JP2005084829A - Information retrieving method, device, and program, and program recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and device for retrieving information, capable of retrieving information using a plurality of voice-recognized recognition results. <P>SOLUTION: This method for retrieving information by voice from a plurality of information supplied by a server via a communication line includes a step for inputting the information of user's requesting retrieval in a terminal by voice, a step for the server to receive from the terminal the voice data regarding the input information, a step for acquiring voice recognition candidate analyzing the contents of the voice data received at the receiving step, using a sound model and a language model and acquiring a plurality of recognition candidates, a step for retrieving information using at least two recognition candidates of the plurality of obtained recognition candidates, a step for transmitting the plurality of information acquired at the information retrieving step to the terminal, and a step for displaying at least one of the plurality of information transmitted in the transmitting step. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an information search method, an information search device, an information search program, and a program recording medium.

  In recent years, mobile terminals such as mobile phones have been widely used, and it is desired to reduce the weight and size of mobile terminals in order to improve portability. For this reason, it is necessary for the portable terminal to reduce or eliminate an input device such as a keyboard as the mobile terminal is downsized.

  On the other hand, various information is provided on the Internet. A user can access necessary information from an arbitrary location using a browser, which is an information access application, from a mobile terminal. However, it is complicated to input characters such as a URL (Uniform Resource Locator) or a search condition from a portable terminal having a small keyboard or no keyboard.

  For this reason, it has been attempted to input information search by voice (see, for example, Patent Document 1). In this document, web pages linked to the displayed web page are pre-read for the number of layers specified in advance, downloaded, and words are extracted from all downloaded web pages for voice search. A voice search method for creating a dictionary and performing a voice search for this word is disclosed.

JP 2002-297372 A (Claims)

 However, in the method described in this document, one recognition result is obtained from the input voice, and a web page is searched from the recognition result. For this reason, when voice recognition is misrecognized, there is a problem that the user is uncomfortable without obtaining a desired search result.

  In the method described in this document, a speech recognition language model or speech recognition dictionary is created only from text such as articles. The language model for speech recognition created in this way has only information such as “whether a sentence can be recognized” or “how much probability that a certain word string exists in the text”. I don't have it. For this reason, a user who views a web page without having a clear purpose cannot fully enjoy it.

  On the other hand, whether a user can enjoy a web page is an important problem for a user who views the web page without having a clear purpose. For such a user, there is no problem as long as the input voice is misrecognized and the web page searched for using a keyword other than the desired keyword can enjoy the web page.

  That is, the present invention has been made in view of the above problems, and an object of the present invention is to provide an information search method and an information search apparatus that can search for information using a plurality of recognition results that have been voice-recognized.

  Another object of the present invention is to provide an information search method and an information search apparatus capable of searching for information by adding another condition to a voice-recognized keyword.

  In order to achieve the above object, an information retrieval method of the present invention is a method for retrieving information by voice from a plurality of information provided by a server through a communication line, and (1) Information requested by a user for retrieval. Input to the terminal by voice, (2) a reception step in which the server receives voice data related to the input information, and (3) the contents of the voice data received in the reception step A recognition candidate acquisition step of analyzing a model and a language model to obtain a plurality of recognition candidates; and (4) searching for information using at least two recognition candidates among the plurality of recognition candidates obtained. (5) a transmission step for transmitting a plurality of pieces of information obtained in the information search step to a terminal; and (6) a copy transmitted in the transmission step. Among the information, having an information display step of displaying at least one.

  According to this configuration, information is searched using at least two recognition candidates among a plurality of recognition candidates obtained as a result of speech recognition. By using a plurality of recognition candidates, the probability that a recognition result close to the user's intention is included in the recognition candidates is increased. As a result, it is possible to reduce the possibility of erroneous recognition of the input voice.

  In addition, at least two recognition candidates are used among the plurality of obtained recognition candidates. As a result, information can be searched using a recognition result that does not necessarily match the keyword intended by the user. That is, when a user searches for information without having a clear purpose, information can be searched using an unexpected recognition result.

  In the information search method of the present invention, the terminal may include a voice analysis step for parameterizing the input voice. Speech recognition is usually performed by a server. However, the voice analysis step does not require a special voice recognition operation. Therefore, according to this configuration, an excessive burden is not given to the terminal.

  Further, in the information search method of the present invention, in the terminal, a speech recognition candidate acquisition step of recognizing the content of speech data using an acoustic model and a language model to obtain a plurality of recognition candidates, and the obtained It is good also as a structure which has a reception step which a server receives a recognition candidate from a terminal.

  The information search method further includes a usage score acquisition step for obtaining a usage score of each information for each of the plurality of information obtained in the information search step, and the plurality of information displayed in the information display step. At least one of the information may be displayed by reflecting the usage score.

  According to this configuration, a usage score is obtained for each piece of information obtained by searching using the speech recognition result. As a result, the information is displayed reflecting the usage score, so that a frequently used website can be accessed. In particular, a user who searches without a clear purpose can access a popular website using an unexpected keyword, and thus does not feel discomfort due to misrecognition.

  Further, at least one of the plurality of pieces of displayed information may be displayed by reflecting a sum of the recognition score of the search candidate used for the search and the usage score. According to this configuration, the sum of the recognition score of the search candidate used for the search and the usage score is reflected in the displayed information. Therefore, information that has a high degree of voice recognition and is used more frequently can be browsed in order.

  The information search method may further include a weighting step for weighting each of the recognition score of the search candidate used for the search and the usage score. According to this configuration, the user can change the weighting of the recognition score of the search candidate used for the search and the usage score depending on the purpose of the search.

  The information retrieval method of the present invention is a method for retrieving information by voice from a plurality of pieces of information provided by a server through a communication line, and (1) an input step in which information requested by the user is input to the terminal by voice. And (2) a reception step in which a server receives voice data relating to the input information from a terminal, and (3) the content of the voice data received in the reception step is recognized using an acoustic model and a language model. A speech recognition candidate acquisition step for obtaining a plurality of recognition candidates; (4) a recognition candidate display step for displaying at least two recognition candidates among the obtained plurality of recognition candidates on a terminal; A recognition candidate selection step in which the user selects one recognition candidate from at least two displayed recognition candidates; and (6) the recognition candidate selected in the recognition candidate selection step. Using an information retrieval step of retrieving information, it is also possible to have a.

  According to this configuration, the user can select one recognition candidate from the displayed plurality of recognition candidates. As a result, information can be searched for using recognition candidates according to the user's purpose. Moreover, even if the recognition result is not scheduled at the time of voice input, a keyword that the user is interested in can be selected, so that more free information retrieval is possible.

  The information search apparatus according to the present invention is an information search apparatus for searching information by voice from a plurality of information provided by a server through a communication line, and (1) a user requests information to be searched by voice. (2) receiving means for receiving voice data related to the input information from a terminal from a terminal; and (3) contents of the voice data received by the receiving means, an acoustic model and a language model. And (4) an information search for searching for information using at least two recognition candidates among the obtained plurality of recognition candidates. Means, (5) transmitting means for transmitting a plurality of pieces of information obtained by the information searching means to a terminal, and (6) expressing at least one of the plurality of pieces of information transmitted by the transmitting means at the terminal. And information display means for, and has a.

  The information retrieval program of the present invention is an information retrieval program for retrieving information by voice from a plurality of information provided by a server through a communication line, and (1) information requested by a user to be retrieved by voice is transmitted to a computer. An input step to input to the terminal; (2) a reception step in which a server receives voice data related to the input information; and (3) a content of the voice data received in the reception step, an acoustic model and a language. A speech recognition candidate acquisition step for recognizing using a model and obtaining a plurality of recognition candidates; and (4) information for searching for information using at least two recognition candidates among the plurality of recognition candidates obtained. A search step; (5) a transmission step of transmitting a plurality of information obtained in the information search step to a terminal; and (6) in the transmission step. Of the transmitted plurality of information, but to execute the information display step of displaying at least one in terminal.

  The information search program may be recorded on a computer-readable recording medium.

  The information search method of the present invention searches for information by using at least two recognition results from among a plurality of recognition candidates obtained by analyzing input speech. As a result, erroneous recognition is reduced, and information closer to the user's intention can be searched.

  Further, the information search method of the present invention can obtain a use score of each information for each of a plurality of obtained information, and can search for information in which the use score of each information is added to a speech-recognized keyword. An information search method and an information search device can be provided.

  The best mode for carrying out the present invention will be described below with reference to the drawings. In addition, this invention is not limited by these.

(Example 1)
FIG. 1 is a block diagram showing a system configuration for realizing an information search method according to an embodiment of the present invention. In FIG. 1, the terminal 100 and the server 200 are connected via a communication line 300.

  The terminal 100 is provided with a voice input unit 101, a display unit 102, and a weight input unit 103. The voice input from the voice input unit 101 such as a microphone is converted into a digital signal by the AD conversion unit 104. The digitized audio data is sent from the transmission control unit 105 to the server 200 via the communication line 300. Further, when the user weights information search, a predetermined weight is input from the weight input unit 103.

  Note that the term “terminal” as used in the present invention does not refer to a “mobile phone” that is generally called, but includes a telephone function and an Internet function, such as a PHS or a next-generation mobile phone, such as a mobile computer or Communication terminals such as car navigation systems are also included in the category.

  Communication lines include public lines, analog telephone lines for LAN, ISDN (combined service / digital network), DSL, Ethernet (registered trademark), optical fiber lines, PHS, mobile phones (line connection, bucket connection), wireless LAN (Local area network), fixed microwave line, satellite communication line and the like.

  The server 200 includes a reception control unit 201 that receives information from the terminal 100 side, a voice analysis unit 212 that analyzes the received voice data, a voice recognition candidate acquisition unit 213 that recognizes the analyzed voice data, An acoustic model table and language model table 223 used for speech recognition, a word dictionary 224 used for speech recognition, an information candidate acquisition unit 214 for acquiring information candidates, and a transmission control unit 226 for transmitting the acquired information candidates to the terminal And an information analysis unit 220 that analyzes necessary information from a web page on the Internet 400, and an information database 221 that stores data relating to the analyzed information. The word dictionary 224 for speech recognition is created by the information analysis unit 220 from the information database 221 that stores character information obtained from a web page on the Internet. The usage score is obtained by analyzing a web page on the Internet 400 by the information analysis unit 220.

  In this embodiment, when the reception control unit 211 of the server 200 receives the voice data from the terminal 100, the voice analysis unit 212 performs voice analysis. Specifically, feature parameters are extracted by, for example, cepstrum analysis and converted into MFCC (Mel-Frequency Cepstrum Coefficient) parameters.

  The speech data that has been converted into feature parameters is subjected to acoustic model processing and language model processing by using the acoustic model and the language model in the speech recognition candidate acquisition unit 213. In this embodiment, since the acoustic model processing and language model processing are performed by the server, a dictionary and grammar with a sufficient number of semantic meanings can be prepared in advance on the server side, and the dictionary and grammar can be downloaded to the terminal side. Since waiting time can be reduced, recognition processing can be performed quickly.

  Further, since the acoustic model processing and the language model processing are performed by the server 200, it is not necessary to store acoustic information and language information on the terminal 100 side. As a result, accurate recognition processing can be performed efficiently even when there are few resources available on the terminal side and the processing capability is insufficient.

  The speech recognition candidate acquisition unit 213 extracts phoneme information (phoneme information) from the feature parameters by using an acoustic model. As the acoustic model, for example, an HMM (Hidden Markov Model) can be used. The phoneme information can be, for example, a phoneme candidate and its likelihood column.

  The extracted phoneme information is then subjected to language level recognition processing using a word model using the word dictionary 224. By this recognition processing, a plurality of recognition result candidates are acquired.

  The language model is created as follows. First, the information analysis unit 220 manually creates a subset of WWW (World Wide Web) that the user needs. Next, words for speech recognition are collected from the text of the Web page on the created subset. Specifically, the text excluding the HTML tag of the Web page is extracted. The extracted text is recorded in the information table 221. Thereafter, the extracted text is subjected to morphological analysis to obtain a noun phrase. The obtained noun phrase is registered in the word dictionary 224. When extracting a hyperlink from an HTML sentence, <A href=“URL”>. . . . It moves to the link destination indicated by </A>, collects speech recognition words, and registers them in the speech recognition word dictionary 224 in the same manner as described above. In this way, by collecting words for speech recognition from the text of the Web page, it is possible to avoid recognizing words that have no meaning even if searched. As a result, the amount of calculation at the time of recognition can be reduced. In this example, the word dictionary 224 is provided separately from the information database 221, but the information database 221 may include a word dictionary.

  There is no particular limitation on the number of words for speech recognition. When collecting words from the WWW, the number of registered words is very large. On the other hand, if the number of recognition target words is too large, the speech recognition performance is greatly deteriorated and a large amount of CPU, memory, etc. are used during speech recognition. For this reason, the same word as the word registered in the recognition word dictionary 224 is not registered. Here, when the word registered in the word dictionary 224 is used a plurality of times on the Web page, the number of uses is counted. In particular, it is preferable to limit the number of words for speech recognition to be registered. For example, 100,000 words from the top are registered in the word dictionary, and other words are treated as unknown words.

  For the Web page on which the speech recognition words are collected, the URL (Uniform Resource Locator) (character string) of the Web page, the title (character string) of the Web page, and the word on the Web page are stored in the information database 221. The noun phrases (character strings) registered in the dictionary, the usage frequency (numerical arrangement) of the noun phrases registered in the word dictionary, the total number (numerical values) of the noun phrases registered in the word dictionary 224, and the like are recorded. .

(Usage)
The usage score of each information is obtained as follows.
Whether or not a certain Web page is used can be determined from the following values.
1) Number of web page accesses 2) Number of web pages linked by other web pages

  As the information usage score, it is most preferable to use the number of accesses of 1). However, not all web pages actually count the number of accesses. Therefore, the method 1) can be used only for Web pages counting the number of accesses.

  For Web pages that do not count the number of accesses, the method 2) is used. When Web page A links Web page B, it is regarded as a support vote for Web page B by Web page A. That is, the web pages are ranked based on the number of linked web pages.

  First, a subset of the WWW is created manually or using an engine using text matching such as Alta Vista. Next, each Web page is classified into “authority” and “hub”. Here, the authority means a page frequently linked from a page having high value as a hub, and the hub is a page on which hyperlinks to the authority are gathered. Then, the respective weights of the hub and the authority are obtained by iterative algorithm. A score of a certain page is obtained from each obtained weight, and this score is used as a usage score. As for Iterative Algorithm, Jon M. et al. Kleinberg, “Authentic Sources in a Hyperlinked Environment”, Proceedings of the ACM-SIAM Symposium on Discrete Algorithms, 1998. The score of the Web page obtained by such a method does not directly reflect popularity, but is considered to reflect popularity to some extent.

  Any of the above methods 1) and 2) may be used, and can be used properly at the time of design depending on the situation. The sum of the scores obtained by the methods 1) and 2) may be used as the usage score. By using these methods, the collected usage score is recorded in the information of each Web page of the information table 221.

(Weighting)
When searching for information reflecting the usage score, the voice recognition score and the usage score are weighted. Weighting is performed by multiplying the voice recognition score and the usage score by the weight value set by the user, respectively, and adding the multiplied voice recognition score and the multiplied usage score. As for the weight value, the speech recognition score is a positive value larger than 0, and the usage score is a positive value larger than 0. By weighting, the user can select which value is to be searched for information among the speech recognition score and the usage score. In particular, by changing the weight value between the voice recognition score and the usage score, when the usage score is smaller than the voice recognition score, it is possible to search for information by reflecting the usage score. When weighting is not performed, the value of the speech recognition score weight and the value of the usage score weight are both 1. In this specification, the sum of the recognition score and the usage score, or the sum of values obtained by multiplying the recognition score and the usage score by the respective weight values may be referred to as a total score.

  In the information candidate acquisition unit 214 by the above method, information is searched using at least two of the plurality of obtained recognition candidates. Since information is searched using a plurality of recognition candidates, erroneous recognition can be reduced. In addition, it is possible to search for information that can be enjoyed by a user who searches for information without having a clear purpose. The obtained information is sent to the terminal 100 and displayed.

  The information candidate acquisition unit 214 may transmit the obtained recognition candidate from the transmission control unit 226 to the terminal 100 via the communication line 300 by the above method. The reception control unit 106 on the terminal 100 side receives the recognition candidates, and the display unit 102 displays at least two of the received plurality of recognition candidates. The user selects a recognition candidate desired to be searched from among the displayed recognition candidates. The selected recognition candidate is sent to the server 200, and the information candidate acquisition unit 214 searches for information based on the recognition candidate.

(Operation 1)
Next, the information search method of the present invention will be described in detail with reference to FIGS. FIG. 2 is a flowchart showing an information search operation according to the present invention. In the example of this figure, the recognition candidates are not weighted. FIG. 3 is a diagram showing an example in which information is actually searched in the present invention. FIG. 3A is a diagram showing the top two of the recognition candidates recognized by speech. FIG. 3B is a diagram for displaying the result of the information searched using each recognition candidate.

[Step S101]
In step S101, speech input of recognition candidates is started. Prior to the start of speech input for recognition candidates, the output buffer is initialized.

[Step S102 to Step S103]
In step S102, the input recognition candidate is recognized by using an acoustic model. Next, in step S103, it is checked whether or not the recognized speech includes a recognized character string in the word dictionary. If no recognized character string exists, the process proceeds to step S111. If a recognized character string exists, the process proceeds to step S104.

[Step S104 to Step S105]
If one or more recognized character strings exist in step S104, the recognized character string is output and displayed. Specifically, recognition candidates as shown in FIG. 3A are displayed. When there are a plurality of recognized character strings, for example, two character strings are displayed on the display unit of the terminal from the highest recognition score. The user selects a recognition candidate that he or she desires to search (step S105).

[Step S106]
In step S106, information in the information database is searched based on the selected recognition candidate. The search is performed depending on whether there is information that matches the keyword corresponding to the recognition candidate. In this case, the match with the keyword does not have to be completely matched with the keyword, and if there is a character string included in the recognized character string, it is regarded as a match.

[Step S107]
In step S106, it is determined whether there is information that matches the keyword corresponding to the recognition candidate. If there is no information matching the keyword corresponding to the recognition candidate, the process proceeds to step 110. If there is information matching the keyword corresponding to the recognition candidate, the process proceeds to step S108.

[Step S108]
In step S108, the total score of the output buffer is updated with the keyword usage frequency.

[Step S109]
Sort the output buffer in order of overall score and output. The output information is sent from the transmission control unit to the terminal. The terminal receives the information sent from the server and displays it on the display unit. As shown in FIG. 3B, information searched based on the selected speech recognition candidate is displayed in the order of the usage score.

[Step S110]
When there is no recognized character string or when there is no information matching the keyword corresponding to the recognition candidate, the fact that the recognized character string or information does not exist is output as an error message (step S110).

  In the case of this operation, for example, when the user speaks “house design” and “picture design” and “house design” are displayed as recognition results in this order, the user recognizes the recognition result. The search target can be changed by referring to. Therefore, it is possible to perform information retrieval with a high degree of freedom for a user who retrieves information without a clear purpose.

(Operation 2)
Next, the information search method of the present invention will be described in detail with reference to FIG. FIG. 4 is a flowchart showing an information search operation according to the present invention. In the example of this figure, when the recognition candidates are not weighted, the information search is performed without selecting one specific recognition candidate from the plurality of recognition candidates.
[Step S201 to Step S203]
Steps S201 to S203 are the same as steps S101 to S103 of the operation 1. Prior to the start of speech input for recognition candidates, the output buffer is initialized.

[Step S204 to Step S207]
Steps S204 to S207 are basically the same as the operations of Step S106 to Step S109 of the above operation 1. However, step S204 to step S207 differ from operation 1 in that information is searched for each recognition candidate when a plurality of recognition candidates are obtained.

[Step S208 to Step S210]
In step S208, the recognition result is displayed. Specifically, recognition candidates as shown in FIG. In step S209, the user selects a recognition candidate for which information is desired. In step S210, information based on the recognition candidate selected by the user is displayed in the order of the usage score, as shown in FIG. 3B, information searched based on the selected speech recognition candidate.

[Step S211]
If the recognized character string does not exist, or if there is no information matching the keyword corresponding to the recognition candidate, it is output as an error message that no recognized character string or information exists (step S211).

  According to this operation, when there are a plurality of recognition candidates, the user can view the search result based on the plurality of recognition candidates. In the example of FIG. 3, when the user utters “house design” and “picture design” and “house design” are displayed as recognition results in this order, from any recognition candidate, , Information can be searched. Therefore, a wide range of information can be searched for a user who searches for information without having a clear purpose.

(Operation 3)
Next, the information search method of the present invention will be described in detail with reference to FIGS. FIG. 5 is a flowchart showing an information search operation according to the present invention. In the example of this figure, the recognition candidates are weighted. FIG. 6 is a diagram illustrating a process in which the retrieved information is weighted. FIG. 6A is a diagram showing the top two recognition candidates recognized by speech recognition. FIG. 6B is a diagram illustrating a result of information searched using each recognition candidate. FIG. 6C is a diagram illustrating a result of weighting each searched information.

[Step S301]
In step S301, a recognition score and a usage score weight value are input. Prior to the start of weight input, the output buffer is initialized.

[Step S302 to Step S307]
Steps S302 to S307 are the same as steps S201 to S206 of the operation 2.

[Step S308]
In step S308, weighting is performed. As shown in FIG. 6A, the speech recognition candidates are the same as those in the above operations 1 and 2. Also. As shown in FIG. 6B, the information obtained from each recognition candidate is the same as in the case of operations 1 and 2. For each piece of information obtained, weighting is performed by adding a value obtained by multiplying the recognition score by the weight value of the recognition score and a value obtained by multiplying the usage score of each information by the value of the weight of the usage score. Do. This added value is taken as the total score. In the example of FIG. 6C, the recognition score weight value and the usage score weight value are each 1.

[Step S309 to Step S310]
In step S309, the recognition result to which the weighted information is output is displayed. Specifically, recognition candidates as shown in FIG. In step S310, as shown in FIG. 3B, information obtained by searching for weighted information in the descending order of the total score is displayed.

[Step S311]
If the recognized character string does not exist, or if there is no information matching the keyword corresponding to the recognition candidate, the fact that the recognized character string or information does not exist is output as an error message (step S311).

  In the case of this operation, as shown in FIG. 6A, for example, even when the recognition score of the recognition candidate “house design” is smaller than the recognition score of “picture design”, the value of the usage score As shown in FIG. 6C, the “house design collection” can be displayed at the top. That is, even if it is erroneously recognized, information retrieved based on recognition candidates having a low recognition rank can be referred to at the same time by reflecting the usage score. Further, even if the unexpected information retrieved by the misrecognized “picture design” is useful, the user can also use the information. In the example of FIG. 6C, information related to “house design” but not obtained when searching by “house design” is obtained, such as “wallpaper design”. In this way, by integrating the recognition score and the usage score, it is possible to obtain a system that is easier to use and enjoyable.

(Example 2)
FIG. 7 is a block diagram showing a system configuration for realizing an information search method according to another embodiment of the present invention.

  As shown in FIG. 7, in the present embodiment, not all recognition processing is performed on the server 200 side, but feature parameter extraction is performed on the terminal 100, and subsequent language level recognition is performed on the server 200 side. That is, instead of the voice analysis unit 213 in the server 200 in FIG. 1, the present embodiment is different from the first embodiment only in that the terminal 100 is provided with the voice analysis unit 107.

  Therefore, in the example of this figure, the terminal 100 performs the extraction of feature parameters obtained by voice analysis. For this reason, it is possible to transmit information necessary for voice recognition without burdening the terminal and with a small communication amount.

(Modified Example)
The information search apparatus according to the present invention is not limited to the voice analysis unit, and a voice recognition candidate acquisition unit may be provided in the terminal. In this case, the necessary acoustic model and language model may be downloaded from the server side.

  Moreover, it is good also as a structure which performs until the extraction of a phoneme level in a speech recognition acquisition part. In this case, since the dictionary and grammar need only be prepared on the server side, even if there are few resources available on the terminal side and the processing capability is insufficient, it can be performed with high accuracy and good recognition. In addition, it is not necessary to send phoneme information to the server via the communication line and send the voice waveform data itself uttered to the server. For this reason, the cost of the equipment at the time of performing recognition processing can be suppressed on the server side, and the voice recognition function can be easily incorporated into the server.

By the way, the function as the information retrieval apparatus in each of the above embodiments is realized by an information processing program recorded on a program recording medium. The program recording medium in each of the above embodiments is a program medium composed of a ROM (Read Only Memory) provided separately from a RAM (Random Accelerator Memory). Alternatively, it may be a program medium that is loaded into an external auxiliary recording device and read out. In any case, the program reading means for reading the information processing program from the program medium may have a configuration for directly accessing and reading the program medium, or a program storage area ( (Not shown) may be downloaded, and the program storage area may be accessed and read. It is assumed that a download program for downloading from the program medium to the program recording area of the RAM is stored in advance in the main unit.
Here, the program medium is configured to be separable from the main body side, and includes a tape system such as a magnetic tape and a cassette tape, a magnetic disk such as a flexible disk and a hard disk, a CD (compact disk) -ROM, and a MO (magneto-optical). ) Discs, optical discs such as discs, MD (mini discs), DVDs (digital versatile discs), card systems such as IC (integrated circuit) cards and optical cards, mask ROMs, EPROMs (ultraviolet erasable ROMs), EEPROMs ( It is a medium that carries a fixed program, including a semiconductor memory system such as an electrically erasable ROM) and a flash ROM.
In addition, the information retrieval apparatus in the above embodiment includes a modem and can be connected to a communication network including the Internet. In this case, the program medium may be a medium that fluidly carries the program by downloading from a communication network or the like. In this case, it is assumed that a download program for downloading from the communication network is stored in the main device in advance. Or it shall be installed from another recording medium.
It should be noted that what is recorded on the recording medium is not limited to a program, and data can also be recorded.

FIG. 1 is a block diagram showing a system configuration for realizing an information search method according to an embodiment of the present invention. FIG. 2 is a flowchart showing an information search operation according to the present invention. FIG. 3 is a diagram showing an example in which information is actually searched in the present invention. FIG. 3A is a diagram showing the top two of the recognition candidates recognized by speech. FIG. 3B is a diagram for displaying the result of the information searched using each recognition candidate. FIG. 4 is a flowchart showing an information search operation according to the present invention. FIG. 5 is a flowchart showing an information search operation according to the present invention. FIG. 6 is a diagram illustrating a process in which the retrieved information is weighted. FIG. 6A is a diagram showing the top two recognition candidates recognized by speech recognition. FIG. 6B is a diagram illustrating a result of information searched using each recognition candidate. FIG. 6C is a diagram illustrating a result of weighting each searched information. FIG. 7 is a block diagram showing a system configuration for realizing an information search method according to another embodiment of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Terminal 101 Voice input part 102 Display part 103 Weight input part 104 AD conversion part 105 Transmission control part 106 Reception control part 107 Voice analysis part 200 Server 211 Reception control part 212 Voice analysis part 213 Voice recognition candidate acquisition part 214 Information candidate acquisition part 220 Information Analysis Unit 221 Information Database 224 Word Dictionary 225 Acoustic Model Table and Language Model Table 226 Transmission Control Unit 227 Communication Line 300 Communication Line 400 Internet

Claims (10)

A method for retrieving information by voice from a plurality of information provided by a server through a communication line,
An input step in which the user requests information to be searched by voice to the terminal;
A receiving step in which a server receives voice data relating to the input information from a terminal;
A speech recognition candidate acquisition step of recognizing the content of the speech data received in the reception step using an acoustic model and a language model, and obtaining a plurality of recognition candidates;
An information search step for searching for information using at least two recognition candidates among the plurality of obtained recognition candidates;
A transmission step of transmitting a plurality of pieces of information obtained in the information search step to a terminal;
An information display step for displaying at least one of the plurality of pieces of information transmitted in the transmission step;
A method for retrieving information, comprising:   The information search method according to claim 1, wherein the terminal includes a voice analysis step of parameterizing the input voice. A method for retrieving information by voice from a plurality of information provided by a server through a communication line,
An input step in which the user requests information to be searched by voice to the terminal;
A speech recognition candidate acquisition step of recognizing the content of the speech data related to the input information using an acoustic model and a language model to obtain a plurality of recognition candidates;
A receiving step in which a server receives the obtained recognition candidate from a terminal; and
An information search step of searching for information using at least two recognition candidates among the plurality of recognition candidates received in the reception step;
A transmission step of transmitting a plurality of pieces of information obtained in the information search step to a terminal;
An information display step for displaying at least one of the plurality of pieces of information transmitted in the transmission step;
A method for retrieving information, comprising: Each of the plurality of information obtained in the information search step has a usage score acquisition step of obtaining a usage score of each information,
The information search method according to any one of claims 1 to 3, wherein in the information display step, at least one of the plurality of pieces of displayed information is displayed by reflecting a usage score.   5. The display device according to claim 4, wherein at least one of the plurality of pieces of displayed information is displayed by reflecting a sum of a recognition score of a search candidate used for the search and the usage score. Information retrieval method described.   The information search method according to claim 5, further comprising a weighting step of weighting each of the recognition score of the search candidate used for the search and the usage score. A method for retrieving information by voice from a plurality of information provided by a server through a communication line,
An input step in which the user requests information to be searched by voice to the terminal;
A receiving step in which a server receives voice data relating to the input information from a terminal;
A speech recognition candidate acquisition step of recognizing the content of the speech data received in the reception step using an acoustic model and a language model, and obtaining a plurality of recognition candidates;
A recognition candidate display step of displaying at least two recognition candidates on the terminal among the plurality of obtained recognition candidates;
A recognition candidate selection step in which the user selects one recognition candidate from the displayed at least two recognition candidates;
An information search step for searching for information using the recognition candidates selected in the recognition candidate selection step;
A method for retrieving information, comprising: An information retrieval device for retrieving information by voice from a plurality of information provided by a server through a communication line,
An input means for inputting information that the user requests to search into the terminal by voice;
Receiving means for receiving voice data relating to the input information from a terminal by a server;
Speech recognition candidate acquisition means for recognizing the content of the voice data received by the receiving means using an acoustic model and a language model to obtain a plurality of recognition candidates;
Information search means for searching for information using at least two recognition candidates among the plurality of obtained recognition candidates;
Transmitting means for transmitting a plurality of information obtained in the information search means to the terminal;
Information display means for displaying at least one of the plurality of information transmitted by the transmission means;
An information retrieval apparatus comprising: An information retrieval program for retrieving information by voice from a plurality of information provided by a server through a communication line,
On the computer,
An input step in which the user requests information to be searched by voice to the terminal;
A receiving step in which a server receives voice data relating to the input information from a terminal;
A speech recognition candidate acquisition step of recognizing the content of the speech data received in the reception step using an acoustic model and a language model, and obtaining a plurality of recognition candidates;
An information search step for searching for information using at least two recognition candidates among the plurality of obtained recognition candidates;
A transmission step of transmitting a plurality of pieces of information obtained in the information search step to a terminal;
An information display step for displaying at least one of the plurality of pieces of information transmitted in the transmission step;
An information retrieval program characterized by causing A computer-readable recording medium on which the information search program according to claim 8 is recorded.

Images