JP2002297372A - Method, device and program for retrieving voice in web page - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a web page voice retrieving method which retrieves voice concerning not only the link item of a web page but also the contents of the web page and also recognizes voice with high precision when a user performs natural speech. SOLUTION: An HTML analyzing part 2 generates a voice retrieval dictionary 3 by connecting a word extracted from the web page with the URL of the web page where the word is extracted. A task control part 5 generates a voice recognition task from the extracted word and selects a language model(LM) and an acoustic model(AM) being optimum to the voice recognition task from a prepared language model and acoustic model group 6. An intention analyzing part 8 analyzes the contents of the user's speech recognized by a voice recognizing part 7 and detects the word which indicates the user's speech contents. A browser control part 9 changes the web page based on the retrieval result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for retrieving voice from a web page, and more particularly, to a method for retrieving voice from a web page using a WWW browser.

[0002]

2. Description of the Related Art Conventionally, there have been proposed many conventional techniques for retrieving a hyperlink embedded in a web page by voice in order to retrieve a web page by voice.
For example, a conventional technology for retrieving a link item of a web page currently displayed on a WWW browser by voice is described below.
This is disclosed in Japanese Patent Application Laid-Open No. 11-184671 entitled "Information Presentation Method and Apparatus, Information Presentation System". Further, a conventional technology for searching by voice for a currently displayed web page and a link item of a web page linked to the currently displayed web page is disclosed in IPSJ Transactions Vol.40 No.2 p443-p452. (1999), "Browsing WWW Using Voice Keywords".

[0003]

However, in the above-mentioned prior art, only the link items of the web page can be searched for the voice, and the content of the web page cannot be searched. Further, in the above-described related art, when recognizing a voice from a user, since the main purpose is to recognize words, only voice recognition using a simple grammar can be performed. Therefore, when performing a voice search based on the natural utterance of the user, there is a problem that the accuracy of voice recognition is low. The present invention has been made to solve these problems, and it is possible to perform a voice search not only on the link items of the web page but also on the contents of the web page, and
It is an object of the present invention to provide a web page voice search method capable of realizing highly accurate voice recognition even when a user speaks naturally.

[0004]

According to a first aspect of the present invention, there is provided a voice search method for a web page, wherein a web page linked to a currently displayed web page is specified in a predetermined hierarchy. Web page acquisition means for prefetching and downloading a number of words, link providing means for associating the extracted word with the URL of the Web page from which the word is extracted, and voice retrieval from the word and the URL associated with the word Dictionary creating means for creating a dictionary for use, task creating means for creating a speech recognition task from words, language model / acoustic model selecting means for selecting an optimal language model and acoustic model for the speech recognition task, and a user uttered Voice recognition means for recognizing voice, intention detection means for analyzing and detecting the intention of the utterance content of the user, and voice detection for detecting the detected intention Search means for searching the book, characterized in that a web page changing means for changing the web page based on the search result. According to the first aspect of the present invention, words can be extracted from all downloaded web pages, and a link can be added to the extracted words to create a voice search dictionary. In addition, voice search can be performed for the contents of the web page. In addition, since a speech recognition task can be created from words extracted from a web page, it is possible to select an optimal language model and acoustic model when performing a speech search on a downloaded web page.

Further, the invention according to claim 2 is characterized in that the word extracting means extracts text from all the web pages downloaded by the web page obtaining means, and then extracts words by morphologically analyzing the text. And According to the invention of claim 2, the word extracting means can analyze the text described in all the web pages downloaded by the web page obtaining means and extract words.

The invention according to claim 3 is characterized in that the link assigning means assigns one or more links to the word extracted by the word extracting means.
Alternatively, the voice search method according to claim 2 is characterized. According to the third aspect of the present invention, the link assigning unit can assign one or more links to the word.

According to a fourth aspect of the present invention, the task creating means compares the word groups extracted from the word extracting means,
A task having a similar concept or a task having a high degree of relevance is a speech recognition task. According to the invention of claim 4, the task creating means can create, from the word group extracted from the web page, a speech recognition task suitable for performing a voice search on the web page.

According to a fifth aspect of the present invention, the language model / acoustic model selecting means compares the speech recognition task created by the task creating means with a task of a language model / acoustic model group prepared in advance, It is characterized by selecting a task with a similar concept or a task with a high degree of relevance. According to the invention of claim 5, based on the speech recognition task created by the task creating means, a language model and an acoustic model optimal for the speech recognition task are selected from a group of language models and acoustic models prepared in advance. Can be.

According to a sixth aspect of the present invention, the language model / acoustic model selecting means matches the speech recognition task created by the task creating means with a task included in a group of speech recognition tasks prepared in advance, and performs semantics. A task close to is selected. According to the invention of claim 6, based on the speech recognition task created by the task creating means, a language model and an acoustic model optimal for the speech recognition task can be selected from a group of speech recognition tasks prepared in advance. .

The invention according to claim 7 is characterized in that, when recognizing a voice uttered by a user, a voice recognition task is dynamically selected to perform voice recognition. According to the seventh aspect of the present invention, since the voice recognition means can dynamically select a voice recognition task suitable for the content of the utterance of the user and perform voice recognition, it is possible to realize highly accurate voice recognition. .

The invention according to claim 8 is characterized in that when recognizing a voice uttered by a user, voice recognition is performed by dynamically combining voice recognition tasks. According to the eighth aspect of the present invention, since the voice recognition means can dynamically perform voice recognition by appropriately combining voice recognition tasks suitable for the utterance content of the user, it is possible to realize highly accurate voice recognition.

According to a ninth aspect of the present invention, there is provided a speech retrieval apparatus for a web page, wherein a web page linked to a currently displayed web page is prefetched by a predetermined number of layers and downloaded. Web page acquisition means, word extraction means for extracting words from all downloaded web pages,
Linking means for associating the URL of the web page from which the word has been extracted, dictionary creating means for creating a voice search dictionary from the word and the URL associated with the word, and creating a speech recognition task from the word Task creating means, a language model / acoustic model selecting means for selecting an optimal language model and acoustic model for the speech recognition task, and a speech recognizing means for recognizing a speech uttered by the user,
Intention detection means for analyzing and detecting the intention of the utterance content of the user, search means for searching for the detected intention from the voice search dictionary, and web page changing means for changing a web page based on the search result It is characterized by having.
According to the ninth aspect of the present invention, it is possible to perform a voice search not only on the link items of the web page but also on the contents of the web page, and to realize highly accurate voice recognition even when the user speaks naturally. It is possible to provide a web page voice recognition device capable of performing the above.

According to a tenth aspect of the present invention, there is provided a voice search program for a web page, wherein a computer for voice search of the web page is controlled in advance by using a web page linked to the currently displayed web page. Web page acquisition means for prefetching and downloading a specified number of layers, word extraction means for extracting words from all downloaded Web pages, and URLs of the Web pages from which the words are extracted for the extracted words. Linking means for associating the word, a dictionary creating means for creating a speech search dictionary from the word, and a URL associated with the word, a task creating means for creating a speech recognition task from the word, Language / acoustic model selection means for selecting the optimal language model and acoustic model for Voice recognition means for recognizing the uttered voice, intention detection means for analyzing and detecting the intention of the utterance content of the user, search means for searching the detected intention from the voice search dictionary, and a web based on the search result. It is characterized by functioning as a web page changing means for changing a page. According to the tenth aspect, it is possible to perform a voice search not only for the link items of the web page but also for the content of the web page, and to realize highly accurate voice recognition even when the user speaks naturally. It is possible to provide a web page voice search program that can be used.

[0014]

After reading and downloading the web pages linked to the currently displayed web page by a predetermined number of layers, words are extracted from all the downloaded web pages. Then, a voice search dictionary is created by associating the extracted word with the URL of the web page from which the word has been extracted. Further, a speech recognition task is created from the extracted words, and a language model and an acoustic model that are optimal for the speech recognition task are selected. Then, the speech uttered by the user is recognized, and a word indicating the uttered content of the user is detected. The detected word is searched from the voice search dictionary, and the web page is changed based on the search result.

[0015]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a functional block diagram showing a system configuration for realizing a web page voice search method according to the present invention. FIG. 2 is a diagram illustrating a method of extracting a hyperlink, and FIG. 3 is a diagram illustrating a method of extracting a word. FIG. 4 is a diagram for explaining a method of associating a hyperlink with an extracted word, and FIG. 5 is a diagram showing a voice search dictionary. Also,
FIG. 6 is a diagram illustrating a method of creating a speech recognition task, and FIG. 7 is a diagram illustrating a method of selecting a language model / acoustic model. FIG. 8 is a flowchart showing a procedure for retrieving a web page by voice.

First, a system configuration for realizing the web page voice search method of the present invention will be described with reference to a functional block diagram shown in FIG. In FIG. 1, 1
Is a WWW browser, 2 is an HTML analysis unit, 3 is a voice search dictionary, 4 is a thesaurus, 5 is a task control unit, 6 is a language model / acoustic model group, 7 is a speech recognition unit, 8 is an intention analysis unit, and 9 is an intention analysis unit. Browser control unit.

The WWW browser 1 is software for browsing files and data in a hypertext format stored in a WWW server on the Internet on a client side, and specifies a URL input from a user or a browser control unit 9. Download and display a web page (hereinafter simply referred to as “page”). In the present embodiment, as the WWW browser 1, for example, Microsoft Internet Explorer, Netscape Communications Netscape Navigator, or the like is used. The WWW browser 1 is not limited to software used in a PC (personal computer), but may be software used in a mobile phone, a PDA (portable information terminal), or the like. Absent.

First, the HTML analysis unit 2 performs HT of all the pages downloaded by the WWW browser 1 to the storage area.
The voice search dictionary 3 is created by analyzing the ML document. Then, the task control unit 5 creates a speech recognition task from the created speech search dictionary 3.

First, a method for creating a voice search dictionary will be described. When creating the voice search dictionary 3, WW
The HTML document downloaded by the W browser 1 is analyzed, and (1) extraction of a hyperlink, (2) extraction of a word,
(3) associating hyperlinks with the extracted words;
I do.

(1) Extraction of Hyperlink When extracting a hyperlink from an HTML document, HT
The text described between the URL specified by the HREF option of the <A></A> tag, which is the ML tag, and the start tag <A> and the end tag </A> is extracted. All the extracted texts are morphologically analyzed and divided into words. Then, the extracted URL and word are stored in the voice search dictionary 3. For example, as shown in FIG. 2, “http: //www.epso” specified by the HREF option of the <A></A> tag
n.co.jp/ ”, start tag <A> and end tag </ A
The text "Seiko Epson Corporation" written between> is extracted. Then, the extracted text "Seiko Epson Corporation" is divided into the words "Seiko Epson" and the word "Corporation" by morphological analysis, and the above "http: //www.epson.c
The URL is stored in the voice search dictionary 3 together with the URL “o.jp/”. The user can arbitrarily specify the level at which the text is to be morphologically analyzed. For example, the text “Seiko Epson Corporation” can be treated as one word “Seiko Epson Corporation” without being divided into the word “Seiko Epson” and the word “stock corporation”. When the hyperlink is extracted, the HTML analysis unit 2 controls the browser control unit 9 to download the page specified by the URL of the extracted hyperlink to the WWW browser 1. This operation is repeated until the number of layers of the page downloaded by the WWW browser 1 reaches a predetermined number of layers. For example, as shown in FIG. 2B, if it is specified in advance to prefetch up to two pages ahead of the current page, first, the WWW browser 1
The pages P2, P3 and P4 linked to are read ahead and downloaded. Then, the WWW browser 1 further prefetches and downloads pages P5, P6, P7, P8 and P9 linked to the downloaded pages P2, P3 and P4, respectively.

(2) Word Extraction When extracting words from an HTML document, first, text is extracted from the HTML documents of all the pages downloaded by the WWW browser 1, and then the extracted text is subjected to morphological analysis. Divide words. When extracting text from an HTML document, in a specific HTML tag, text described immediately after the tag or at a specified position in the tag is extracted. When a text is extracted, positional information (row and column) of the extracted text is stored. HTML tags for text extraction include <TITLE> tags for setting titles and <
A> Set tag, <IMG> tag to display image, set heading <
H> tag, <UL> tag used when creating lists, <O>
L> tag and <DL> tag. Also, ALT of <IMG> tag
Text specified by options is also subject to extraction.
For example, as shown in FIG. 3, when the text “Seiko Epson Corporation” written immediately after the <TITLE> tag is extracted, the extracted text “Seiko Epson Corporation” is subjected to morphological analysis and It is divided into the word “Seiko Epson” and the word “stock”. As in the case of “(1) Extraction of hyperlink”, the user can arbitrarily specify the level at which the text is to be morphologically analyzed. For example, in the example of FIG.
The text “Seiko Epson Corporation” can be used as it is as a single word “Seiko Epson Corporation”.

(3) Associating a Hyperlink with an Extracted Word When associating a hyperlink with an extracted word,
The hyperlink of the page from which the word was extracted is associated with the word extracted from the HTML document in “(2) Extraction of word”. Specifically, a word extracted from the HTML document is provided with a hyperlink based on the URL of the page on which the word is described. At the link destination of the word, an <A NAME> tag indicating a reference within the same page is embedded as link information in the HTML document. For the location where the <A NAME> tag is embedded,
The position information (row and column) of the extracted text stored when the text is extracted in “(2) Extracting words” is referred to. The word to which the hyperlink is assigned is stored in the voice search dictionary 3 together with the URL of the hyperlink assigned to the word. For example, as shown in FIG. 4A, when the word “printer” is extracted from the HTML document, the word “printer” is added to the word “printer”.
The URL “http: //localhost/index.html#printer” is assigned based on the URL of the page in which the word “http: //localhost/index.html#printer” is described. Then, the word "printer" is replaced with "http: // loc
The URL is stored in the voice search dictionary 3 together with the URL “alhost / index.html # printer”. In addition, as shown in FIG. 4B, the link destination of the word "printer"
A tag “<A NAME={Printer}> Printer </A>” is embedded as link information in the L document.

FIG. 5 shows an example of the voice search dictionary 3. The voice search dictionary 3 includes words extracted from an HTML document and URLs of hyperlinks assigned to the words. For example, as shown in FIG. 5, the word “Seiko Epson” is stored in the voice search dictionary 3.
Indicates the link destination of the word "http://www.epson.co.jp
/ "With the URL. Similarly, in the voice search dictionary 3, the word "printer" includes "http: // localhost / ind" indicating the link destination of the word.
ex.html # printer ".

Next, a method for creating a speech recognition task from the created speech search dictionary 3 will be described. When creating a voice recognition task, first, “(2) Word extraction”
In WWW, a group of words extracted from the HTML document of all pages downloaded to the WWW browser 1 is
It is a group of keywords for grasping the contents of the page downloaded to the browser 1. Then, the task control unit 5 creates a speech recognition task using the keyword group and the thesaurus 4 as an external database. For example, as shown in FIG. 6, when the words (keywords) extracted from the HTML document are “printer”, “memory”, “hard disk”, “product”, and “purchase”, refer to the thesaurus 4 Then, the task control unit 5 compares the relevance of those keywords, and creates a voice recognition task of “purchasing a personal computer-related device”.

The thesaurus 4 is a database in which synonyms, synonyms, and the like are collected. The relation between the words extracted by the HTML analysis unit 2 is arranged, the mutual relation between the words and the application range of the words are defined. In the analysis unit 2, the HT
It assists in creating a speech recognition task from a group of words extracted from an ML document. Further, as described later, the thesaurus 4 includes a language model (LM) and an acoustic model (AM) from the language model / acoustic model group 6 in the task control unit 5.
Assist in choosing

The task control unit 5 creates a speech recognition task and, based on the created speech recognition task,
An optimal language model (LM) and acoustic model (AM) for the created speech recognition task are selected from a group of six language models and acoustic models. When selecting the language model (LM) and the acoustic model (AM), the created speech recognition task is compared with a task of the language model / acoustic model group 6 to determine a task having a similar concept or a high degree of association. Select a task. When collating the speech recognition task with the tasks of the language model / acoustic model group 6, the thesaurus 4 is used. For example, as shown in FIG.
Is a "purchase of personal computer-related equipment", and the tasks of the language model / acoustic model group 6 are "place name search task", "hotel reservation task",
When "shopping task" and "computer-related article reading task" were performed, referring to thesaurus 4,
"Purchase PC-related equipment" and "PC-related article reading task" are recognized as similar concepts,
A language model (LM) and an acoustic model (AM) relating to the "PC-related article reading task" are selected. In addition,
In the present embodiment, when the speech recognition task and the task of the language model / acoustic model group 6 are compared with each other, the thesaurus 4 which is an external database is used. The method of matching the task with the acoustic model group 6 is not limited to the method of the present embodiment, and various methods can be used. In addition, a language model (LM) and an acoustic model (AM)
When selecting a speech recognition task group 10 as shown in FIG. 8, a speech recognition task group 10 is prepared in place of the language model / acoustic model group 6. It is also possible to select a task semantically similar to the above. Alternatively, the speech recognition task and the language model (LM) / acoustic model (AM) may be associated with each other in advance. Further, in the present embodiment, the speech recognition task and the language model (LM) / acoustic model (AM) corresponding to the speech recognition task are determined on a one-to-one basis. A language model (LM) and an acoustic model (AM) may be combined.

The language model / acoustic model 6 group is composed of a plurality of language models (LM) and acoustic models (AM), and each language model (LM) / acoustic model (AM) is the content of each speech recognition task. With data suitable for That is, the language model (LM) is created from text data related to a specific field, and the acoustic model (AM) is
It is created from audio data for a specific field. Note that, for the language model (LM) and the acoustic model (AM), a probabilistic / statistical model using statistics and a structural model using grammar can be used. Each language model (LM) has a vocabulary dictionary for speech recognition. In other words, the vocabulary dictionary for speech recognition is a language model (LM)
It is prepared for each. For example, as a language model (LM) / acoustic model (AM) corresponding to a voice recognition task called “PC-related article reading task”, a language model (LM) created from an article related to a personal computer and an article related to a personal computer Model created from the utterance (A)
M). The language model / acoustic model group 6 is
Not only those installed in a local environment such as a PC (personal computer) but also those installed on a network such as the Internet can be used.

The voice recognition unit 7 uses the language model (LM) and the acoustic model (AM) selected by the task control unit 5 to target the page downloaded by the WWW browser 1 and reproduces the voice uttered by the user. recognize. When the voice recognition unit 7 performs voice recognition, first, a voice uttered by the user is analyzed to extract information necessary for voice recognition.
Then, the language model (LM) selected by the voice recognition task
-With reference to the acoustic model (AM), a word string indicating the utterance content is acquired. Note that the word sequence here is a sequence of vocabulary of the language model (LM). Also, the words acquired from the page downloaded by the WWW browser 1 are:
Although the word may not be included in the vocabulary of the language model (LM), the process for such an unknown word is omitted for simplification of description.

The intention analyzing unit 8 analyzes the word string acquired by the speech recognizing unit 7 and detects words indicating the contents of the utterance of the user. Then, based on the analysis result, a page desired by the user is searched using the voice search dictionary 3. Analysis of a word string is performed using natural language processing such as pattern matching, semantic analysis, morphological analysis, and case structure analysis.
Then, the word detected as the analysis result is searched from the voice search dictionary 3, and the page to be moved is determined. For example, in the case of using the pattern matching method, first, from the word string acquired by the voice recognition unit 7, a part “I want to see” or “I want to know” is found,
A word corresponding to the part of "~" is detected from those parts. Then, the detected word is searched from the voice search dictionary 3. The searched page is determined as a page desired by the user, and the page is determined as a page to be moved,
Get the URL of the page.

The browser control unit 9 controls the WWW browser 1 based on the analysis result of the intention analysis unit 8. Specifically, the URL acquired by intention analysis section 8 is
Input to browser 1 and change to page. When moving to the same page as the currently displayed page, instead of newly downloading the page from the network and acquiring the page, the page acquired when analyzing the HTML document is passed to the WWW browser 1. .

Next, a procedure for searching the WWW by voice will be described with reference to a flowchart shown in FIG. First, in step S1, the user sets the WWW browser 1
Enter any URL. The URL is input by the user using a keyboard or a mouse. Further, the URL may be input by the user by voice. In the following step S2, the WWW browser 1 downloads and displays the page indicated by the URL. In step S3, it is determined whether the page displayed in step S2 is a page desired by the user. If the displayed page is the page desired by the user, the operation is terminated. Conversely, if the user desires another page, the process proceeds to the next step S4.

In step S4, the HTML analysis unit 2
An HTML document is acquired from a page currently displayed on the WWW browser 1 and a page linked to the currently displayed page. In the following step S5,
The HTML document acquired in step S4 is analyzed to extract hyperlinks and words. This operation is repeated until the number of layers of the page downloaded by the WWW browser 1 reaches a predetermined number of layers. Then, in step S6, it is determined whether the number of layers of the page downloaded by the WWW browser 1 has reached the number of layers specified in advance. When the number of layers of the page downloaded by the WWW browser 1 has reached the number of layers specified in advance, the process proceeds to the next step S7. Also, WWW
If the number of layers of the page downloaded by the browser 1 does not reach the number of layers specified in advance, the previous step S
Return to 4.

In step S7, the HTML analysis unit 2
The voice search dictionary 3 is created from the hyperlinks and words extracted in the previous step S5. Subsequently, in step S8, the task control unit 5 creates a speech recognition task from the speech search dictionary 3 created in step S7. In step S9, based on the speech recognition task created in step S8, a language model (LM) / acoustic model (A
Select M).

Then, when the user speaks in step S10, in the following step S11, the voice recognition unit 7
Uses the language model (LM) / acoustic model (AM) selected in step S9 to recognize the speech uttered by the user and obtain a word string indicating the uttered content. In the following step S12, it is determined whether or not the voice recognition unit 7 has been able to recognize the utterance of the user. If it is determined that the utterance of the user has been recognized, the process proceeds to the next step S13. If it is determined that the utterance of the user has not been recognized, the process returns to step S10.

In step S13, the intention analyzing unit 8 analyzes the word string acquired in the previous step S11 to detect a word indicating the utterance content of the user. In the following step S14, the word detected in step S13 is searched from the voice search dictionary 3, and the URL of the page to be moved is obtained. Then, in step S15, the browser control unit 9 transmits the U acquired in step S14.
The RL is input to the WWW browser 1 to change the page.

As described above, an embodiment of the method for retrieving voice of a web page according to the present invention has been described. However, the present invention is not limited to the above-described embodiment, and can be widely modified. For example, in the present embodiment, a case has been described in which an HTML document downloaded to the WWW browser 1 is analyzed.
(Extensible Markup Language) or the like. Further, in the present embodiment, the WWW browser program is used to download the web page, but the web page may be downloaded by another method. Further, by storing information of a page once analyzed as a history, it is possible to omit processes such as pre-reading the next and subsequent pages, creating a voice search dictionary, and creating a voice recognition task.

[0037]

According to the present invention, a voice search can be performed not only for a link item of a web page but also for the content of the web page. Therefore, a voice search with a high degree of freedom can be realized. In addition, since a speech recognition task can be created from words extracted from a web page, it is possible to select an optimal language model and acoustic model when performing a speech search on a downloaded web page. Therefore, highly accurate speech recognition can be realized.

[Brief description of the drawings]

FIG. 1 is a functional block diagram showing a system configuration for realizing a web page voice search method according to the present invention.

FIG. 2 is a diagram illustrating a method of extracting a hyperlink.

FIG. 3 is a diagram illustrating a method for extracting words.

FIG. 4 is a diagram illustrating a method of associating a hyperlink with an extracted word.

FIG. 5 is a diagram showing a voice search dictionary.

FIG. 6 is a diagram illustrating a method of creating a voice recognition task.

FIG. 7 is a diagram illustrating a method of selecting a language model / acoustic model.

FIG. 8 is a diagram illustrating another method for selecting a language model / acoustic model.

FIG. 9 is a flowchart illustrating a procedure for performing a voice search on a web page.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 WWW browser 2 HTML analysis part 3 Voice search dictionary 4 Thesaurus 5 Task control part 6 Language model / acoustic model group 7 Voice recognition part 8 Intention analysis part 9 Browser control part 10 Voice recognition task group

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI theme coat ゛ (reference) G10L 15/18 G10L 3/00 537J 15/00 551A 15/28 551P F-term (reference) 5B075 KK07 ND16 PP10 PP14 PP24 PQ02 QM07 UU40 5D015 GG01 KK01 LL10

Claims (10)

[Claims] 1. A voice search method for a web page, comprising: a web page acquisition unit for prefetching and downloading a web page linked to a currently displayed web page by a predetermined number of layers; Word extracting means for extracting a word from all the web pages, linking means for associating the extracted word with the URL of the web page from which the word was extracted, and the word and the URL associated with the word. Dictionary creation means for creating a speech search dictionary from; a task creation means for creating a speech recognition task from the words; a language model / sound model selection means for selecting an optimal language model and an acoustic model for the speech recognition task; A voice recognition means for recognizing a voice uttered by the user, and analyzing and detecting an intention of the uttered content of the user A voice of a web page, comprising: a diagram detection means; a search means for searching a detected intention from the voice search dictionary; and a web page change means for changing a web page based on a search result. retrieval method. 2. The method according to claim 1, wherein the word extracting unit extracts text from all the web pages downloaded by the web page obtaining unit, and then extracts words by morphologically analyzing the text. 1
Voice search method described in. 3. The voice search method according to claim 1, wherein the link assigning unit assigns one or more links to the word extracted from the word extracting unit. 4. The method according to claim 1, wherein the task creating unit compares the word groups extracted from the word extracting unit, and sets a task having a similar concept or a highly related task of the word group as a speech recognition task. 4. A method for retrieving a speech of a web page according to claim 1. 5. The language model / acoustic model selecting means, wherein the speech recognition task created by the task creating means is
5. The task according to claim 1, wherein a task having a similar concept or a task having a high degree of relevance is selected by comparing the task with a task included in a language model / acoustic model group prepared in advance.
Method for retrieving web pages according to any of the above. 6. The language model / acoustic model selecting means, wherein the speech recognition task created by the task creating means is
Collating with a task of a voice recognition task group prepared in advance, and selecting a semantically close task,
The method for retrieving speech of a web page according to any one of claims 1 to 5. 7. The speech recognition device according to claim 1, wherein the speech recognition means dynamically selects the speech recognition task and performs speech recognition when recognizing a speech uttered by the user. Voice search method for web pages described in any of them. 8. The speech recognition apparatus according to claim 1, wherein said speech recognition means performs speech recognition by dynamically combining said speech recognition tasks when recognizing speech uttered by a user. A method for retrieving voice from web pages described in Crab. 9. A web page voice retrieval apparatus, comprising: a web page acquisition means for prefetching and downloading a web page linked to a currently displayed web page by a predetermined number of layers; Word extracting means for extracting a word from all the web pages, linking means for associating the extracted word with the URL of the web page from which the word was extracted, and the word and the URL associated with the word. Dictionary creation means for creating a speech search dictionary from; a task creation means for creating a speech recognition task from the words; a language model / sound model selection means for selecting an optimal language model and an acoustic model for the speech recognition task; A voice recognition means for recognizing a voice uttered by the user, and analyzing and detecting an intention of the uttered content of the user A voice of a web page, comprising: a diagram detection means; a search means for searching a detected intention from the voice search dictionary; and a web page change means for changing a web page based on a search result. Search device. 10. A Web page acquisition means for prefetching a Web page linked to a currently displayed Web page by a predetermined number of layers and downloading the computer in order to perform a voice search for the Web page; Word extracting means for extracting a word from all of the extracted web pages, link providing means for associating the extracted word with the URL of the web page from which the word has been extracted, and the word and the URL associated with the word Dictionary creation means for creating a speech search dictionary from the above; task creation means for creating a speech recognition task from the words; language model / sound model selection means for selecting a language model and an acoustic model optimal for the speech recognition task And voice recognition means for recognizing the voice uttered by the user, and the intention of the uttered content of the user. Intention detecting means for analyzing and detecting, a searching means for searching the detected intention from the voice search dictionary, and a voice search for a web page for functioning as a web page changing means for changing a web page based on a search result. program.