JP2001092845A - Document acquisition method and recording medium - Google Patents

Abstract

(57) [Summary] (with correction) [PROBLEMS] To provide a document acquisition method for recognizing a user's voice and acquiring a next linked document via a network. A dictionary creation step of analyzing an acquired document and registering corresponding word information for associating a next link destination in a speech recognition dictionary, and a process of creating a corresponding word information and an identifier for obtaining a next link destination document. A table creation step of associating and registering in the management table, a voice recognition step of recognizing a voice input by the user, an identifier acquisition step of obtaining an identifier from the management table, and changing the contents of the voice recognition dictionary and the management table And a narrowing-down step of narrowing down the number of identifiers based on the changed voice recognition dictionary and management table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention recognizes speech input from a user and, based on the recognition result, converts a next link destination document from a hypertext or other document obtained through a telecommunication line such as the Internet. The present invention relates to a method for acquiring a document and a recording medium on which the program is recorded.

[0002]

2. Description of the Related Art In recent years, hypertext has been used as a method for easily obtaining information existing on www (Word Wide Web), and access to related information has been realized through a client called a browser.

[0003] The browser requests the server to transmit the information using a unique identifier called a URL.

Hypertext is an HTML (Hyper Te
xt Mark-up Language) is defined by a grammar called the layout and style of objects such as characters and photos, the definition of anchors that are symbols that indicate the existence of links to other information, and the identifier of the link destination Is
It consists of a script that performs correspondence with URLs and specific control at the rank and the like.

[0005] The browser performs each of the above identifications using TAGs, and processes the contents of the range specified by the TAG start signal and end symbol according to the interpretation corresponding to each TAG.

When the browser obtains the HTML text, the browser analyzes the document, creates data for output to the display means based on the TAG, and displays a part or all of the information.

[0007] By the way, the anchor uses a different expression from the others, so that the user can easily identify the anchor. Usually, the font color is changed or an underline is added for display. The link destination information (URL) related to the anchor is also managed, and when the user designates a specific anchor using a pointing device or a keyboard, a browser corresponding to the anchor is used for the UR corresponding to the anchor.
Using L, request information transmission from a server uniquely determined by the URL.

As described above, the user can access related information one after another by designating an anchor. Usually, the system is realized as a program on an information processing apparatus having communication means, distributed through a recording medium or a communication line, and stored in the recording means of the information processing apparatus.

Recently, devices and methods using voice recognition have been proposed as means for specifying an anchor. this is,
This is because it is difficult to secure a place for operating the pointing device in the portable information device, or it is difficult to operate the pointing device.

As a proposal for solving such a problem, there is a conventional example 1 (Japanese Patent Laid-Open No. Hei 10-124293). This Conventional Example 1 will be described with reference to the flowchart of FIG.

In the first conventional example, an HTML document is received (step 401), the HTML document is analyzed, and an anchor and a link destination URL are obtained (step 40).
2) Access the reading dictionary database to create reading data of the anchor character string (step 403), and register the obtained reading data in the recognition dictionary (step 40).
4) A pair of read data and link destination information is created and registered in a table (step 405), and a voice input from the user is waited for (step 406).

When the user reads out the desired anchor,
The voice recognition means processes the input voice signal to create data for recognition, compares the data with data registered in the dictionary for recognition, and gives a score to each. If the threshold value is exceeded, the one with the highest score is output as a recognition result (step 407).

The system searches the above table using the obtained recognition result as a key, acquires link destination information (step 408), and transmits an information request to a specific server using the information (step 409). .

As described above, it is possible to access the related information using voice input, but the user needs to read out the sentence of the anchor itself, and if the anchor is long, it becomes a burden on the user, and The likelihood of misreading increases, which can cause speech recognition to fail.

In order to solve such a problem, a second conventional example (Japanese Patent Laid-Open No. 11-25098) has been proposed. This Conventional Example 2 will be described with reference to the flowchart of FIG.

In the conventional example 2, instead of using the sentence of the anchor as it is, the HTML sentence is analyzed to acquire the anchor (step 411), and the morphological analysis of each anchor is performed. , And only the noun phrases are selected from them (step 412), and the only noun phrase is selected so that the anchor sentence can be determined from the selected noun phrases (step 413), and the reading data is added to the selected word. (Step 414), registration in the voice recognition dictionary (step 415), and creation of a table in combination with the link destination information (step 414).

Thus, the user need only speak the noun phrase instead of reading out the sentence of the anchor, which reduces the load on the user and the occurrence of recognition errors as compared with the previous system.

Further, the second conventional example is characterized in that the character display of the corresponding portion in the anchor is emphasized or the color is changed so that the noun phrase can be easily identified.

However, if only the noun phrase presented by this system is accepted, and if all the noun phrases included in the anchor are used for identifying the previous anchor,
There is a case where noun phrase for solely determining an anchor cannot be selected, a case where no noun phrase exists in the anchor, a case where no noun phrase reading data exists, and a case where no noun phrase exists. It is not possible to cope with problems that occur in actual use, such as reading data of less frequently used names such as person names, place names, and technical terms, which may not be included due to dictionary size restrictions.

[0020]

As described above, conventionally, there has been a problem that only a specific noun phrase can be used, and a problem that a link destination cannot be accessed when a valid noun phrase cannot be selected.

Therefore, when recognizing the voice input from the user and obtaining the next linked document from the document obtained through the telecommunication line based on the recognition result,
Provided is an invention capable of narrowing down to a single identifier even if a plurality of identifiers are obtained based on information instructed by a user by voice.

Further, the present invention provides an invention capable of obtaining an identifier even when the analyzed corresponding word information does not correspond to the identifier.

Furthermore, the present invention provides an invention capable of presenting only necessary information to a user corresponding to the content displayed on the display unit or the content not displayed.

The present invention provides a method for realizing access to a link destination by accepting one or more words, and for realizing access to link destination information even when a word valid for an anchor cannot be selected. The purpose is to:

[0025]

According to a first aspect of the present invention, a voice input from a user is recognized, and a next linked document is obtained from a document obtained through a telecommunication line based on the recognition result. A document acquisition method, comprising: analyzing the acquired document; and registering, in a speech recognition dictionary, corresponding word information on a word to which a next link can be associated, and a correspondence registered in the speech recognition dictionary. A table creation step of creating a management table by associating word information with an identifier for acquiring the next linked document, and inputting by the user based on the corresponding word information registered in the speech recognition dictionary A voice recognition step of recognizing the obtained voice, and an identifier acquisition step of obtaining an identifier from the management table based on the recognition result in the voice recognition step. In the case where there are a plurality of identifiers obtained in the identifier obtaining step, the contents of the speech recognition dictionary and the management table are referred to based on the plurality of identifiers. From among them, a narrowing change step of further changing the number of identifiers so as to be narrowed down, based on the changed voice recognition dictionary and the management table, performing the voice recognition step and the identifier obtaining step, And a narrowing down step for narrowing down the number.

According to a second aspect of the present invention, there is provided a document acquisition method for recognizing a voice input by a user and acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. Analyzing the obtained document,
A dictionary registration step of registering corresponding word information relating to a word to which a next link destination can be associated in a speech recognition dictionary; and a corresponding word information registered in the speech recognition dictionary; and acquiring the next link destination document. A table creation step of creating a management table by associating with an identifier, a speech recognition step of recognizing a speech input by the user based on corresponding word information registered in the speech recognition dictionary, and a speech recognition step A document acquisition method for acquiring an identifier from the management table on the basis of the recognition result in the document acquisition method, wherein a correspondence determination step of determining whether or not the analyzed corresponding word information and the identifier can be associated with each other; If the association is not possible, the data to be registered in the speech recognition dictionary, A document acquisition method characterized in that it comprises a reading information creation step of creating and registering a special reading information on the voice recognition dictionary corresponding to the corresponding word information.

According to a third aspect of the present invention, there is provided a document acquisition method for recognizing a voice input by a user and acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. Analyzing the obtained document,
A dictionary registration step of registering corresponding word information relating to a word to which a next link destination can be associated in a speech recognition dictionary; and a corresponding word information registered in the speech recognition dictionary; and acquiring the next link destination document. A table creation step of creating a management table by associating with an identifier, a speech recognition step of recognizing a speech input by the user based on corresponding word information registered in the speech recognition dictionary, and a speech recognition step A document acquisition method for acquiring an identifier from the management table based on the recognition result in the step (a), wherein a display object for determining whether or not the content is displayed on a screen of a display unit for presenting a document to a user A determination step, and a speech recognition word according to the content displayed on the screen of the display unit or the content not displayed. A document acquisition method characterized by having a display registration changing step of changing the corresponding word information to be registered in.

According to a fourth aspect of the present invention, there is provided a document acquiring method for recognizing a voice input from a user and acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. A recording medium on which a program is recorded, wherein the acquired document is analyzed, and a dictionary registration function for registering corresponding word information about a word to which a next link can be associated in a speech recognition dictionary; A table creation function for creating a management table by associating the registered corresponding word information with an identifier for acquiring the next linked document, and the corresponding word information registered in the speech recognition dictionary. A voice recognition function for recognizing voice input by a user, and an identifier for obtaining an identifier from the management table based on a recognition result in the voice recognition function If a plurality of identifiers obtained by the identifier obtaining function are present in a recording medium on which a program for realizing the obtaining function is recorded, the contents of the speech recognition dictionary and the management table are determined based on the plurality of identifiers. From among a plurality of identifiers, a narrowing-down change function for further changing the number of identifiers so as to be narrowed down, and performing the voice recognition function and the identifier acquisition function based on the changed voice recognition dictionary and management table. , A program for realizing a narrowing-down function for narrowing down the number of identifiers is recorded.

According to a fifth aspect of the present invention, there is provided a document acquisition method for recognizing a voice input by a user and acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. A recording medium on which a program is recorded, wherein the acquired document is analyzed, and a dictionary registration function for registering corresponding word information about a word to which a next link can be associated in a speech recognition dictionary; A table creation function for creating a management table by associating the registered corresponding word information with an identifier for acquiring the next linked document, and the corresponding word information registered in the speech recognition dictionary. A voice recognition function for recognizing voice input by a user, and an identifier for obtaining an identifier from the management table based on a recognition result in the voice recognition function In a recording medium on which a program for realizing the acquisition function is recorded, a correspondence determination function for determining whether or not the analyzed corresponding word information and the identifier can be associated with each other. Data for registering in a recognition dictionary, and a reading information creation function of creating special reading information corresponding to the analyzed corresponding word information and registering the special reading information in the speech recognition dictionary. This is a recording medium for a document acquisition method that is a feature.

According to a sixth aspect of the present invention, there is provided a document acquisition method for recognizing a voice input by a user and acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. A recording medium on which a program is recorded, wherein the acquired document is analyzed, and a dictionary registration function for registering corresponding word information about a word to which a next link can be associated in a speech recognition dictionary; A table creation function for creating a management table by associating the registered corresponding word information with an identifier for acquiring the next linked document, and the corresponding word information registered in the speech recognition dictionary. A voice recognition function for recognizing voice input by a user, and an identifier for obtaining an identifier from the management table based on a recognition result in the voice recognition function In a recording medium on which a program for realizing the acquisition function is recorded, a display target determination function for determining whether or not the content is displayed on a screen of a display unit for presenting a document to a user; And a display registration change function for changing corresponding word information registered in the speech recognition dictionary in accordance with the content being displayed or the content not being displayed. It is.

According to the first and fourth aspects of the present invention, even if a plurality of identifiers are obtained based on the corresponding word information instructed by the user by voice, the contents of the management table are changed and the contents are changed according to the changed contents. By having the user input the voice again, the identifier can be further narrowed down.

According to the second and fifth aspects of the present invention, even if the analyzed corresponding word information and the identifier do not correspond to each other, special reading information is created, and if this special reading information is indicated by a voice by the user, An identifier corresponding to the special reading information can be obtained.

According to the third and sixth aspects of the present invention, only necessary information is presented to the user in order to register the corresponding word information corresponding to the content displayed on the display unit or the content not displayed. it can.

[0034]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (First Embodiment) A first embodiment of the present invention will be described with reference to FIGS.

1. Configuration FIG. 2 is an example of a block diagram illustrating a configuration of the information processing apparatus according to the first embodiment.

As shown in FIG. 2, it is realized as a program that operates on an information processing apparatus including an arithmetic unit 1, a display unit 2, a storage unit 3, a communication unit 4, a keyboard 5, an instruction unit 6, and a voice input unit 7. I do. CD-ROM
It is assumed that the data is read into the processing device via a recording medium such as a FD or an FD, or the communication unit 6.

The display means 2 comprises a video memory for temporarily storing display contents, a DA converter, and a CRT or liquid crystal display for presenting the display contents to the user.

As the storage means 3, a device such as a RAM used for temporarily storing data and a hard disk for storing data permanently are used. The RAM stores work variables such as this program, dictionary data for recognition, and link management tables. The hard disk stores a cache of received HTML sentences, a phoneme data dictionary that is a dictionary for speech recognition, and a reading data dictionary. Record.

The communication means 4 is a wired LA using a cable.
It is realized by N, RS232C, Centronics, SCSI, USB, IEEE1394, etc., or wireless LAN, infrared communication, PHS data communication, etc. without using cables. For example, communication is performed through the Internet.

The indicating means 6 includes a mouse, a trackball,
It is realized by accu points, pen input on display means, etc.

The audio input means 7 is connected to a microphone or an external amplifier, and converts an input audio signal into a digital signal through AD conversion. The input data is stored in the memory.

The computing means 1 is also called a CPU and is connected to the above-mentioned means via the bus 9 to transmit and receive data. It also serves as voice recognition means.

The process of analyzing the HTML text and displaying it on the screen is possible by using a browser module that is made into components.

2. Processing FIG. 1 shows a flowchart as an example of an embodiment of the first embodiment.

This program acquires the designated HTML text from the server 8 by the communication means 6 through an electric communication line such as the Internet, and records it in the memory (step 101).

Next, the HTML text is analyzed based on the TAG information (step 102).

FIG. 3 shows Htm for explaining the present embodiment.
1 shows an example of a sentence. Note that TAG information not related to the present embodiment is not shown for simplification of the description.

For example, the analysis result of <A href="http://www.zzz.com.news/news1.htm">today's top </A> described in the example shows that the anchor is "today's Is the top of
The link identifier (URL) is "http://www.zzz.com.new
s / news1.htm ".

The extraction of each data can be realized by searching a character string for symbols such as <>"".

Next, the anchor character string is compared with dictionary data including grammar and part of speech information, and morphological analysis is performed. Taking "Today's top" as an example, it is divided into "Today", "of", and "Top". If only a noun phrase is selected therefrom, the words "Today" and "Top" can be selected as keywords (step 103). ).

The words "today" and "top" selected in step 103 are temporarily recorded in the work area, compared with the reading dictionary of words and phrases, and the reading data of each word, "honjintsu" and "top". Is acquired (step 104).

The reading data obtained in step 104 is sequentially added to the data of the dictionary for voice recognition (step 10).
5). The data of the dictionary for voice recognition is Html to be displayed.
If the text is changed, delete it all. Also, even if the same word is registered, since it is meaningless, a duplication check is performed before registration. This records the words registered in the dictionary in the work area,
This is realized by searching using the word to be registered as a key.

FIG. 4 shows word data to be registered in the dictionary for speech recognition.

"Hontsutsu", "Top", "Niisu", "Koi", "" obtained by analyzing the HTML sentence
In addition to keywords such as "Igai", words for system control are also registered. The role of the control words will be described later.

After obtaining the read data, the read data, the identifier, and the anchor character string are registered as a set in the management table (step 106). FIG. 5 shows information registered in the management table.

Registration in the management table can be performed later at a high speed by registering the read data as a key. Note that an index corresponding to the read data may be registered without registering the read data.

The procedure from step 103 to step 106 is executed for the number of noun phrases in the anchor. After checking all noun phrases, return to step 102 and
Continue parsing the sentence.

After analyzing all the contents of the HTML text, a command to activate the voice recognition means of the arithmetic means 1 is transmitted. As a result, the system enters a standby state until the user speaks (step 107).

The outline of the method of implementing speech recognition in the speech recognition means of the arithmetic means 1 will be described.

As an example, a hidden Markov model is used.

When a signal is input from the voice input means 7 by the user's utterance, the signal is recorded in the memory as digital data via AD conversion, and a signal corresponding to a predetermined time is converted into a frequency signal by Fourier transform. And create feature data.

Further, a phoneme model of each word is created from the words registered in the dictionary for speech recognition and the phoneme model, and the distance between the word and the feature data string composed of the plurality of data is calculated. Ask for.

If the calculation result exceeds a predetermined threshold, the word with the best score is output as the recognition result. If the threshold is not reached, an error is output.

If the recognition result is an error, step 1
Returning to step 07, the apparatus enters a standby state.

If the recognition is successful, the above-mentioned table is searched using the word of the recognition result as a key (step 1).
09).

In the case where there is one identifier When the result of the search is that there is only one identifier, a command for requesting information using the obtained identifier is transmitted through the communication means 6 (step 110).

For example, when the user speaks “today”,
The information "honjintsu" is obtained from the voice recognition means. Searching the table using this as a key gives the identifier "http://www.zzz.com.news/news1.htm".

As a result, the next linked HTML document can be obtained.

When there are a plurality of identifiers When there are a plurality of identifiers, steps 111 to 116 are executed.

For example, when the user utters "news", the information "Nice" is obtained from the voice recognition means. When the management table is searched using this as a key, two identifiers, "http://www.zzz.com.news/news2.htm" and "http://www.zzz.com.news/news3.htm" Is obtained.

Using the obtained two identifiers as keys, other read data corresponding to these identifiers is obtained so that the two identifiers can be narrowed down to one identifier. Then, the contents of the dictionary are updated with the read data excluding the read data obtained by speech recognition from the obtained read data (step 111).

More specifically, since the above two identifiers are identifiers obtained by "Nice", read data other than "Nice" is used to further narrow down the information. In other words, the read data of “don't come”, “ni-su”, “kaiai”, “ni-su” is obtained,
"Nice" is a recognized word and is not used.Before changing the dictionary, a speech recognition invalidation signal is transmitted.

When the anchor is "news" only, if the above algorithm is applied, other valid read data will be lost. This problem can be avoided by assigning a serial number to each combination to prevent this and using the reading of the serial number.

As shown in FIG. 6, the dictionary contains "Kako" and "Koi" obtained by a table search using an identifier as a key.
In addition to the word "kaigai", the words "most" and "niban" are added.In this embodiment, the serial numbers are assigned, but alphabets and the like may be assigned.

The contents of the management table are updated by combining the read data, the identifier, and the anchor character (step 1).
12).

As shown in FIG. 7, in the management table, in addition to the originally obtained combination of the reading and the identifier, the combination with the reading data of the number is also registered.

FIG. 8 is an example of a screen shown on the display means 2 for prompting the user to select a narrowing down. By listing the overlapping options with numbers, the correspondence between the anchors and the numbers can be understood. (Step 11
3).

The above steps 111 to 113
After executing the above, a command for enabling the voice recognition means is transmitted. This puts the system in a standby state until the user speaks.

Hereinafter, steps 114 to 117 will be described.
Performs the same processing as steps 107 to 109. That is, the user utters again based on the screen prompting the selection of narrowing down, and one identifier is narrowed down from a plurality of identifiers based on the content of the voice data again and the content of the changed dictionary.

If a plurality of identifiers are obtained again, the procedure returns to step 111 and the same procedure is repeated to finally select only one identifier.

When the recognition result is a control command When the recognition result is a control command "return" or "home",
Executes processing corresponding to each instruction.

In the case of "return", the UR acquired immediately before
A processing procedure for displaying the HTML sentence corresponding to L again is executed.

In the case of "home", if the program is displayed first when the program is executed, a preset HTML text is displayed again. After completing the acquisition and display of the HTML text, the processing procedure is continued from step 101.

If the recognition result is "cancel" in step 115, the process returns to step 102. In this case, the user speaks when the speech recognition result is wrong and the user falls into a situation that is not desired. Whether the recognition result is wrong can be understood by the user by the anchor displayed on the narrowing-down screen.

3. Modification of First Embodiment In the first embodiment, the reason for not only updating the content of the management table but also updating the dictionary for speech recognition is that the possibility of erroneous recognition occurring due to the reduction of recognition candidates is high. This need not be changed if the performance of the recognition means is good. In other words, step 111 may be omitted.

[0086] 4. Effects of the First Embodiment Selecting only a noun phrase from an anchor requires an operation that cannot be a user, and it is considered effective to change the display method to another. However, JP-A-11-250
The highlighting and the change of the font color described in Japanese Patent No. 98 become a problem when the anchor is originally highlighted or a color considering the background color is added to the anchor. The method of giving the phonetic mark may significantly impair the original screen layout when the read data is long.

Therefore, in this embodiment, a method of inserting a specific symbol before or after a noun phrase is used. This method is advantageous in that the layout is not significantly impaired since there is little data to be newly inserted, which is not affected by the expression method, as compared with the conventional example.

As described above, according to the first embodiment,
The user can access the link destination information by speaking a desired keyword in the anchor. In addition, even when the uttered word is included in a plurality of anchors, it is possible to cope by having steps 111 to 116 for narrowing down.

(Second Embodiment) A second embodiment of the present invention will be described with reference to FIGS.

This explanation is based on the first embodiment, and only different points from the first embodiment will be described.

Although the first embodiment is used for the description, it is not necessary to satisfy all the necessary conditions for realizing the first embodiment. An anchor in which no noun phrase having reading data exists in the anchor is referred to as “exception data”.

Step 2 in the flowchart of FIG.
After executing the morphological analysis of the 03 anchor sentence, the number of noun phrases is measured. If the number of noun phrases is 0, the process of step 2111 is performed. If the number of noun phrases is one or more, after performing the processing from step 204 to step 206 for the number of existing noun phrases, measurement of noun phrases having valid reading data is performed. If the measured result is 0, the process of step 211 is performed in the same manner as described above.

In step 211, the identifier and the anchor information are associated with each other and registered in the exception management table. Also,
Change the contents of the anchor information displayed in the browser. Thus, the user knows that special processing is required for the anchor.

An example of a method of specifying exception data by voice will be described.

A specific word in advance, in this case, temporarily “not considered”
Assume that If the user utters “no reading” in the voice input waiting state in step 207 and obtains “not considered” information from the recognition result, exception processing is determined in step 7. When it is determined that the instruction is “not regarded”, the processing from step 213 to step 217 is executed.

In step 213, all the contents of the dictionary for speech recognition are deleted, and the read data of the serial numbers corresponding to the number of the exception data are registered in this dictionary. Also, a control command such as "cancel" is added in the same manner as in the first embodiment.

In step 214, as shown in FIG. 10, a serial number corresponding to the beginning of the anchor of the exception data is assigned, and a selection screen is presented on the display means 2. Specifically, in FIG. 10, the user utters the serial number "1" to indicate "no leakage".

At step 216, when the serial number data (special reading data) is obtained by the voice recognition means,
The identifier of the exception data corresponding to the number is selected (step 217).

Thereafter, a request for link information is transmitted using the identifier selected in step 210.

In the present embodiment as described above, if no noun phrase having valid reading data exists in the anchor, it is determined whether or not the data is exceptional data. This can be dealt with by preparing special reading data (serial number in the above example), which is the selection means.

In this embodiment, the serial number is used for identifying exception data. However, a method using alphabets, Japanese syllabary, special symbols, and the like may be used.

As another designation method, when it is determined in step 203 that there is no valid registration, the display content is changed by adding a serial number to the beginning of the anchor, and the reading data of the number is stored in the speech recognition dictionary. A method of registering and registering the reading data, the link identifier, and the anchor correspondingly to the management table is also conceivable.

(Third Embodiment) A third embodiment of the present invention will be described with reference to FIGS.

In the third embodiment, when a plurality of noun phrases having reading data in one anchor exist in the present invention, the number of words registered in the recognition dictionary increases, and the possibility of erroneous recognition increases. It concerns the problem.

In some cases, the display effective area of the display means 2 is smaller than the display screen created by analyzing the HTML text. In this case, the browser displays a scroll bar on the screen and enables access to a screen that is not included in the effective area.

From the user's point of view, anchors whose display area is narrow and have not been displayed at all are not considered as targets for access. Therefore, registering the reading data included in the non-displayed anchor in the dictionary for recognition increases the possibility of erroneous recognition and is not useful.

The processing procedure according to the third embodiment will be described with reference to the flowcharts of FIGS. 1 and 12 and only the differences from the first embodiment. Although the first embodiment is used for the description, it is not necessary to satisfy all the necessary conditions for realizing the first embodiment.

The registration in the dictionary for speech recognition in step 105 in FIG. 1 is not executed in the third embodiment. In addition, the management table of step 106 includes the read data, the identifier,
In addition to the anchor, the display start x-coordinate and y-coordinate, the width of the anchor, and the height of the anchor on the display unit 2 are also registered as shown in FIG.

As the information on the display of each anchor, the information calculated when the display screen is created is used. Or Document
It can also be obtained from a browser using a technology called Object Model.

When the user instructs the browser to change the display area of the display means 2 by operating the scroll bar in the state of waiting for voice input in step 107 in FIG.
The procedure shown in FIG. 12 is performed.

In step 301, information on the size of the current effective display area of the display means 2 and which area is being displayed is obtained. Here, it is assumed that the size of the area is assumed to be width 320 and height 400, and information of width 320 and height 400 from (0, 140) in the created screen data is presented. This is an area of (0,140)-(319,539) when represented by the coordinates of the created screen.

In step 302, all the contents registered in the dictionary for voice recognition are deleted.

Steps 303 and 304 are executed by the number of items recorded in the management table. In step 303, the content of the Nth element is obtained, and it is determined whether the anchor is currently displayed.

In the example, items 1 and 2 correspond to the area (100, 48).
0)-(195, 495), and is included in the valid display area, so that it is determined that it is displayed. Similarly, it is determined that items 3 and 4 are displayed, but it is determined that items 5 and 6 are not displayed because they are outside the effective area.

In step 304, only the reading data of the item determined to be displayed in step 303 is registered in the dictionary for speech recognition.

Of course, the procedure shown in FIG. 12 is executed even when data is first displayed on the screen. Accordingly, it is possible to cope with the case where the screen change instruction is not issued.

With the above processing, the reading information registered in the dictionary for voice recognition can be minimized, and the possibility of erroneous recognition can be reduced.

In the present description, the word of the anchor included in the displayed area is registered in the dictionary for speech recognition, but a method of registering the word once displayed may be considered.

In this case, this can be realized by adding an item of a flag indicating whether the data is already displayed or not to the data to be registered in the table shown in FIG. .

Further, contrary to the above description, the word of the anchor included in the non-displayed area may be registered in the dictionary for speech recognition.

According to this, when the user turns the screen and stops displaying the page, the identifier can be narrowed down without displaying the page again.

[0122]

According to the first and fourth aspects of the present invention, even if a plurality of identifiers are obtained based on the corresponding word information instructed by the user by voice, the contents of the management table are changed and the contents are changed. By having the user input the voice again according to the content, the identifier can be further narrowed down.

According to the second and fifth aspects of the present invention, even if the analyzed corresponding word information and the identifier do not correspond to each other, special reading information is created, and if this special reading information is indicated by a voice, An identifier corresponding to the special reading information can be obtained.

According to the third and sixth aspects of the invention, only necessary information is presented to the user in order to register the corresponding word information corresponding to the content displayed on the display unit or the content not displayed. it can.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a processing procedure according to a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of the information processing apparatus according to the first embodiment.

FIG. 3 is an explanatory diagram showing an example of an HTML sentence.

FIG. 4 is an explanatory view showing dictionary data for speech recognition.

FIG. 5 is an explanatory diagram showing registration of read data and an identifier in a management table.

FIG. 6 is an explanatory diagram showing data of a dictionary for speech recognition after a change.

FIG. 7 is an explanatory diagram showing the contents of registration of read data and identifier management tables after change.

FIG. 8 is an explanatory diagram of a narrowing-down candidate screen.

FIG. 9 is a flowchart illustrating a processing procedure according to the second embodiment;

FIG. 10 is an explanatory diagram of an exception data selection screen.

FIG. 11 is an explanatory diagram showing registration of reading data, an identifier, and coordinate data in a management table.

FIG. 12 is a flowchart illustrating a processing procedure according to a third embodiment;

FIG. 13 is a flowchart showing a processing procedure of Conventional Example 1.

FIG. 14 is a flowchart illustrating a processing procedure of Conventional Example 2.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Calculation means 2 Display means 3 Storage means 4 Keyboard 5 Instruction means 6 Communication means 7 Voice input means 8 Server 9 Bus

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (Reference) G10L 15/00 G10L 3/00 551P 5K034 15/28 H04L 13/00 307Z 9A001 H04L 29/08 F-term (Reference) 5B009 KB00 5B075 PP07 5B082 AA11 EA07 GC04 5B089 GA25 GB03 HA10 JA24 KA03 KB07 KC15 KH16 5D015 GG01 GG03 KK02 LL10 5K034 AA18 BB06 FF01 FF17 HH01 HH02 HH14 HH17 HH26 LL01 9A001 BB03 JJ03HJ BB03

Claims (6)

[Claims] 1. A document acquisition method for recognizing a voice input by a user and acquiring a next linked document from a document acquired through a telecommunication line based on a result of the recognition. A dictionary registration step of registering corresponding word information on a word that can be associated with a next link destination in a speech recognition dictionary; anda corresponding word information registered in the speech recognition dictionary; A table creation step of creating a management table by associating with an identifier for obtaining, and a speech recognition step of recognizing a speech input by the user based on corresponding word information registered in the speech recognition dictionary, An identifier obtaining step of obtaining an identifier from the management table based on a recognition result in the voice recognition step. In the case where there are a plurality of identifiers obtained in the identifier obtaining step, the contents of the speech recognition dictionary and the management table are extracted from the plurality of identifiers based on the plurality of identifiers, and the number of identifiers is further increased. And a narrowing-down step of narrowing down the number of identifiers by performing the voice recognition step and the identifier acquisition step based on the changed voice recognition dictionary and management table. A document acquisition method characterized in that: 2. A document acquisition method for recognizing a voice input from a user and acquiring a next linked document from a document acquired through a telecommunication line based on the result of the recognition. A dictionary registration step of registering corresponding word information on a word that can be associated with a next link destination in a speech recognition dictionary; anda corresponding word information registered in the speech recognition dictionary; A table creation step of creating a management table by associating with an identifier for obtaining, and a speech recognition step of recognizing a speech input by the user based on corresponding word information registered in the speech recognition dictionary, An identifier obtaining step of obtaining an identifier from the management table based on a recognition result in the voice recognition step. In addition, a correspondence determination step of determining whether or not correspondence between the analyzed corresponding word information and an identifier is possible; and, if the correspondence is not possible, data for registering in the voice recognition dictionary. A step of generating special reading information corresponding to the analyzed corresponding word information and registering the special reading information in the speech recognition dictionary. 3. A document acquisition method for recognizing a voice input from a user and acquiring a next linked document from a document acquired through a telecommunication line based on the result of the recognition. A dictionary registration step of registering corresponding word information on a word that can be associated with a next link destination in a speech recognition dictionary; anda corresponding word information registered in the speech recognition dictionary; A table creation step of creating a management table by associating with an identifier for obtaining, and a speech recognition step of recognizing a speech input by the user based on corresponding word information registered in the speech recognition dictionary, An identifier obtaining step of obtaining an identifier from the management table based on a recognition result in the voice recognition step. A display target determining step of determining whether or not the content is displayed on a screen of a display unit that presents a document to a user; and a content displayed on the screen of the display unit, or A display registration change step of changing corresponding word information registered in the speech recognition dictionary according to the content that does not exist. 4. A recording which records a program for realizing a document acquisition method for recognizing a voice input from a user and acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. A medium, wherein the acquired document is analyzed, and a dictionary registration function for registering corresponding word information on a word to which a next link destination can be associated in a speech recognition dictionary; and corresponding word information registered in the speech recognition dictionary. And a table creation function for creating a management table by associating the identifier with the identifier for acquiring the next linked document; and a voice input by the user based on corresponding word information registered in the voice recognition dictionary. A speech recognition function for recognizing, and an identifier acquisition function for obtaining an identifier from the management table based on a recognition result in the speech recognition function. In the recording medium on which the program to be recorded is recorded, if there are a plurality of identifiers obtained by the identifier acquisition function, the contents of the speech recognition dictionary and the management table are updated based on the plurality of identifiers. From inside
Further, a narrowing change function for changing the number of identifiers so as to be narrowed down, based on the changed voice recognition dictionary and the management table, performing the voice recognition function and the identifier acquisition function,
A recording medium for a document acquisition method, characterized by recording a program for realizing a function of narrowing down the number of identifiers and 5. A recording device for recognizing a voice inputted by a user and recording a program for realizing a document acquisition method for acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. A medium, wherein the acquired document is analyzed, and a dictionary registration function for registering corresponding word information on a word to which a next link destination can be associated in a speech recognition dictionary; and corresponding word information registered in the speech recognition dictionary. And a table creation function for creating a management table by associating the identifier with the identifier for acquiring the next linked document; and a voice input by the user based on corresponding word information registered in the voice recognition dictionary. A voice recognition function for recognizing the information and an identifier acquisition function for obtaining an identifier from the management table based on the recognition result in the voice recognition function are realized. And a correspondence determination function of determining whether or not the analyzed corresponding word information and the identifier can be associated with each other on a recording medium on which a program for realizing the above is recorded. Data for registering in a dictionary, and a reading information creating function for creating special reading information corresponding to the analyzed corresponding word information and registering the special reading information in the speech recognition dictionary. Recording medium of the document acquisition method to be described. 6. A recording device for recognizing a voice input by a user and recording a program for realizing a document acquisition method for acquiring a next linked document from a document acquired through a telecommunication line based on the recognition result. A medium, wherein the acquired document is analyzed, and a dictionary registration function for registering corresponding word information on a word to which a next link destination can be associated in a speech recognition dictionary; and corresponding word information registered in the speech recognition dictionary. And a table creation function for creating a management table by associating the identifier with the identifier for acquiring the next linked document; and a voice input by the user based on corresponding word information registered in the voice recognition dictionary. A voice recognition function for recognizing the information and an identifier acquisition function for obtaining an identifier from the management table based on the recognition result in the voice recognition function are realized. And a display medium determining function for determining whether or not the content is displayed on a screen of a display unit for presenting a document to a user on a recording medium on which a program for realizing the program is realized. And a display registration change function for changing corresponding word information registered in the speech recognition dictionary in accordance with contents present or not displayed, and a program for realizing the following.