WO2001055901A1

WO2001055901A1 - Machine translation system, translation server thereof, and client thereof

Info

Publication number: WO2001055901A1
Application number: PCT/JP2001/000343
Authority: WO
Inventors: Hiroyasu Kikuchi
Original assignee: Joyport Incorporated
Priority date: 2000-01-25
Filing date: 2001-01-19
Publication date: 2001-08-02
Also published as: TW501030B; AU2707301A; JP2001209643A

Abstract

A machine translation system by the transfer translation method in which units are distributed by use of the Internet as a medium for processing distribution such that an analyzing unit (12) is provided in a client, and a conversion unit (23) and a generating unit (24) are provided in a server. A translation server (200) used in a machine translation system for translating an original text to a text in a second language by use of the Internet comprises a conversion unit (23) that receives original text analysis data representing the result of analysis of the original text in units of a word and converts the original analysis data to second language analysis data on each word in the second language according to a appropriate translation dictionary and/or original text language word dictionary and translated word dictionary, and a generating unit (24) that generates a translated text in the second language according to the second language analysis data, translation rules, and translated word dictionary. The generating unit (24) generates a synthesis text for displaying a pair of the original text and the translated text and outputs the synthesis text.

Description

Specification

Machine translation system, its translation server and its client

TECHNICAL FIELD The present invention relates to a machine translation system that translates an original sentence into another language according to a translation program distributed using a network such as the Internet, a translation server and a client thereof, and a recording medium storing the program.

Scenic technology Translation involves re-expressing the content expressed in one language in another language, and consists of a process of understanding the content of the expression and a process of expressing the understood content again in the target language. It is difficult to realize this process strictly in a so-called computer environment.Machine translation employs a method that replaces the source language expression with the target language expression by using grammar and semantic correspondence between languages. . Machine translation methods are broadly divided into transfer methods that perform this replacement between two languages, and intermediate language methods that convert the target language through a universal language as the third language. In practice, the intermediate language method requires much time, so the transfer method is the mainstream in most cases.

The transfer method includes a direct translation method that replaces words at the word level, a semantic analysis translation method that analyzes and constructs the meaning of a language, an example-based translation method that uses a bilingual corpus (aggregated words), and a knowledge base that uses common sense. Translation is being carried out by considering a form translation method and an eight-bridge type translation method that mixes these. Machine translation is an applied technology that represents natural language processing during computer processing.The transfer method uses morphological analysis, syntax analysis, semantic analysis, and other language analysis, and second, syntax conversion and vocabulary conversion. Thirdly, it is composed of three processes, such as translation selection and morphological adjustment.

In language analysis technology, the accuracy of analysis has been greatly improved. However, in the case of long sentences, the prospect of semantic analysis technology is opening up with the development of a semantic dictionary containing a large number of case frames.

As for conversion technology, wide-area conversion technology is being attempted, which converts even expressions that span multiple words collectively. In the generation technology, the generation of determinants (articles and numbers) is a problem, but other technologies include omission complementary technology that applies context processing, and automatic conversion of sentences into texts that are easy to translate in the system. Replacement source text Automatic rewriting technology, technology for automatically acquiring knowledge of grammar and vocabulary from a large corpus, etc.

A wide variety of knowledge such as linguistic knowledge, situational knowledge, and specialized knowledge is required as translation knowledge, and it is necessary to convert these into rules and machine dictionaries. Recently, e-mail translation services are expected as an application that emphasizes the immediacy of translation.

Here, as a conventional technique, a similar example translation system is disclosed in Japanese Patent Application Laid-Open No. H10-3012382.

According to the publication, (1) an input device for inputting an original sentence to be translated, and (2) a single sentence extracting means for cutting out the input original sentence by sentence and outputting the original sentence,

(3) An example database that stores the dictionary database and example sentences and their translations as pairs. — A database section consisting of evening bases. (4) A complete translation that searches the example database and outputs an example sentence that completely matches the original sentence. A matching sentence translating means; and (5) a similar example translating means for searching an example database to extract an example sentence that is partially inconsistent with the original sentence, and translating the translated sentence by changing the translated sentence according to the original sentence; (6) Example sentences that match the original sentence by more than a predetermined rate by searching the example database A sentence segmentation method that divides the original sentence according to a predetermined rule when no sentence is found, and (7) a sentence perfect match sentence that searches the example database and outputs a bilingual sentence of an example sentence that completely matches the segmented original sentence A translation means; and (8) a divided sentence similar example translating means for searching an example database, extracting an example sentence that is partially inconsistent with the divided original sentence, and translating by changing the bilingual sentence according to the divided original sentence. And (9) a new sentence translator for searching and translating the dictionary database when an example sentence that matches the divided original text by a predetermined rate or more cannot be found by searching the example database; and (10) each of the divided original texts. And a split translation assembly means for assembling the translation of and outputting a translation of the original sentence.

In addition, in this publication, a machine translation server is installed at a central location to use the similar example translation system via the Internet, and access is made from terminals via the Internet to translate. It states that the terminal can use machine translation from any terminal and from any location as long as it has a function to access the Internet.

As described above, according to the publication, the example database is searched, and a bilingual translation is obtained from an exact matching example sentence, or translation is performed using a similar example sentence. Therefore, translation can be performed quickly with a simple configuration. Since the previously translated sentences are used, duplicated translation work can be omitted, the translated sentences can be unified, and in the case of post-editing, the translated sentences can be corrected by referring to the example database, making editing work easy. According to the company, translation can be performed via the Internet, so users can use machine translation from anywhere with only a small terminal device.

However, the role of the network in the conventional machine translation is, as described in Japanese Patent Application Laid-Open No. Hei 10-32122, a user interface that connects the input / output device unit and the translation processing unit. There was no idea other than the use of path resources for, and there was no idea of distributed processing of machine translation literature itself. Therefore, effective use of servers connected to the network and having infinite resources is ineffective. Is desired. In particular, it has long been desirable to disperse over-intensive translation work by configuring the machine translation system as a distributed processing system on the Internet or other networks and distributing and using huge databases. Had been.

Also, for example, when e-mail is transmitted across countries, it can be delivered via a network and translated into the local language rather than displayed on the destination client in the original text. It was also desired in order to carry out simple and prompt processing of the wastewater.

The present invention relates to a transfer method, which is a method of machine translation, by using a network as a medium for processing distribution and distributing an analysis unit to a client side, and a conversion unit and a generation unit to a server side. It is an object of the present invention to reduce processing at a server where processing is concentrated and to perform high-speed translation processing from a client.

In addition, for the conversion from a given language to another multiple languages, resources for conversion / generation to other languages are added to the conversion unit and generation unit on the server side without depending on the client. This makes it possible to build a machine translation server that supports other languages.

DISCLOSURE OF THE INVENTION In order to solve the above-mentioned problems, the present invention provides a translation server used in a machine translation system that translates an original sentence into another language using a network. A conversion unit that receives and converts the original sentence analysis data into multilingual analysis data of the other language of each word based on the appropriate translation dictionary and / or the original sentence co-lingual word dictionary and the translation word dictionary, and is converted by the conversion unit A generating unit that generates a translated sentence that is a sentence in a foreign language from the foreign language analysis data based on a translation rule and a translation dictionary, wherein the generating unit A composite sentence that can be displayed as a pair with the translated sentence is generated and output. In addition, the present invention analyzes the original text using a network, entrusts the analysis result to an agent as an analysis file, and converts the analysis file from the analysis file to a foreign language analysis database based on a dictionary database. A communication processing unit capable of communicating with a network in a client used for a machine translation system that generates a translation from a foreign language analysis data for each word and sends the translation to a client at a destination address; An input unit for inputting an original sentence, an analyzing unit for separating and analyzing the original sentence in word units, and an output unit for outputting analysis data analyzed by the analyzing unit to the communication processing unit as an e-mail, The analysis unit analyzes the word for each space according to each language, or extracts a word according to the symbol.

The present invention also provides a machine translation system for translating an original sentence into another language using a network, comprising a client, a translation server, a destination client, and an agent mediating between them, wherein the client is A communication processing unit capable of communicating with the network, an input unit for inputting the original sentence, an analyzing unit for separating and analyzing the original sentence in units of words, and an analysis data analyzed by the analyzing unit as an e-mail. An output unit for outputting to the communication processing unit, wherein the translation server receives original sentence analysis data obtained by analyzing the original sentence in units of words, and converts the original sentence analysis data into a suitable translation dictionary and a Z or original co-language word dictionary. A conversion unit for converting each word into the multilingual analysis data of the other language based on the translation dictionary; and a translation rule from the multilingual analysis data converted by the conversion unit. And a generator for generating a translated sentence is another language sentence based on the fine translation dictionary, in front Symbol generation section generates and outputs a synthesized statement that can be displayed as pairs and the translation and the original text.

The present invention also provides a computer-readable recording medium storing a translation program installed by a translation server used in a machine translation system that translates an original sentence into another language using a network, The translation server analyzes the first analysis data obtained by analyzing the original sentence for each word of the minimum unit by the client based on the word dictionary of the language and the language installed again, and the second analyzed data is used. The analysis data is converted into translation data based on the translation dictionary installed, and the translated translation data is generated based on the installed translation rules and appropriate translation rules. A computer-readable recording medium characterized by transmitting a translation to a destination client is provided.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a conceptual block diagram of a conceptual translation of the present invention.

FIG. 2 is a configuration diagram of the machine translation system of the present invention.

FIG. 3 is a configuration diagram of the machine translation system of the present invention.

FIG. 4 is a diagram showing a physical configuration of conceptual translation of the present invention and a conventional example.

FIG. 5 is a block diagram of the machine translation system of the present invention.

FIG. 6 is an overall schematic flowchart of the machine translation system of the present invention. FIG. 7 is a flowchart of the analysis unit of the machine translation system of the present invention. FIG. 8 is a flowchart of the conversion unit of the machine translation system of the present invention. FIG. 9 is a flowchart of the generation unit of the machine translation system of the present invention. FIG. 10 is an explanatory diagram of an actual translation example of the machine translation system of the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION

[First Embodiment]

(1) Description of configuration

FIG. 1 shows a schematic configuration diagram of the machine translation system of the present invention. In Figure 1, an e-mail is sent from an English client 100 to a Japanese client. Then, the client 100 creates “This is an applej” as the body of the mail and sends it to the translation server 200 as an analysis file with an analyzed intermediate file obtained by analyzing the body using the analysis dictionary. The machine translation server 200 is usually located in the Japanese country, converts the analyzed intermediate file into the target language based on the translated word dictionary, and matches the translation rules. A translation is generated based on the translation dictionary, and at that time, the translation may include advertisement information.The client 3000 in the Japanese country checks whether an e-mail has arrived at the translation server 200. Upon confirmation, you will receive an e-mail from the English client 100. The e-mail contains the following text: "This is an applej" and the completed translation "Lingo is here." And an example If the display coexist and corporate ads that have been commissioned by the company.

Here, the network network includes a public network, a world-wide Internet network, LAN, WAN, WLL (Wireless Local Loop), etc., which are connected in series or in parallel. It includes a network network, and if a translation program for a translation server is installed on a client, the client can execute the original text analysis by the originating client and then convert and generate the analysis data. It may be.

FIG. 2 is a configuration diagram showing a flow of the machine translation system of the present invention. Although the configuration and operation of the general machine translation system have been described with reference to FIG. 1, further details will be described with reference to FIG.

In FIG. 2, first, a predetermined translation program is distributed and installed on servers and clients on the Internet, and the specificity of each server and client, a unique translation, and features are used. First, the text is created by the text creation unit 11 from the client 100 having a predetermined resource, the sent text is analyzed by the analysis unit 12, and the analysis result is analyzed by the agency 400. An analysis file is created in the analysis file section 22 from the UR L (Un if rm resource Locator), send it in the form of resource utilization on the Internet, and output the analysis file to the conversion unit 23 on the server 200 having the predetermined resource, where it is converted into the predetermined language. Then, the generation unit 24 translates the predetermined language into another predetermined language, completes the translation, and generates translation display data that can be displayed by the translation generation unit 32 of the output client 300. Then, the translated sentence is displayed and output on the display of the client 300.

Here, instead of seeing the text as a mere list of words, we regard it as a set (corpus) with some structure, extract its structure, and intermediately express it with figures and mathematical symbols. The parser generates translations while utilizing it as an auxiliary means.The analysis unit 12 has analysis methods such as morphological analysis, syntactic analysis, and semantic analysis of the text. Analyze based on. The conversion unit 23 has a technique such as syntax conversion and vocabulary conversion, and converts each minimum unit word into a foreign language using a translation dictionary or the like. The generation unit 24 has a method of selecting a translated word, adjusting morphemes, and the like, and generates a translated sentence according to a translation rule such as an arrangement of words.

The learning function generation unit 25 updates various dictionaries used by the client 100 and the server 300 as the latest dictionary, and outputs, for example, the original text from the client that created the text. If the translation is more appropriate, the feedback is provided as learning data 251, and the result of comparison between the completed translated sentence from the server 200 and client 300 and the text is provided as learning data 25 2. It has a function to update various dictionaries with the latest dictionaries. This learning function is appropriate, especially when translating into many other languages.

In addition, since the generation unit 24 of the server 200 outputs the composite sentence 26, it is possible to display the text 14 and the translation created by the generation unit 24 in parallel on the display of the client 300. Then, it is generated using codes such as HTML of image forming language such as page script. As a result, on the output client side, It makes it easy to compare the text with the translated text, and it is also effective for client users who have some knowledge of both languages as a means of checking the differences between the text and the translated text.

The distributed translation program described above may be a program in which the analysis, conversion, and generation programs are stored in a storage medium with the entire program as one package software, or as an analysis program for a client and a translation server. Package software may be used individually as the conversion 'generation program. In each case, various dictionaries are also included, and the analysis program uses the source language co-language word dictionary, and the conversion and generation program includes the source language co-language word dictionary, the translation dictionary, the translation rules, and the appropriate translation dictionary. I just need to add it. However, these dictionaries are also equipped with programs that are updated as needed to reflect the learning function. Further, as a recording medium storing the translation program, a hard disk, a floppy disk, CD-R 、 M, DVD, MO or the like is used. In addition, it is preferable to use a network such as the Internet as a transmission medium for delivering the translation program, so that it is not necessary to store the translation program in a package software.

FIG. 3 shows an example in which the same reference numerals as in FIG. 2 have the same functions, and an analysis unit 12 of the client 100 is provided with a conversion unit 13 for adding a conversion function. For example, if it is better to write the text itself in the translated sentence, the conversion unit 13 may analyze the word in the minimum unit by the analysis unit, and may not translate a certain word. By performing the conversion process, the load on the translation server can be reduced.

Figure 4 shows an illustration of the distributed processing of machine translation. Fig. 4 (A) shows the case of a conventional stand-alone translation system.In the case of single-transfer translation, the analysis unit 51 analyzes the text and the analyzed data is converted to a minimum by the conversion unit 52. Each word in the unit is converted into another language, and the converted data in another language is generated as a translation of a sentence by the generation unit 53 and output. On the other hand, in FIG. 4 (B), as described above, the client 100 analyzes the text, the conversion section 23 of the server 200 converts the analysis result, and the generation section generates a translated sentence. Output to the destination clients 301 and 302 via the Internet. In this case, the destination client may be singular, plural in the multicast format, or the majority in the broadcast format, and when there are multiple destination clients, each destination client is in one language. However, each may be in a different language zone, and there is no restriction. This distributed processing can significantly reduce the load on the server, and can process translation requests from many clients at high speed.

FIG. 5 shows an overall block diagram of the machine translation system of the present embodiment. In addition to Fig. 2, the point that various dictionaries based on data are specified is added. In FIG. 5, a sender client 100 has a text creator 11 for inputting a text which is an original text from an input device, and a text dictionary for inputting a text and a word dictionary in the same language as the text. An analysis unit 12 that analyzes the smallest word unit based on the word dictionary 14, and a file transmitting and receiving agent 15 that outputs the text analysis data and text analyzed by the analysis unit 12 to the Internet Consists of

Further, the translation server 200 inputs the file transmission / reception agency 16 received from the file transmission / reception agency 15 via the Internet, the text analysis data from the file transmission / reception agency 16 and the text. The conversion unit 23 converts the text analysis data into other languages of the receiving client in word units, and the generation unit 24 generates translations of other languages according to the translation rules and the appropriate translation dictionary. A file transmission / reception agency 17 for inputting the translation and its text and outputting it to the Internet, a server source co-language word dictionary 29 used for conversion by the conversion unit 23, and a conversion unit A translation dictionary 27 used for the conversion in 23 and the generation of the translation by the generation unit 24; a translation rule and a suitable translation dictionary 28 used for the generation of the translation in the generation unit 24; Clients connected to the Internet And a learning function means 25 for updating the various dictionaries 27 to 29 to the latest dictionary in accordance with various translation data from a translation server or the like.

Here, a suitable translation dictionary is also referred to as a so-called proper translation rule.For example, when there are multiple translations for one word, a rule for determining which one to select is stored in a recording medium. It stores translation rules as a database in the same way as dictionaries, and expresses the appropriate translation dictionary and the appropriate translation rule equally. Further, the receiving client 300 receives a file transmission / reception agency (not shown) corresponding to the file transmission / reception agency 17 and a text and a translation from the file transmission / reception agency (not shown) and outputs them on a display or the like. Means 3 3. Note that the clients 100 and 300 operate in the opposite manner in the transmission / reception relationship.

(2) Description of operation

Next, FIG. 6 shows a schematic flowchart of the entire machine translation system. First, the sender client 100 creates the body in the body creation unit 11 and starts the process of requesting translation to the body of the e-mail to other countries (S11). Then, the original text of the text is extracted from the file (S12). Next, the original sentence is analyzed with reference to the client original sentence synonym word dictionary 14. For example, in the case of an English sentence, the space is used as a reference, and in the case of Japanese, two words are used as a reference to determine whether or not the word is in the word dictionary. 13 ) . The presence or absence of unregistered words is determined by analyzing the original sentence (S14). If there is an unregistered word, an unprocessed flag is assigned to the entire original sentence, and the original sentence is adopted (S15). An analysis file as a storage means is generated by the ant 100 or the file transmission / reception agency 15 (S16). The file transmission / reception agency 15 transmits the original file and the analysis file to the translation server 200 via the Internet, using the address of the translation server 200 as the destination address (S17). The original file is sent to the translation server 200 by the file transfer agency 16 And the analysis file are received. The conversion unit 23 determines whether or not the analysis file has an unprocessed flag (S18). If the unprocessed flag is present, the server is provided with a reanalysis of the unprocessed portion. The execution is performed with reference to the word dictionary 29 of the original language (S 19). Next, the presence or absence of unregistered words in the analysis file is determined (S 20). Adopt as it is (S21).

Next, each word in the minimum unit is converted from the analysis file into a translated word in accordance with the translated word dictionary (S22). The converted file converted into the translated word is rearranged and combined in the generator 24 according to the translation rule and the translation dictionary of the translation rule, and “.” And! By inserting “,” “,” and the like, and translating and generating idioms and maxims according to an example dictionary storing a considerable number of grammatical or phrasal cases, a translated text is generated (S23).

Next, when the text creation is completed, the file transmission / reception agency 17 transmits the original text and the translation of the translated text to the designated URL (Uniform Resource Locator). At the designated URL, a corporate advertisement for advertising revenue is added to the original and translated texts, and output via the Internet to the destination client address. The destination client receives the original sentence, the translated sentence, and the corporate advertisement as an e-mail and displays it on an output means such as a display. When displaying on the display, the original and translated texts are displayed side by side, and the system operator who displayed the corporate advertisement information and distributed the translation program to the client or server gained advertising revenue and gained public benefit. To meet the social demands for

Next, the operation of the analyzing unit 12 of the transmitting client 100 will be described with reference to the flowchart of FIG. First, the original sentence is extracted from the file in which the original sentence is stored (S31), and the original sentence is analyzed for each word by processing symbols and spaces in order to decompose and analyze the original sentence. For example, in the case of English, it is decomposed with spaces and symbols placed before and after the word, such as ".", "One", """, and in the case of Japanese, the symbols", "". "" And the like (S32). Next First, a word search is started while referring to the client original sentence co-language word dictionary from the beginning of the sentence (S33), and the presence or absence of an unregistered word is determined (S34). The number of characters is changed, and the search is performed again while referring to the client original co-language word dictionary (S35). Next, when there is no unregistered word, it is further determined whether or not there is an unregistered word (S36). If there is an unregistered word, an unregistered flag is added to the entire e-mail and the original text remains as it is. It is stored in the storage means as the analysis data to be used, or the process directly proceeds to the next step (S37). Next, the analysis data is generated together with the original text as an analysis file (S38), and transmitted to a predetermined URL by the file transmission / reception agency 15 (S39). The given URL is the operator of the translation system, counts how much of this analytics data is being transferred, and, at this stage, a request from an advertising supplier who wishes to advertise internationally. A corporate advertisement may be added to the analysis data in accordance with.

Next, the operation of the conversion unit 23 of the translation server 200 will be described with reference to the flowchart of FIG. First, it is determined whether there is an unprocessed flag in the analysis data based on the original text received by the file transmission / reception agency 16 and the analysis data (S41). If there is an unprocessed flag, the unprocessed unregistered word is determined. Is searched with reference to the word dictionary 29 in the same language as the original sentence owned by the translation server 200 (S42). Next, together with the case where there is no unprocessed flag, the presence / absence of an unregistered word is determined (S43). Search (S44). Next, the presence / absence of an unregistered word is determined (S45). If there is an unregistered word, an inaccurate notification is given like a flag and the original sentence is output to the next step. In this case, the unregistered words of the original sentence are sent to the learning function unit 25 in a feedback manner, and are reflected in the client original sentence co-lingual word dictionary 14 so that the original sentence words are analyzed as they are.

Next, search the translation word dictionary 27 from the analysis data of the e-mail with no unregistered words and the analysis data to which the incorrect The translation is performed (S47). Next, it is determined whether or not there is an unregistered word in the translated word data that does not exist in the translated word dictionary (S48). If there is an unregistered word, the word of the original sentence is adopted (S49), and the process ends. Then, the result is output to the generation unit 24, and the result is output to the learning and maintenance processing of the translation dictionary for another language.

The operation of the generation unit 24 of the translation server 200 will be described with reference to the flowchart of FIG. First, the translation unit 23 receives the translated words and creates a target table for grammar operation.For example, to clarify the arrangement of the subject, predicate, object, complement, conjunction, etc. A subject, a predicate, an object, a complement, a conjunctive, etc., and a subject, a predicate, an object, a complement, a conjunction, etc. of a translated word are created in a target table (S51). Next, referring to the translation rule and the appropriate translation dictionary 28 (S52), apply the translation rule to the translated word data (S53), and then determine whether there is an unregistered rule (S54). If there is an unregistered rule, a reference is made to the example dictionary, and a translation table is searched from the example dictionary for custom words and proverbs in the target table created from the analysis data (S55). In this case, as in the case of a normal example dictionary, it may be checked whether or not the word is in the example dictionary not only for each word but also in sentences. Next, it is determined whether or not there is an unregistered example (S56). If there is an unregistered example, then an approximate example is extracted from an approximate example dictionary (not shown) (S57), and an unregistered example is extracted. At the same time, the translated words are combined with reference to the translation rules and the appropriate translation dictionary to obtain a translated sentence of the e-mail (S58). On the other hand, when the approximation example is used in step S57, the approximation example is stored in the example dictionary as one of the learning functions of the example dictionary, and is used when the next reference is made to the example dictionary. It can be useful.

By combining the translated words, a translated sentence is created, sent to a predetermined URL together with the original sentence, a corporate advertisement is attached, and output to the destination client via the Internet.

The translation program delivered by the operator of the machine translation system is divided into an analysis unit and a conversion / generation unit, and the analysis of the flow chart described in Fig. 7 is performed. By installing the program for translation and the program for converting and generating flowcharts similar to those shown in Figs. 8 and 9 on the sending client and the translation server in each country, respectively, Can be distributed.

In addition, by displaying the corporate advertisement in parallel with the original and translated text in e-mail, etc., the operator of the machine translation system can also earn advertising revenue. Also, in the above embodiment, an example is shown in which the text and the translation are displayed in parallel on the display of the receiving client. However, even if there is no text, if the translated text alone does not hinder the receiving client, It is good to display only the translation. In such a case, the translation server can simply generate the translation by ignoring the text and converting the data from the analysis data, so that the translation processing can be easily and quickly performed.

(3) Example of machine translation

Figure 10 shows an example of the case of handling by this machine translation system in the case of converting from Korean to English and from Japanese to English.

Figure 10 (A) shows the case of translation from Korean to English. (A) In the analysis part processing, the original sentence is divided into words separated by [violence], analyzed, and (b) in the conversion part processing. In addition, the results are shown by analyzing the original sentence co-lingual word dictionary in the translation server, separating the original sentence by [na], and then extracting the corresponding translation. After that, (c) shows the result of generating the translated sentence by the generation unit processing.

Similarly, Fig. 10 (B) shows the case of translation from Japanese to English. (A) In the processing of the analysis part, the analysis is started from two characters of the original sentence, and is converted to three or four characters. As a result of collation with the language word dictionary, the original sentence is divided into words separated by [Okina] and analyzed. Next, (b) in the conversion unit processing, the original sentence is further analyzed by the same-language word dictionary in the translation server, and the original sentence is divided by [na] for each word of the minimum unit, and then the corresponding translation word is extracted from the translation word dictionary. I do. Then, (c) in the generation unit processing, according to the English translation rules and the appropriate dictionary, for example, the first character is In capital letters, leave a space between words, and in question text [? ] Shows the result of generating a translation according to rules such as loading

Note that Korean and Japanese have the same word order, such as subject, object, and verb, so mutual translation can be performed without using a translation table from the conversion unit to the generation unit. Direct translations can be combined.

In the above embodiment, the transfer method is described as a distributed method. However, even in the case of the pivot method, a delimiter is provided for each sentence on the client side by a program of a one-sentence analysis unit, and a server side uses an example dictionary or the like. By performing the conversion, the translation according to the example can be performed, so if the analysis unit is loaded on the client and the translation server is equipped with a conversion unit, the translated sentence can be output as in the transfer method.

In the above embodiment, an example of an e-mail to a foreign language mainly using another language has been described. However, when translating to another language, the e-mail is not limited to an e-mail that needs to be translated in real time. Even if it is an e-mail with an attached document, the attached document may be translated, or the present invention may be used to translate a general document. In addition, not only between languages, but also domestic languages such as standard language, Osaka dialect and Satsuma language may be applied to the machine translation system.

In addition, in a telephone using the Internet called an Internet telephone, the original speech is decoded as text, and the text is analyzed using the machine translation system of the present invention, and the provider or the bridge is used. A real-time Internet telephone system may be operated by performing conversion Z generation processing in an intermediate processing server or the like such as J / RO and converting it into a predetermined voice.

Industrial applicability According to the above-described present invention, when a translated sentence is created from an original sentence by a transmitting client and a translation server via the Internet, an analysis part of a transfer method is used instead of a case where the translation is performed collectively in the server. It is executed by the client, and the translation and generation parts are distributed and processed by the translation server, reducing the load on the translation server. In particular, since a translation server corresponding to each language is used, there is a translation server Because the native language dictionary can be used, translations can be expressed in everyday words, and can be displayed in a way that is unique to the Internet compared to expressions that are literary and rigid.

Claims

The scope of the claims

1. A translation server used for a machine translation system that translates an original sentence into another language using a network, receives the original sentence analysis data obtained by analyzing the original sentence in units of words, and converts the original sentence analysis data into an appropriate translation dictionary. A conversion unit that converts each word into the multilingual analysis data of the other language based on the Z or the original sentence synonym word dictionary and the translation word dictionary; and a translation rule and a translation rule based on the other language analysis data converted by the conversion unit. A generating unit that generates a translated sentence that is a sentence in another language based on the translated word dictionary, wherein the generating unit generates and outputs a synthesized sentence that can display the original sentence and the translated sentence as a pair. Characteristic translation server.

2. In the translation server according to claim 1, the original sentence may be output as it is based on the original sentence co-language word dictionary for unparsed words included in the original sentence analysis data. Translation server to do.

3. Analyze the original text using the network, leave the analysis result to the agent as an analysis file, convert the analysis file from the analysis file to foreign language analysis data based on the dictionary database, and translate each word into another language. The client on the sending side used for the machine translation system that generates a translation from the analysis data and sends it to the client at the destination address,

A communication processing unit capable of communicating with the network; an input unit for inputting the original sentence; an analyzing unit for separating and analyzing the original sentence in word units; and an analysis unit for analyzing the analysis data analyzed by the analyzing unit. An output unit that outputs the word as an email to the communication processing unit.The analysis unit separates and analyzes the word for each space according to each national language, or extracts the word according to a symbol. Client to do.

4. In a machine translation system that translates the original text into another language using a network, It consists of a client, a translation server, a destination client, and an agent that mediates them.

A communication processing unit capable of communicating with the network; an input unit for inputting the original text; an analyzing unit for separating and analyzing the original text in word units; and analysis data analyzed by the analyzing unit. And an output unit for outputting an e-mail to the communication processing unit.

The translation server receives original sentence analysis data obtained by analyzing the original sentence in units of words, and, based on the proper sentence analysis data and the appropriate translation dictionary and / or the original sentence co-lingual word dictionary and the translated word dictionary, translates the foreign language of each word. A conversion unit that converts the data into multilingual analysis data; and a generation unit that generates a translation sentence in another language based on the translation rules and the translation dictionary from the other language analysis data converted by the conversion unit. A machine translation system, wherein the generation unit generates and outputs a composite sentence capable of displaying the original sentence and the translated sentence as a pair.

5. The machine translation system according to claim 4, wherein an advertisement is placed alongside the display of the translation to the destination client.

6. On a computer-readable recording medium storing a translation program installed by a translation server used in a machine translation system for translating an original text into another language using a network,

The translation server analyzes the first analysis data obtained by analyzing the original sentence for each word of the minimum unit by the transmitting client based on the word dictionary of the language synonym re-installed, and The translated data is converted to translated data based on the translated dictionary installed for the analysis data in Step 2, and the translated translated data is generated and generated based on the installed translation rules and appropriate translation rules. A computer-readable recording medium storing a program, wherein the translated sentence is transmitted to a destination client.