, and used for the search Search keyword storage means for storing keywords that are stored text, and keyword group creation means for calculating a relationship between keywords stored in the search keyword storage means and creating a keyword group in which large related keywords are collected And a total usage frequency calculating means for calculating the total usage frequency of a plurality of keywords in each of the created keyword groups, selecting a keyword group based on the calculated total usage frequency, and selecting the keyword group The gist is to make each keyword function as keyword group selection means for registering in the speech recognition dictionary.
[0028]
In the present invention described in claim 3, a speech search keyword is received from the user terminal, the speech keyword is speech-recognized by referring to the speech recognition dictionary, and the speech-recognized text keyword and the Internet terminal are used. The keyword that is the text that is sent is retrieved as a search request, information corresponding to the keyword is retrieved and output to the user terminal or text terminal, the keyword that is the text used for the search is saved, and this saved Calculate the relationship between keywords, create a keyword group that summarizes highly related keywords, calculate the total frequency of use of multiple keywords in each keyword group, and determine the keyword group based on this total usage frequency. Select each keyword in this selected keyword group To register in the speech recognition dictionary as a valid word in the search, it is possible to improve the recognition accuracy while realizing coverage sufficient word search service.
[0029]
According to a fourth aspect of the present invention, there is provided a voice utilization type information retrieval program that is executed by a voice utilization type information retrieval device that retrieves corresponding information based on a search keyword inputted by voice from a user terminal and outputs the information to the user terminal. A speech recognition unit that receives a search keyword input by voice from the user terminal, recognizes the keyword by referring to a speech recognition dictionary, and outputs a keyword that is a text, and is speech-recognized by the speech recognition unit A search means for searching for a keyword that is text and a keyword that is text sent from an Internet terminal as a search request and searching for information corresponding to the keyword, and a search result output for outputting the searched information to a user terminal or an Internet terminal And keywords that are the text used in the search Search keyword storage means for storing together with the time, and selecting a keyword stored within a predetermined period based on the search time stored in the search keyword storage means, and calculating the relationship between the selected keywords, Keyword group creation means for creating a keyword group that summarizes related keywords, total usage frequency calculation means for calculating the total usage frequency of a plurality of keywords in each created keyword group, and the calculated total usage The gist is to select a keyword group based on the frequency and to function as keyword group selection means for registering each keyword of the selected keyword group in the speech recognition dictionary.
[0030]
In the present invention described in claim 4, a speech search keyword is received from the user terminal, the speech keyword is speech-recognized by referring to the speech recognition dictionary, and the speech-recognized text keyword and the Internet terminal are used. The keyword that is the text that is sent is retrieved as a search request, information corresponding to the keyword is retrieved and output to the user terminal or text terminal, and the keyword that is the text that was used for the search is stored, and together with this keyword Select keywords saved within a specified period based on the saved search time, calculate the relationship between the selected keywords, create a keyword group that summarizes the relevant keywords, and create a keyword group within each keyword group. To calculate the total usage frequency of multiple keywords Select a keyword group based on the frequency of use, and register each keyword in this keyword group in the speech recognition dictionary as an effective word for search, improving recognition accuracy while realizing sufficient word coverage for search services can do.
[0035]
Further, the gist of the present invention described in claim 5 is that the voice utilization type information retrieval program according to claim 3 is recorded on a computer-readable medium.
[0036]
In the present invention according to claim 5, a relationship between stored keywords is calculated to create a keyword group in which a large number of related keywords are collected, and the total use frequency of a plurality of keywords in each keyword group To select a keyword group based on the total usage frequency, and register each keyword of the selected keyword group in the speech recognition dictionary as a valid word for search. Since the program capable of improving the recognition accuracy while being realized is recorded on the recording medium, the distribution of the program can be improved by using the recording medium.
[0037]
The gist of the present invention described in claim 6 is that the voice utilization type information retrieval program according to claim 4 is recorded on a computer-readable medium.
[0038]
In the present invention according to claim 6, a keyword stored within a predetermined period is selected based on a search time stored together with the keyword, a relationship between the selected keywords is calculated, Create a keyword group that summarizes, calculate the total usage frequency of multiple keywords within each keyword group, select a keyword group based on this total usage frequency, and enable each keyword in this keyword group for search Since the program that can improve the recognition accuracy while realizing the word coverage sufficient for the search service is recorded on the recording medium because it is registered in the voice recognition dictionary as a simple word, the recording medium is used. , Can improve its distribution.
[0039]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a voice-based information search apparatus according to an embodiment of the present invention. The voice utilization type information retrieval apparatus of the embodiment shown in the figure is configured as a voice portal 10 that retrieves corresponding information based on a search keyword inputted by voice from the user terminal 1 and transmits it to the user terminal. 10 receives a speech recognition dictionary 5 that stores words used in speech recognition in correspondence with the reading thereof, and a search keyword inputted by speech from the user terminal 1 via the circuit switching network 21, and receives the speech recognition dictionary 5, a speech recognition unit 3 that is a speech recognition unit that recognizes a keyword by speech, and a search unit that searches for information corresponding to the keyword using the keyword recognized by the speech recognition unit 3 as a search request. Search engine 8 and a search result output means for transmitting the search result to the user terminal 1 or the Internet terminal 2 via the Internet 22 And a search log database (DB) 7 which is a search keyword storage means for storing a keyword used for a search in the search engine 8 and a search keyword effective from the stored keywords. A recognition dictionary registration unit 6 that is a registration unit that registers a word as a word in the speech recognition dictionary 5 and a transmission that transmits a search result to the user terminal 1 when the search result transmission unit 4 transmits the search result to the user terminal 1 It consists of a user profile database (DB) 9 which is output format storage means for storing the format.
[0040]
The user terminal 1 is, for example, a mobile phone terminal connected to the circuit switching network 21 and the Internet 22 and having a voice call function. The user terminal 1 transmits a search keyword by voice to the voice recognition unit 3 of the voice portal 10 and transmits a search result. A search result is received from the unit 4 via the Internet 22. Note that the terminal that transmits the search keyword and the terminal that receives the search result are not necessarily the same terminal, and may be different terminals. In the present embodiment, both functions are included in the same terminal. The case will be described.
[0041]
When using the search service, the Internet terminal 2 transmits a keyword input by a keyboard or the like to the search engine 8 via the Internet 22 and receives the search result from the search result transmitting unit 4. Yes. Note that this is a form of Internet service that is currently generally used.
[0042]
The voice recognition unit 3 receives a keyword based on voice spoken by the user from the user terminal 1, performs voice recognition processing with reference to the voice recognition dictionary 5, and supplies the keyword to the search engine 8 in a text format. For the speech recognition processing, see, for example, “Sound Communication Engineering, Nobuhiko Kitawaki, Corona, 1996”.
[0043]
The search result transmission unit 4 receives the search result from the search engine 8 and transmits it to the user terminal 1 and the Internet terminal 2. In the voice recognition dictionary 5, a dictionary used for voice recognition in the voice recognition unit 3 is registered. The recognition dictionary registration unit 6 receives from the search log database 7 the keywords used for the search, the search time, and the terminal ID, and registers the keywords that are frequently used and effective for the search in the speech recognition dictionary 5. To do. The search log database 7 stores keywords, search times, and terminal IDs input to the search engine 8 and used for the search in the order of input.
[0044]
When the keyword to be searched is input, the search engine 8 outputs the URL of the document including the keyword and the URL of the document including the keyword in the Web search. This is because a search keyword may be inputted by voice from the user terminal 1 by the voice portal service, or text information is directly designated and inputted from the Internet terminal 2 such as a PC without using the voice portal service. Also good.
[0045]
The user profile database 9 stores a transmission format for transmitting search results from the search engine 8 to the user terminal 1 or the Internet terminal 2.
[0046]
FIG. 2 is a diagram showing a data configuration stored in the user profile database 9. As shown in the figure, the user profile database 9 stores a transmission format and a transmission destination corresponding to each user's terminal ID as a transmission format for outputting the search result to the user. The terminal ID is information for identifying a terminal supplied from a public network, and is a caller number in a subscriber line or an ISDN line. It is assumed that registration from each user in the transmission format to the user profile database 9 has been performed in advance.
[0047]
Next, with reference to the flowchart shown in FIG. 3, the search process by the voice interface, that is, the search service providing process, which is the operation of the voice-use type information search apparatus configured as described above, will be described.
[0048]
In FIG. 3, when the user terminal 1 utters a search keyword by voice, the voice search keyword is supplied from the user terminal 1 to the voice recognition unit 3 via the circuit switching network 21, and also the terminal of the user terminal 1. An ID is also supplied (step S201). The speech recognition unit 3 recognizes the speech search keyword input from the user terminal 1 while referring to the speech recognition dictionary 5, and outputs text information of the recognition result (step S202).
[0049]
The text information of the recognition result output from the voice recognition unit 3, that is, the search keyword is input to the search engine 8 as a search request (step S203). The search engine 8 searches for information corresponding to the search keyword in response to the search request, and supplies the search result to the search result transmission unit 4. The search result transmission unit 4 searches the user profile database 9 with the terminal ID of the user terminal 1 and transmits a transmission format corresponding to the user terminal 1, such as an email format, a homepage format, and a file format as shown in FIG. The transmission format and destination desired by the user are searched, and the search result from the search engine 8 is transmitted via the Internet 22 to the destination of the user terminal 1 in the searched transmission format (step S204). The destination is the destination address in the case of e-mail, the display URL in the case of a home page, and the transfer destination address in the case of file transfer.
[0050]
In the search process, the search engine 8 searches for information corresponding to the search keyword in response to a search request from the user terminal 1, and is used for this search every time the search result is supplied to the search result transmission unit 4. The retrieved keywords are stored in the search log database 7 as a time-series list together with the search time and terminal ID. That is, the search log database 7 stores a keyword used for the search by the search engine 8, a search time when the search is performed, and a terminal ID of the user terminal 1 that has requested the search as a time-series list. It is like that.
[0051]
Next, with reference to a flowchart shown in FIG. 4, a process in which the recognition dictionary registration unit 6 selects a word effective for search from the search log database 7 and registers this word in the speech recognition dictionary 5 will be described.
[0052]
In FIG. 4, first, all the words registered in the speech recognition dictionary 5 are deleted (step S300). Then, a time-series list of keywords, search times, and terminal IDs used for search by the search engine 8 is acquired from the search log database 7 (step S301).
[0053]
Next, the recognition dictionary registration unit 6 receives as input a search log stored in the search log database 7, that is, a time-series list of keywords, search times, and terminal IDs used for the search. The relationship between keywords is calculated by a needs extraction method, and a word group, that is, a keyword group, is created by collecting keywords having a large relationship, and the total frequency Vs of a plurality of keywords constituting the group within a predetermined period is calculated for each group. Calculate and output (step S302). For the information needs extraction method, refer to, for example, Information Needs Extraction Based on WWW Search Log, Okubo et al., Information Processing Society of Japan Journal, Vol. 39, No. 7, 1998.
[0054]
The total frequency Vs is the total number of inputs to the search engine 8 for each keyword in the group during a predetermined period of time when the calculation is performed. For example, an output example of the total frequency Vs is as follows.
[0055]
[Table 1]
Words in group (keywords): Total frequency Vs
Christmas, winter, skiing: 456
Summer, diving, Okinawa: 189
Autumn, autumn leaves, fallen leaves: 233
……
Next, the group having the largest total frequency Vs calculated as described above is selected (step S303). Then, it is determined whether or not the total frequency Vs of the selected group is greater than or equal to the registered minimum value Vmin (step S304). If the total frequency Vs is smaller than the minimum registration value Vmin, the present process is terminated. If the total frequency Vs is larger, the word that is each keyword of the selected group is registered in the speech recognition dictionary 5 corresponding to the reading. (Step S305). The reason for comparing the total frequency Vs and the registered minimum value Vmin is that even if a keyword that is used less frequently in the search, that is, a word is registered, it is unlikely that the keyword is used as a keyword in the search by the search engine 8. Even if such a word is registered in the speech recognition dictionary 5, the number of words in the dictionary is simply increased, and the influence of deteriorating the recognition accuracy is relatively large. This is to prevent registration.
[0056]
Next, it is determined whether or not the number of words registered in the speech recognition dictionary 5 has reached the maximum registration number Nmax (step S306). As a result of this determination, if the number of words registered in the speech recognition dictionary 5 has reached the maximum registration number Nmax, this processing ends. If not, all groups output in step S302 Whether or not the process has been performed is determined (step S307). If all groups have been processed, this process is terminated. If not, the process returns to step S303 and the same process is repeated for all groups.
[0057]
The word registration process in the speech recognition dictionary 5 shown in FIG. 4 needs to be performed before the search process shown in FIG. 3, that is, the voice portal service is provided, and the word registration process shown in FIG. By periodically performing the above, it is possible to register a keyword that varies depending on the time in the speech recognition dictionary 5.
[0058]
As described above, in the present embodiment, a search keyword input by voice from the user terminal 1 is received, the voice keyword is recognized by referring to the voice recognition dictionary 5, and the voice recognized keyword is used as a search request. Information corresponding to the keyword is searched by the search engine 8 and output from the search result transmission unit 4 to the user terminal 1, and the keyword used for the search is stored in the search log database 7. Since words that are effective for search are registered as words in the speech recognition dictionary 5 from among them, it is possible to realize a sufficient word coverage for the search service while suppressing a decrease in recognition accuracy due to an increase in the number of registered words. Specifically, the words to be registered in the speech recognition dictionary 5 are not limited to a specific genre such as “place name, company name”, and words for various fields can be registered. It is possible to receive a word in the field as a search input by voice.
[0059]
In the embodiment described above, the relationship between keywords stored in the search log database 7 is calculated, a group is created by grouping together large keywords, and the total frequency of each keyword constituting the group within a predetermined period. Is calculated, the group to be registered in the speech recognition dictionary 5 is selected, and the words that are the keywords of this group are registered in the speech recognition dictionary 5. However, the present invention is not limited to this, for example, The recognition dictionary registration unit 6 selects a word effective for the search from the keywords stored in the search log database 7 based on the frequency of use of the keywords stored in the search log database 7 without creating a keyword group. The selected word may be registered in the speech recognition dictionary 5, or the recognition dictionary registration unit 6 may detect the selected word. Calculate the relationship between keywords stored in the log database 7, create a keyword group that summarizes highly related keywords, and calculate the total usage frequency of multiple keywords in each group regardless of the predetermined period Then, a group may be selected based on the total use frequency, and each keyword of the selected group may be registered in the speech recognition dictionary 5 as a valid word for search. Alternatively, the total frequency used per unit time may be calculated for each keyword group and registered in the recognition dictionary registration unit 6 in the descending order of the total frequency.
[0060]
It should be noted that the voice-utilization type information search processing procedure of the above embodiment is recorded on a recording medium as a program, and the recording medium is incorporated into a computer system, and the program recorded on the recording medium is downloaded or installed in the computer system. Of course, by operating the computer system with the program, it is possible to function as a speech-based information retrieval device that performs speech-based information retrieval processing. It can improve the nature.
[0061]
【The invention's effect】
As described above, according to the present invention, a search keyword inputted by voice from a user terminal is received, the voice keyword is recognized by referring to the voice recognition dictionary, and the voice recognized keyword is used as a search request. Since the information corresponding to the keyword is searched and output to the user terminal, the keyword used for the search is stored, and a word effective for the search is registered as a word in the voice recognition dictionary from the stored keyword. Thus, it is possible to achieve a word coverage sufficient for the search service while suppressing a decrease in recognition accuracy due to an increase in the number of registered words.
[0062]
Further, according to the present invention, a word that is effective for search is selected from keywords based on the frequency of use of the stored keyword, and the selected word is registered in the speech recognition dictionary. The recognition accuracy can be improved while realizing the rate.
[0063]
Further, according to the present invention, the relationship between the saved keywords is calculated, a keyword group in which large keywords are related is created, and the total use frequency of a plurality of keywords in each keyword group is calculated. A keyword group is selected based on the total frequency of use, and each keyword of the selected keyword group is registered in the speech recognition dictionary as a word that is effective for search. Therefore, recognition is performed while realizing sufficient word coverage for the search service. Accuracy can be improved.
[0064]
According to the present invention, a keyword group stored in a predetermined period is selected based on a search time stored together with a keyword, a relationship between the selected keywords is calculated, and a keyword group in which large keywords related to each other are collected. Create, calculate the total usage frequency of a plurality of keywords in each keyword group, select a keyword group based on the total usage frequency, and recognize the speech as a valid word for each keyword in this keyword group Since it is registered in the dictionary, recognition accuracy can be improved while realizing a word coverage sufficient for the search service.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a voice-based information retrieval apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram showing a data configuration stored in a user profile database used in the voice-based information retrieval apparatus shown in FIG.
FIG. 3 is a flowchart showing a search process by a voice interface of the voice-usage type information search apparatus shown in FIG. 1, that is, a search service providing process.
FIG. 4 is a flowchart showing a process of selecting a word effective for search from the search log database and registering it in the voice recognition dictionary in the voice-based information search apparatus shown in FIG. 1;
[Explanation of symbols]
1 User terminal
2 Internet terminal
3 Voice recognition unit
4 Search result transmission part
5 Speech recognition dictionary
6 recognition dictionary registration part
7 Search log database
8 Search engine
9 User profile database
21 Circuit switching network
22 Internet

Claims (6)

A voice-based information retrieval device that retrieves corresponding information based on a search keyword input by voice from a user terminal and outputs the information to the user terminal,
Voice recognition means for receiving a search keyword input by voice from a user terminal, referring to a voice recognition dictionary, voice-recognizing the keyword, and outputting a keyword that is text;
Search means for searching for information corresponding to the keyword by using a keyword that is text recognized by the voice recognition means and a keyword that is text sent from an Internet terminal as a search request;
Search result output means for outputting the searched information to a user terminal or an Internet terminal;
Search keyword storage means for storing a keyword that is text used in the search ;
A keyword group creating means for calculating a relationship between keywords stored in the search keyword storing means and creating a keyword group in which large related keywords are collected;
A total usage frequency calculating means for calculating a total usage frequency of a plurality of keywords in each of the created keyword groups,
Select a keyword group based on the calculated total use frequency, audio-using information retrieval apparatus characterized by having a keyword group selecting means for registering each keyword in the selected keyword group before Symbol speech recognition dictionary . A voice-based information retrieval device that retrieves corresponding information based on a search keyword input by voice from a user terminal and outputs the information to the user terminal,
Voice recognition means for receiving a search keyword input by voice from a user terminal, referring to a voice recognition dictionary, voice-recognizing the keyword, and outputting a keyword that is text;
Search means for searching for information corresponding to the keyword by using a keyword that is text recognized by the voice recognition means and a keyword that is text sent from an Internet terminal as a search request;
Search result output means for outputting the searched information to a user terminal or an Internet terminal;
Search keyword storage means for storing a keyword, which is text used for the search, together with a search time;
Select the keyword stored within a predetermined period of time based on the search time stored in said analyzing Sakuki keyword storage means, an association between the selected keywords to calculate summarizes the major related keywords keywords A keyword group creation means for creating a group;
A total usage frequency calculating means for calculating a total usage frequency of a plurality of keywords in each of the created keyword groups,
Select a keyword group based on the calculated total use frequency, audio-using information retrieval apparatus characterized by having a keyword group selecting means for registering each keyword in the selected keyword group before Symbol speech recognition dictionary . A speech utilization type information search program to be executed by a speech utilization type information retrieval device that retrieves corresponding information based on a search keyword inputted by speech from a user terminal and outputs the information to the user terminal,
The voice-based information retrieval device is
Voice recognition means for receiving a search keyword input by voice from a user terminal, referring to a voice recognition dictionary, voice-recognizing the keyword, and outputting a keyword that is text;
Search means for searching for information corresponding to the keyword by using a keyword that is text recognized by the voice recognition means and a keyword that is text sent from an Internet terminal as a search request;
Search result output means for outputting the searched information to a user terminal or an Internet terminal;
Search keyword storage means for storing a keyword that is text used in the search;
A keyword group creating means for calculating a relationship between keywords stored in the search keyword storing means and creating a keyword group in which large related keywords are collected;
A total usage frequency calculating means for calculating a total usage frequency of a plurality of keywords in each of the created keyword groups,
Keyword group selection means for selecting a keyword group based on the calculated total use frequency and registering each keyword of the selected keyword group in the speech recognition dictionary;
A voice-based information retrieval program to make it function . A speech utilization type information search program to be executed by a speech utilization type information retrieval device that retrieves corresponding information based on a search keyword inputted by speech from a user terminal and outputs the information to the user terminal,
Voice recognition means for receiving a search keyword input by voice from a user terminal, referring to a voice recognition dictionary, voice-recognizing the keyword, and outputting a keyword that is text;
Search means for searching for information corresponding to the keyword by using a keyword that is text recognized by the voice recognition means and a keyword that is text sent from an Internet terminal as a search request;
Search result output means for outputting the searched information to a user terminal or an Internet terminal;
Search keyword storage means for storing a keyword, which is text used for the search, together with a search time;
Based on the search time stored in the search keyword storage means, a keyword stored within a predetermined period is selected, a relationship between the selected keywords is calculated, and a keyword group in which large related keywords are collected A keyword group creation means to create,
A total usage frequency calculating means for calculating a total usage frequency of a plurality of keywords in each of the created keyword groups,
Keyword group selection means for selecting a keyword group based on the calculated total use frequency and registering each keyword of the selected keyword group in the speech recognition dictionary;
A voice-based information retrieval program to make it function . 4. A medium on which a voice-based information retrieval program according to claim 3 is recorded on a computer-readable recording medium. 5. A medium on which a voice utilization type information retrieval program according to claim 4 is recorded on a computer readable recording medium.

Images