JP2001109747A - Translation device and recording medium with translation program recorded thereon - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain translation results which can easily be understood even about a sentence where a plurality of areas coexist and a proper noun is included without increasing a burden on a user. SOLUTION: A computer (C) executes information update processing, homepage translation display processing, etc., by executing a translation program. The information update processing acquires area selection information B1 in which proper nouns representing various areas described in the sentence are registered in each area and dictionary information B2 including a specified dictionary in each area from a computer (B) and always updates the latest contents. The homepage translation display processing divides the sentence being a translation object, extracts proper nouns from each sentence and decides the area of contents described in the sentence by referring to the area selection information. Then, each sentence is translated by using a specified dictionary corresponding to the decided area.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translator for translating a sentence in one language into a sentence in another language, and a recording medium on which a translation program is recorded.

[0002]

2. Description of the Related Art In recent years, a program for mechanically translating a sentence in a certain language into a sentence in another language by a personal computer (hereinafter referred to as a translation program) has been used. Among them, with the spread of the Internet, the WWW (World Wide Web)
Some websites automatically translate and display the homepage downloaded from the server.

[0003]

However, in machine translation using a conventional translation program, one word usually has many completely different meanings (translated words), and many different meanings thereof. In many cases, it was not possible to select a meaning appropriate for the text to be translated from among them. In particular, sentences used in specific sports, industries, academic fields, etc. may use common words with completely different meanings, so if you translate a sentence in such a field, The translated text that was output tended to be meaningless.

[0004] In order to avoid such unknown meaning, there has been a method in which a specialized dictionary is set in advance. For example, a specialized dictionary suitable for baseball is created in advance, and when translating a sentence of baseball, this specialized dictionary is designated in advance. Thus, when selecting a translation of a word, the translation program can perform correct translation by preferentially using a specialized dictionary over a basic dictionary in which a normal meaning is registered.

However, if an ordinary sentence is translated while the baseball specialty dictionary is set, a meaningless translation is performed on the contrary. In order to avoid these mistakes, it is necessary to analyze the context before and after using a translation program to detect that it is not a topic of baseball, but the processing load of the translation program increases.

[0006] Translation programs are often used when browsing foreign language homepages on the Internet. In such a case, it is very difficult for the user to switch the specialized dictionary for each homepage content. Annoying and unrealistic. For example, if you are watching a site news such as a newspaper, you may find that there are a mix of texts in various fields such as political articles, investment articles, sports articles, science articles, etc. Every time you look at an article, you have to switch the specialized dictionary, and even for the same sports article, you have to switch depending on the event, such as baseball, soccer, or basketball.

[0007] Furthermore, it is another case that a person who has the ability to understand a foreign language uses a translation program as an aid, but in the first place, he does not know the foreign language used in the article, so he uses the translation program. Some people find it difficult to know what the article is about before translating it, so it is not at all practical to switch between specialized dictionaries before translating.

[0008] As described above, in the conventional machine translation, a person who hardly understands a target language may not be able to output an appropriate translated word, particularly in a case of viewing homepages in various fields on the Internet, which may cause a problem. There were many.

[0009] The present invention has been made in view of the above-described problems, and does not increase the burden on the user.
It is an object of the present invention to provide a translation device in which a plurality of fields are mixed and a translation that can easily understand a translation result even for a sentence including a proper noun, and a recording medium on which a translation program is recorded.

[0010]

The present invention provides a proper noun extracting means for extracting a proper noun from a text to be translated,
Proper nouns representing various fields described in the text are referred to the field selection information registered for each field, and the text to be translated based on the proper noun extracted by the proper noun extracting means. Field determination means for determining the field of the content described in the field, corresponding to each of the fields determined by the field determination means, a specialized dictionary used for translation processing, according to the field determined by the field determination means A translation unit that translates the sentence to be translated using the specialized dictionary.

[0011] The present invention is characterized in that an information updating means for acquiring the field selection information and the specialized dictionary from the outside via a network and automatically updating the contents is provided.

The text to be translated is a text described on a page obtained by WWW,
Is divided into sentences for each field based on tags, and translation is executed by applying a specialized dictionary to each of the divided sentences.

[0013]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an example of a translation system using the translation device according to the present embodiment.
The translation device according to the present embodiment is configured by a computer which reads a translation program recorded on various recording media and whose operation is controlled by the read translation program. In FIG. 1, the computer (C)
Executes the translation program C1 to realize a translation device.

The computer (C) is connected to another computer (A) via a network such as the Internet.
(B) can be communicated. The computer (A) holds the contents of an English homepage (English article information A1) that the user using the computer (C) intends to download and, based on a request from the computer (C), downloads the contents. (C)
Send to The computer (C) has a translation program C1 to which the present invention is applied.
Is executed, the English homepage sent from the computer (A) is automatically translated and displayed on the display device. Computer (B) is a translation program C
1 that has the latest field selection information B1 and dictionary information B2 used in machine translation accompanying the execution of computer 1
Is sent to the computer (C). The field selection information B1 is referred to when determining the field of the text to be translated, and a proper noun representing each field is registered. The dictionary information B2 includes field selection information B1 in addition to a basic dictionary for performing general translation.
Specialized dictionaries corresponding to each of the fields identified by are included. Details of the field selection information and the dictionary information will be described later. The computer (B) uses the field selection information B1.
And the dictionary information B2 are constantly updated to the latest contents. The computer (B) obtains necessary information from another computer (server) via the Internet, for example, and obtains the field selection information B1 and the dictionary information B2.
Has been updated.

FIG. 2 is a block diagram showing a configuration of the translation apparatus realized by the computer (C). As shown in FIG. 2, the translation device (computer (C)) has a CPU
10, a memory 12, a display device 14, an input device 16, a communication device 17, and a storage device 18.

The CPU 10 controls the entire apparatus, activates a program stored in a program area 12a of the memory 12 in response to a key operation signal from the input device 12, and realizes various functions according to the program. I do. For example, browsing a homepage by a browser program stored in the program area 12a (browser)
The function is executed, and the machine translation function is executed by the translation program 12b. In addition, the CPU 10 includes the storage device 1
The program code recorded on a storage medium (not shown) can be read by the program memory 8 and stored in the program area 12a for execution.

The memory 12 has a program area 12a for storing system programs for controlling the entire translation apparatus, programs corresponding to various functions (browser program, translation program 12b, etc.), and information downloaded via the Internet. Information buffer 12c for temporarily storing the information being downloaded, the type name 12 of the information being downloaded indicating the type of the information being downloaded.
d, an area for storing data such as a download information version number 12e indicating the version of the information being downloaded and a downloaded data amount 12f indicating the amount of downloaded data stored in the download information buffer 12c is required. Stored accordingly.

The display device 14 is used to display the contents of various data when executing a homepage browsing function, a machine translation function, or the like. The input device 16 is for inputting instructions and data that define the operation of the translation device (computer (C)), and is constituted by a pointing device such as a keyboard and a mouse. The communication device 17
When the homepage is browsed via the Internet by the homepage browse function, communication with the outside is performed in order to obtain field selection information, dictionary information, and the like used in the machine translation function via the Internet.

The storage device 18 has a recording medium,
Various programs (browser program, translation program, etc.) for this recording medium, field selection information 18a obtained from another computer (B) via the Internet used by the machine translation function, dictionary information 1
Various data such as 8b (see FIG. 3 described later) are recorded. The storage medium is a magnetic or optical storage medium or a semiconductor memory.
To be fixedly mounted on the camera, or to be detachably mounted. The program, data, and the like stored in the storage medium may be configured to be received and stored from another device connected via a communication line or the like. A storage device provided with a storage medium may be provided on the device side, and a program and data stored in the storage medium may be used via a communication line.

Next, the operation of the translation device (computer (C)) in the present embodiment will be described. here,
An example in which English is translated into Japanese will be described. FIG.
Shows an example of information used when the computer (C) executes machine translation. Computer (C)
Obtains the information shown in FIG. 3 from the computer (B) via the Internet, that is, the field selection information 18a and the dictionary information 18b, and obtains the storage device 18 (such as a hard disk).
Is stored in Note that the downloading information buffer 12c, downloading information type name 12d, downloading information version number 12 shown in the memory 12 of FIG.
e, the data of the downloaded data amount 12f is also interrupted when the information update process described later is interrupted or when the translation program 1
When 2b is terminated, it is stored in the storage device 18.

In the field selection information 18a, proper nouns representing various fields are registered for each field in order to determine which field the text to be machine-translated relates to. For example, in the field of baseball, proper nouns such as names of all registered players, names of all teams, names of all stadiums, names of all coaches, names of all coaches, and names of all owners of the major league are registered. Usually, English baseball information translated into Japanese is
This is almost enough, since most of the major leagues in the United States are concerned. If necessary, the names of minor league teams and famous players, the names of famous professional baseball players in Japan, and in the year of the Olympics, the names of representative players in each country are also registered. In the field of soccer, the names of players, managers, teams, etc. of famous world leagues are registered. In the field of politics, the names of politicians such as heads of countries and international organizations are registered.

The dictionary information 18b includes, in addition to the basic dictionary 18b1, a specialized dictionary corresponding to each field determined by the field selection information 18a, for example, a baseball specialized dictionary 18b2,
A football specialized dictionary 18b3, a political specialized dictionary 18b4, and the like are included. In the basic dictionary 18b1, words and general translations of the words, as well as additional information such as parts of speech necessary for translation are registered. Baseball dictionary 18b2
Is a word that is unique to baseball and a word that is an ordinary word but has a meaning that is unique to baseball, along with its translation. Further, the field selection information 1 is stored in the baseball specialized dictionary 18b2.
A proper noun related to baseball and its translation registered in 8a are registered, and only the last name of a person registered on behalf of baseball and its translation are registered in the field selection information. Although not registered in the field selection information 18a, other proper nouns that may be used in sentences in the field of baseball and their translations may also be registered. Soccer Dictionary 18
The same applies to b3 and the political specialized dictionary 18b4. Although only three specialized dictionaries are shown for the sake of simplicity, specialized dictionaries for other fields may be used.

The storage device 18 of the computer (C)
The version numbers 18c1 to 18c5 are added to each of the field selection information 18a and the dictionary information 18b (the basic dictionary 18b1 and each specialized dictionary) stored in.
The version number corresponds to the information stored in the storage device 18 acquired from the computer (B).
Is referred to in order to determine whether the information is the same as the latest information currently held. The version number added to the field selection information 18a and the dictionary information 18b stored in the storage device 18 is compared with the version number added to the field selection information B1 and the dictionary information B2 held by the computer (B). If it is the same version number, it can be determined that it is the latest information.

Information published via the Internet is constantly changing. For example, politicians who were previously unknown in the world due to elections and political changes sometimes suddenly appeared in political articles on the Internet one day, and in baseball and soccer, when a new season began, many Players are replaced, and unknown players who have been in the minor leagues suddenly start playing in the middle of the season, and that information may be disclosed.

In the computer (B), the field selection information B1 and the dictionary information B2 are updated with the latest information obtained through the Internet, which changes from moment to moment. By acquiring the latest information and performing machine translation using the information, it is possible to execute appropriate translation based on the latest information.

Next, an information updating process for updating information stored in the storage device 18 by the computer (C) using the computer (B) will be described with reference to a flowchart shown in FIG. This information update process
The processing is started when the translation program 12b is started,
Until the translation program is terminated, the processing is continued in the background when there is little load without translation processing in parallel with other processing.

If there is no data being downloaded, the CPU 10 transmits the dictionary information including the field selection information B1 and various specialized dictionaries to the computer (B) via the Internet at a predetermined timing during operation according to the translation program 12b. The latest version number of B2 is inquired (steps A1, A4).

The CPU 10 compares the version number added to the field selection information 18a and the dictionary information 18b recorded in the storage device 18 with the version number obtained from the computer (B). As a result, the storage device 18
If the information stored in is updated (step A5), the CPU 10 stops the information updating process for a while (predetermined time) (step A6). This is to prevent an increase in load due to unnecessary processing caused by frequently inquiring the latest version number to the computer (B) since the latest version information is already stored in the storage device 18.

On the other hand, if it is determined that the information stored in the storage device 18 is not the latest, the CPU 10 sets the type name of the information determined not to be the latest to the downloading information type name 12d (step). A7). For example, when the baseball specialized dictionary version number of the dictionary information 18b shown in FIG. 3 is old, “baseball specialized dictionary” is set as the type name in the information type name 12d being downloaded.

Further, the CPU 10 sets the latest version number obtained from the computer (B) corresponding to the information determined as the old version information as the downloading information version number 12e (step A8), and Data amount 12f is "0"
(Step A9).

In the following, the CPU 10 downloads the next download target information indicated by the downloaded data amount 12f from the computer (B) via the Internet, stores it in the downloading information buffer 12c, and stores the downloaded data. The amount is added to the downloaded data amount 12f (step A
10).

Hereinafter, if the download of the information to be downloaded has not been completed (step A1)
1), the CPU 10 continues downloading information from the computer (B). If the download is to be continued (step A1), the CPU 10 sets the version number set in the downloading information version number 12e,
While determining whether the version number of the information to be downloaded held by the computer (B) matches, that is, whether or not the information has been updated to the latest content in the computer (B), it is determined based on the version number. Execute (Step A2). If the information of the computer (B) is not updated during the download, the information to be downloaded next indicated by the downloaded data amount 12f is transferred to the computer (B) via the Internet.
Downloading information buffer 12 from
c and add the downloaded data amount to the downloaded data amount 12f (step A).
10). Hereinafter, downloading is continued in the same manner.

On the other hand, if the computer (B) determines that the information has been updated to the latest content during the download, the CPU 10 proceeds to the download information buffer 12c, the download information type name 12d, and the download information. The version number 12e and the downloaded data amount 12f are cleared (step A3), and the latest information is downloaded again in the same manner as described above (steps A4 to A11).

When the download is completed, the CP
U10 replaces the information specified by the downloading information type name 12d stored in the storage device 18 with the information stored in the downloading information buffer 12c. Also, the version number of the information whose information has been replaced is replaced with the version number stored in the downloading information version number 12e.

Since the downloaded data amount 12f indicates the downloaded data amount, for example, when the translation program 12b is terminated during the download, the downloaded data amount 12f is stored in the storage device 18.
When the translation program 12b is started next time, the download can be restarted from the middle of the information where the download was interrupted in accordance with the amount of downloaded data stored in the storage device 18.

In this way, the computer (C) acquires the latest field selection information B1 and dictionary information B2 held by the computer (B) while executing the translation program C1, and stores it in the storage device 18. Since the field selection information 18a and the dictionary information 18b stored in the database are constantly updated, an appropriate translation result can be obtained without increasing the burden on the user.

Next, a homepage translation display process for translating and displaying a sentence described on a homepage downloaded from a WWW server will be described with reference to a flowchart shown in FIG. First, the homepage (Web page) downloaded by the browser function
Is specified, and when the user instructs execution of the translation process for this homepage, the CPU 10
Suspends the information update process that is operating in parallel in the background in order to output the translation result in a short time (step B1).

The CPU 10 receives the data of the designated page via the Internet (step B2), and divides the contents of the page into sentences which are considered to be in the same field before starting to translate the data. (Step B3). This is to prevent sentences in different fields from being mixed in the sentences to be processed without being divided.

In order to divide a sentence into fields, it is necessary to analyze the content to determine whether the field belongs to the same field. However, this requires complicated processing that can determine the content of the sentence. I will. Therefore, here, HTML (Hyper Text Markup Lan) which defines the content of the page
guage) is used to divide the text. Various methods can be considered as a method of dividing a sentence using an HTML tag.

For example, using a tag for designating a character size or a typeface, a method of converting a large character or bold sentence into a single sentence to be translated from immediately before the next large character or bold sentence, There is a method of dividing the document according to the position of the sentence using a tag for designating a position to be performed, and a method of dividing the document for each paragraph or ruled line using a tag for designating a paragraph or a ruled line.

FIG. 6 shows an example of a homepage (part), and FIG. 7 shows an HT corresponding to the homepage shown in FIG.
An example of the description in ML is shown. FIG. 7 shows an HTML description of only the first two parts of the homepage (FIG. 6).
Only the portion corresponding to (range shown in middle A) is shown. The homepage shown in FIG. 6 is related to sports and has articles about baseball (“Baseball”).

On this page, headlines of various sports articles and outlines of the articles are lined up separated by spaces, and details of the articles are displayed by clicking a character string portion to which a link is set. It has become. In the example of the page shown in FIG. 6, the fields are classified by characters indicating the field of sports, such as "Baseball" and "Hockey". Therefore, a method is conceivable in which the content of the page is divided using this to determine the field of each part. However, such division into the contents of a sentence in this manner requires an analysis process, which complicates the process. In addition, since there are many homepages in which fields are not divided in such an easy-to-understand format, this method cannot be applied to all homepages and processed.

As shown in FIG. 6, in this home page, descriptions in the same format (the part under “ANGELS4.
ORES4 ... the following parts ") are lined up. Correspondingly, the same HTML tag is used in the HTML description, except for the content of the article. In the present embodiment, an HTML tag commonly used in the respective units is used to divide by a description part in the same format.

For example, it is separated by a <P> tag indicating a paragraph break shown in B in FIG. 7, or a font size shown in C in FIG. 7 is controlled. <FONT SIZE = −
If the method of dividing by 1> is used, a sentence can be easily divided. Further, in order to divide a sentence according to various situations, it may be possible to make comprehensive judgment using more tags and divide the sentence. Note that when a sentence is divided using a plurality of tags, a plurality of sentences in the same field may be further divided. However, there is no problem unless the text is divided so finely as to impose a load on the translation processing itself. By dividing the text described on the homepage in this way, the divided text becomes a text included in one field.

The CPU 10 determines the field of the sentence by referring to the field selection information 18a stored in the storage device 18, using each of the divided sentences as a translation processing unit, and determines a basic dictionary corresponding to the determined field. Alternatively, translation is sequentially performed for each document using a specialized dictionary (step B4). In addition,
The details of the translation process will be described later (see FIG. 10). When the translation of all sentences is completed, the CP
U10 displays a character string (in this case, a sentence in Japanese) as a translation result in accordance with the format of the original homepage using, for example, a browser function (step B6). Thereafter, the CPU 10 restarts the suspended information updating process.

In this way, the content described on the home page is divided based on the tags in the HTML description that defines the content of the page, and a specialized dictionary is created using the divided text as a translation target unit. Apply and perform translation. For this reason, even when sentences in a plurality of fields are mixed, machine translation using the dictionary information 18b (basic dictionary, various specialized dictionaries) updated to the latest contents stored in the storage device 18 is not performed. Can be performed.

Next, details of the field selection information 18a will be described. FIG. 8 is a diagram illustrating an example of information registered as the field selection information 18a. For simplicity, the field selection information 18a shown in FIG. 8 shows only three fields, baseball, soccer, and politics, and only a part of proper nouns representing each field are shown. The groups from "Brewers" to "Bob Wiclman ..." in FIG. 8 are proper nouns related to the field of baseball. In addition, as described above, the names of all players, all managers, all coaches, The names of all stadiums are registered. Also, "Chri
The groups from "Stain Vieri" to "Serie A" are proper nouns related to soccer, and the remaining groups are proper nouns related to politics.

Further, for each proper noun registered in the field selection information 18a, a point corresponding to the degree of representing each field, for example, the frequency used in the text of the relevant field is given. (Hereinafter, this point is referred to as a field selection point). A sentence in which a proper noun having a high field selection point of a specific field appears is likely to be a text related to the field.

For example, a sentence in which a proper noun having a field selection point of “10” in the field of baseball and “0” in another field such as soccer and politics is a text on the topic of baseball. The possibility is very high, and it can be determined that the possibility of the topic of soccer or politics is low. Therefore, in the case of a sentence including the proper noun “Hideo Nomo”, since only the field selection point of this proper noun is “10” only in the field of baseball, it is very likely that the word is related to baseball, It can be determined that it is not a text related to the field of politics.

When the field selection point of the field of baseball is "5" and the field selection point of other fields is "0", it is possible that the sentence is not a topic of soccer or politics but a topic of baseball. High, but the previous field selection point is "10"
This means that the probability is lower than in the case of. For example, it is the name of a baseball player, not the name of a soccer player or a politician, but corresponds to a case where the player is not very famous or a general name.

In the example of the field selection information 18a shown in FIG. 8, the proper noun “Eric Owen
The field selection points for each field of "s" are baseball "5", soccer "5", and politics "0". For all other proper nouns, only one field has field selection points, and other fields The field selection point is “0”. That is, the proper noun (name) "Eric Owens" is set to appear in sentences in both the fields of baseball and soccer. This proper noun `` Eric Owen
The sentence in which "s" is used may be a topic of baseball or soccer, but it is not as likely in any field as the sentence using a proper noun whose field selection point is "10". Show. For example, it could be a sentence describing a topic about the common man named "Eric Owens". The details of the determination of the field of the sentence will be described later (FIG. 10).

In FIG. 8, for the sake of simplicity, it is shown that proper nouns are registered for each field, but actually, as shown in FIG. It is assumed that nouns are registered (FIG. 9 shows only proper nouns shown in FIG. 8). Since the proper nouns registered in the field selection information 18a are sorted in alphabetical order, it is possible to quickly determine whether a certain word in the text to be translated is registered in the field selection information 18a.

Next, a translation process for performing machine translation on a sentence as a processing unit will be described with reference to a flowchart shown in FIG. The process of dividing each sentence as a processing unit is described in the above-described home page translation display process (FIG. 5, step B3).

First, the CPU 10 sets the field selection information 18a.
Is used to clear the field selection point total (SPx) corresponding to each field used in the processing for determining the field of the text to be translated (step C1). It is assumed that a storage area for storing the value of the field selection point total (SPx) is secured in the memory 12. Each numerical value used in the following description is also stored in the memory 12 as needed.

The CPU 10 searches the sentence to be translated for a word that is not at the beginning of the sentence and starts with a capital letter, that is, a word that is a proper noun (step C2). Here, in order to reduce the number of searches for the field selection information 18a and to realize high-speed processing, the head of the sentence is always capitalized and it is highly likely that a useless search will be performed. I do. Also, for example, "The more I watch hi
m, the more I'm impressed with him. ", also exclude words that are not clearly proper nouns, such as" I ".

Here, if a word starting with a capital letter can be searched for at a word that is not at the beginning of the sentence, and the word following the word also starts with a capital letter, the CPU 10 extracts it as a series of proper nouns following the preceding word ( Step C3). However, when a delimiter between words such as “,” is specified between words, only the preceding word is extracted.

If the extracted word does not already appear in the text to be translated, the CPU 10 searches whether the extracted word matches a proper noun registered in the field selection information 18a (step C5). If the proper noun corresponding to the field selection information is registered, the field selection point for each field set in the proper noun is added to the field selection point total (SPx) for the field (step C6). . However, if the extracted word has already appeared in the text to be translated, it is not added again to the field selection point total (SPx). This is because, for example, it is considered that it is more likely that the topic of baseball players appears once when the names of two baseball players appear once than when the name of one baseball player appears twice. is there. Note that the processing here can use another method, for example, when the same proper noun appears for the second time, a method of adding half of the field selection points can be used.

FIG. 11 is a diagram showing the progress of appearance of proper nouns for a sentence to be translated. In FIG.
“Appearance order” indicates the order in which the proper nouns that have not been excluded in advance appear in the text to be translated. "SP baseball"
Indicates the total value of the field selection points for baseball in the total field selection points (SPx), and “SP soccer” and “SP politics” indicate the total field selection points for the fields of soccer and politics, respectively. ing.

In the example shown in FIG. 11, the proper noun that first appeared is "Brewers", indicating that this proper noun was registered in the field selection information 18a. For the proper noun “Brewers”, 8 points are set only in the field of baseball in the field selection information 18a, and 0 points are set in the field of soccer and politics.
Points are added.

Since the next proper noun "Draw Olson" (name of the reporter) is not registered in the field selection information 18a, it is not added to the field selection point total (SPx).
The same processing is performed hereinafter, except that “Brewers” in the order of appearance 10
Has already appeared in the appearance order 1, so there is no addition to the field selection point total (SPx). Appearance order 31
"Eric Owens" has five points for baseball and five points for soccer according to the field selection information 18a, so the field selection points total (SPx) for both fields is added by 5 points. When the processing is completed for all proper noun candidates, “SP baseball” is replaced by “SP soccer”.
Since this is greater than “SP politics”, this is the maximum value MAXSP of the cumulative field selection points (SPx) for each field.

When the search for all proper nouns is completed from the text to be translated (step C2), the CPU 10 calculates the total volume VOL of the text (step C7). Assuming that the total number of words is used as the total volume VOL of sentences, for example, the total amount of sentences for a sentence having 617 words is VOL = 6
It becomes 17.

The CPU 10 calculates the maximum point density MAXDEN by dividing the maximum value MAXSP in the total field selection points (SPx) of each field by the total volume VOL of the text (step C8). In this case, the maximum point density MA
XDEN = 125/617 = 0.2.

Here, the CPU 10 determines whether or not the value of the maximum point density MAXDEN exceeds a field selection threshold DENTHR set in advance to determine the field of the text (step C9). It should be noted that an appropriate value of the field selection threshold DENTHR is determined in advance based on analysis of many sample data.

If the value of the maximum point density MAXDEN does not exceed the field selection threshold DENTHR, the CPU 10
Does not set a specialized dictionary as a dictionary used for machine translation,
Only the basic dictionary 18b1 in the dictionary information 18b is set (step C10). Thereafter, the CPU 10 returns to the basic dictionary 1
Machine translation is performed on the document to be translated using 8b1 (step C11).

On the other hand, if the value of the maximum point density MAXDEN exceeds the field selection threshold DENTHR, the CPU
In step C12, it is determined that the text to be translated is a text in a field where the cumulative field selection points (SPx) are the largest, and a specialty dictionary in that field is set as a specialty dictionary to be used thereafter (step C12). Thereafter, the CPU 10 executes machine translation on the document to be translated by using the set specialized dictionary with higher priority than the basic dictionary 18b1 (step C13).

For example, the field selection threshold DENTHR = 0.
If it is 1, the maximum point density MAX in the above example
DEN is 0.2, maximum point density MAXDEN
Is larger than the field selection threshold DENTHR, it is determined that the English sentence to be translated relates to the field of baseball. As a result, the baseball specialized dictionary 18b is used as a specialized dictionary.
2 is set and used in preference to the basic dictionary 18b1.

FIG. 12 shows an example of the baseball dictionary 18b2. As shown in FIG. 12, the proper noun and base name of only the last name (for example, Nomo) registered in the field selection information 18a are registered in an alphabetical order. Many other words that have baseball-specific translations (eg, "walk", "bottom", "start", "double"
Etc.) are also registered. For each word, in addition to the part of speech and the translated word, a lot of information (not shown) used for translation is registered in association with each other.

Using a baseball specialized dictionary 18b1 in which words having proper nouns relating to baseball and translations unique to baseball as shown in FIG. 12 are registered, machine translation is performed on sentences determined to be contents relating to baseball. As a result, a translation suitable for baseball is selected, and a meaningless translation is not performed. In addition, proper nouns such as player names are surely translated. Also, the maximum point density MAXDEN
Is smaller than the field selection threshold DENTHR, it is determined that the text to be translated is not a field in the field registered in the field selection information 18a, and only the basic dictionary 18b1 in which general meaning is registered is used. Since the machine translation is performed, a translation unique to baseball is not selected and its meaning is not unknown. At this time, the proper noun is a proper noun registered in the basic dictionary 18b1 (in a personal name,
Only common names) will be translated.

The translation of a proper noun such as a person's name is a translation of a proper noun when the field of the text to be translated cannot be specified, or a proper noun corresponding to another field when the field is specified. It is also possible to configure so that the translated word of can be selected.

As described above, in the translation apparatus according to the present embodiment, the field of the content of the text is determined based on the words (proper nouns) included in the text to be translated, and the field is prepared in advance according to this field. Automatically switches between multiple specialized dictionaries, so there is no need for the user to switch between specialized dictionaries for each sentence to be translated or for the user to specify in advance the correspondence between the homepage and the specialized dictionaries, and without increasing the burden on the user. You can get translation results.

As the information that is used as the basis for making the decision for switching the specialized dictionary, a proper noun beginning with a capital letter in the text to be translated is used. For this reason, the process of searching for the determination information (proper noun) can be executed easily and at high speed, and there is less error than the determination using a general word. For example, when trying to determine the baseball field,
The batter “batter” or the home run “homer” may be determined by using words, but these words have meanings other than baseball, so that mistakes are likely to occur. In addition, when searching for proper nouns, words at the beginning of a sentence or special words such as "I" are excluded in advance, so the number of times of searching the field selection information 18a is reduced, and the processing is faster.

Further, in the field selection information 18a used for discriminating the field of the text, a plurality of points (field selection points) indicating the possibility of being a text in each field are attached to each proper noun. When there is a possibility in a plurality of fields or the degree of showing each field can be adjusted, it is possible to cope with various situations. When determining the field of a sentence,
Since a value (maximum point density MAXDEN) divided by the amount of sentences is used instead of the field selection point total (SPx) itself, correct judgment can be made without being affected by the amount of sentences or the method of dividing sentences. Further, even if the field selection point total (SPx) reaches a certain value, the specialized dictionary is not used unless the maximum point density MAXDEN exceeds a preset field selection threshold DENTHR.
It is rare that machine translation is executed by mistake using a specialized dictionary for sentences having general contents.

Even when there are sentences in different fields on the same page, the contents of the page are divided into a plurality of parts so as not to include the different fields, and automatically translated to a specialized dictionary suitable for each part. Therefore, even a page in which sentences in different fields are mixed can be correctly translated without hassle. In the present embodiment, as a method of dividing a page into sentences for each field, an HTML tag for defining the description of the page is used. By performing the processing, the processing can be executed at high speed, and erroneous division can be reduced.

Also, proper nouns which are likely to appear in sentences in each field are registered as field selection information 18a, and their translations and related translations (such as translations of only the last name) are also registered. Proper nouns are not mistranslated or output as a translation result without being translated in the original language, and a translated sentence that can be easily understood even by a person with little knowledge of the original language is easily created.

In the translation device (computer (C)) according to the present embodiment, the field selection information 18a and the dictionary information 18b (basic dictionary 18b1 and various specialized dictionaries 18b2 to 18b-1)
8b4), etc., to the computer (B) having the latest field selection information B1 and dictionary information B2 at all times. If it is confirmed that the information is outdated, the computer (B) is automatically checked. Since the information is downloaded from and updated, it is possible to correctly execute the translation even for a sentence in which the latest topic is described without any trouble of the user.

Further, since downloading of the latest information such as the field selection information B1 and the specialized dictionary B2 is executed when there is no translation process, there is no influence that the speed of the translation process becomes slow. Further, even if the translation is started during the download or the download is interrupted on the way due to the termination of the translation program 12b, the download can be resumed from the continuation based on the downloaded data amount 12f. Therefore, even if there is a large amount of information to be updated, it can be efficiently downloaded. Therefore,
The user can end the translation program 12b without being aware that the download has ended, and can easily handle the translation program 12b. Also, if there is information to download, even if the download is interrupted, the translation can be performed using the information before the update, and the latest information is used when the download is completed. You can start the translation.

The translation device according to the present invention can also be configured as follows. In the above description, the basic dictionary 18b1 and the specialized dictionaries 18b2 to 18b4 are described as separate and independent information. However, as one dictionary (comprehensive dictionary), a general translation and a translation in a specialized field such as baseball are included. May be registered. in this case,
Select a translation as a translation result from among the translations corresponding to the field of the text to be translated currently. FIG. 13 shows a configuration example of information about the word "walk" registered in the comprehensive dictionary (however, the contents of the specialized dictionary are shown only in the field of baseball). As shown in FIG. 13, a plurality of translated words (meanings) for each of “general” and “field” and their parts of speech are registered in association with words in the language to be translated (other words). Information necessary for translation processing is also included.) When this comprehensive dictionary is used, if the text to be translated is a content related to baseball, a translation (meaning) in the field of "baseball" is selected as a translation for "walk" in the text.

In the above description, the field selection information 1
The proper nouns registered in 8a and their translations are also registered in the specialized dictionary, and after the field is determined, the specialized dictionary is searched and its translation is extracted. However, the translation of the proper noun is also registered in the field selection information 18a. In addition, the translation of the proper noun may be extracted at the same time when the field selection information 18a is searched. Although the full name is registered for the personal name in the field selection information 18a, only the last name may be registered. In this case, it is better to set the field selection point of the proper noun for the last name lower than the field selection point for the full name. Further, although the field selection information 18a is provided separately from the basic dictionary and the specialized dictionary, they may be integrated. Also,
When searching for proper noun candidates, all words at the beginning of the sentence are excluded, but if certain conditions are satisfied, the words at the beginning of the sentence may be included. For example, if the word following the first sentence is also capitalized,
`` Hideo Nomo can't explain why he pitches so well
in SunFrancisco. "," Hideo Nomo "will also be adopted as a candidate for proper noun. Furthermore, all words starting with a capital letter, including the beginning of a sentence, may be used as candidates.

Although the number of fields that can be classified is fixed, the number of fields is automatically updated by automatically updating the field selection information 18a and automatically adding or deleting a specialized dictionary. May be increased or decreased. For example, only during the Olympics, fields such as judo and marathon can be set up, or fields that are temporarily popular can be set up and abolished as the fashion goes away. In this case, the update of the field selection information 18a and addition or abolition of the specialized dictionary corresponding to a new field can be performed by information downloadable from the computer (B).

In the above description, the field selection information 18a stored in the storage device 18 of the computer (C)
In order to update the information and the dictionary information 18b, it is assumed that the other computer (B) obtains the latest field selection information and dictionary information from the other computer via the Internet. Any configuration can be used as long as the latest information can be obtained from an external computer (such as a server device) via the PC.

[0081]

As described above, according to the present invention, information used for translation processing is automatically updated via an external network, and the contents described in the text based on the updated information are updated. The translation is performed by applying the specialized dictionary for each domain, and the translation result can be easily translated even for sentences that include multiple proper fields and proper nouns without increasing the burden on the user. It will be possible to translate in an understandable way.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an example of a translation system using a translation device according to an embodiment.

FIG. 2 is a block diagram showing a configuration of a translation device realized by a computer (C).

FIG. 3 is a view showing an example of information used when a computer (C) executes machine translation.

FIG. 4 is a flowchart for explaining an information updating process for updating information stored in a storage device by a computer (C).

FIG. 5 is a flowchart for explaining a homepage translation display process for translating and displaying a sentence described on a homepage.

FIG. 6: Homepage to be translated (part)
FIG.

FIG. 7 is a view showing an example of a description in HTML corresponding to the homepage shown in FIG. 6;

FIG. 8 is a diagram showing an example of information registered as field selection information 18a.

FIG. 9 is a diagram showing an example of information that is sorted and registered in alphabetical order as field selection information 18a.

FIG. 10 is a flowchart for explaining a translation process of executing machine translation for a sentence as a processing unit.

FIG. 11 is a view showing the progress of appearance of proper nouns in a sentence to be translated.

FIG. 12 is a diagram showing an example of a baseball specialized dictionary 18b2.

FIG. 13 is a diagram showing a configuration example of information on a word “walk” registered in a comprehensive dictionary.

[Explanation of symbols]

 (A) (B) (C) Computer 10 CPU 12 Memory 12 b Translation program 12 c Downloading information buffer 14 Display device 16 Input device 17 Communication device 18 Storage device 18 a Field selection information 18 b Dictionary information

Claims (4)

[Claims] 1. A proper noun extracting means for extracting proper nouns from a text to be translated, and referring to field selection information in which proper nouns representing various fields described in the text are registered for each field. A field determining means for determining a field of the content described in the text to be translated based on the proper noun extracted by the proper noun extracting means; and a correspondence between the fields determined by the field determining means. A specialized dictionary used for translation processing, and a translation means for performing translation on the text to be translated using the specialized dictionary corresponding to the field determined by the field determination means. A translation device characterized by the above-mentioned. 2. The translation apparatus according to claim 1, further comprising information updating means for acquiring the field selection information and the specialized dictionary from outside via a network and automatically updating the contents. 3. The text to be translated is WWW (Worl
d Wide Web) is a sentence described on the page acquired, HTML (Hyper Text Markup Language)
2. A sentence in the page described in (1) is divided into sentences for each field based on a tag, and translation is executed by applying a specialized dictionary to each of the divided sentences.
Translation device as described. 4. A proper noun extracting means for causing a computer to extract proper nouns from a text to be translated, and field selection information in which proper nouns representing various fields described in the text are registered for each field. A field determining means for determining a field of the content described in the text to be translated based on the proper noun extracted by the proper noun extracting means, and a field determined by the field determining means A computer-readable recording medium in which a translation program for causing a translation unit to execute translation of a sentence to be translated using the specialized dictionary corresponding to the translation program is recorded.