JP2021022203A - Translation evaluation device, program for controlling translation evaluation device, and translation evaluation method using translation evaluation device - Google Patents

Abstract

To provide a translation evaluation device capable of easily evaluating a translation sentence, such as checking whether there is no omission of translation or not.SOLUTION: A translation evaluation device evaluates a result in which a sentence described in a first language is translated into a sentence described in a second language. The sentence described in the first language includes a plurality of character strings. The translation evaluation device includes: determination means for determining whether or not a character string corresponding to the character string included in the first language is included in the sentence described in the second language; and presentation means for presenting the determination result by the determination means.SELECTED DRAWING: Figure 1

Description

The present invention relates to a translation evaluation device, a control program of the translation evaluation device, and a translation evaluation method using the translation evaluation device.

Research on the basic techniques of machine translation (MT), such as morphological analysis and dependency analysis, began in the 1950s. In the 1980s, as the performance of computers improved, rule-based machine translation technology came to have some practicality.

In the 1990s, a method of statistical machine translation (STM: Statistical Base Machine Translation) was developed. This statistically processes the correspondence between words and sentence structures between different languages, and selects words and translated sentences that are highly likely to be translated from translation candidates. In statistical machine translation, the translation model determines the words corresponding to the original words and the word order of the words according to probability. In addition, the language model determines the correct arrangement of translated words according to probability. In statistical machine translation, a large amount of bilingual data (training data) is required to build a translation engine (translation model and language model). The translation engine is tuned appropriately by comparing the translated sentence using the translation engine with the correct translation.

In statistical machine translation, in order to correctly determine the word order, the word order of the words in the original language is preliminarily changed to the word order of the destination language by parsing the sentences in the original language and using the dependency relationship. The technology to convert is known.

In the 2010s, neural machine translation (NMT: Neural Machine Translation) using deep learning by neural networks appeared. It has been put into practical use rapidly since about 2015, and is replacing statistical machine translation. Neural machine translation generally translates a sentence directly (end to end) with a neural network without preprocessing the original translated text (such as changing the word order).

In neural machine translation, words are treated as real values of vectors and as distributed representations of hundreds of dimensions. That is, the vector expresses the meaning and syntactic information of a word (Word Embedding). From the semantic relationship between one original word and its translated word, the semantic relationship between the other original word and its translated word can be inferred to some extent, so that flexible translation is possible.

In machine translation, the length of the input data (the length of the original text) is not constant, and it is necessary to use the data input in the past. Therefore, in neural machine translation, a recurrent neural network (RNN) is generally used. When the words (or characters) of the original sentence are sequentially input to the recurrent neural network and the code indicating the end of the sentence (EOS: End of String) is output, the translation of one sentence is completed.

In neural machine translation, since the weighting coefficient of each neuron (node) is the translation knowledge, a table for translation (a table for associating the original word with the translated word such as a phrase table) is not required. Unlike statistical machine translation, neural machine translation does not translate by replacing and rearranging the constituent words of the original text, but creates a new translation from the input text so that it matches the learned language model. It can be said that it is a thing.

JP-A-2018-120584

Compared to rule-based machine translation and statistical machine translation, neural machine translation allows for flexible translation of unlearned sentences (by Word Embedding, etc.) even if the translated sentences are fluent. There is an advantage that it is.

However, unlike rule-based machine translation and statistical machine translation, neural machine translation is not a technology that translates by replacing the original text, so it is difficult to translate the information contained in the input text exactly and exactly. There was a problem (translation may be omitted). In addition, duplicate translations may be output.

Furthermore, in neural machine translation, it is not possible to use the method used in statistical machine translation to convert the word order of the words in the original text into the word order of the translation destination language in advance, so the learning that has been used so far. There was a problem that it was difficult to use data (dictionary data, word order data, etc.).

Furthermore, in neural machine translation, while Word Embedding enables flexible translation, words with completely different meanings may be selected to create translated sentences with completely different meanings (especially words that appear infrequently). Infrequently occurring proper nomenclature, etc.). In addition, in the case of long sentences and complicated sentences, translated sentences that do not make sense may be created.

Furthermore, in general, neural machine translation, unlike rule-based machine translation and statistical machine translation, makes it difficult for humans to understand the translation process, and predicts whether or not an accurate translation will be output. There was a problem that it was difficult to do. As a result, in neural machine translation, there is a problem that mistranslation is likely to occur and the mistranslated sentence is easily provided as the final product.

For everyday conversations where you only need to get an overview even if there are some mistranslations or omissions in translations, or for documents for communicating less important information, the fluency and flexibility of the generated translations rather than the accuracy of the content. May be emphasized. Neural machine translation is extremely useful for such documents.

On the other hand, for documents that require strict accuracy such as laws and patents, mistranslations and omissions of translations are fatal. Therefore, even if neural machine translation is used, it is necessary to spend a considerable amount of time and effort to check for mistranslations and omissions in translations. In addition, due to lack of checks, fatal mistranslations and omissions of translations may occur, resulting in property damage.

Another issue with neural machine translation is the handling of character decorations and styles, such as superscripts, subscripts, italics, bold, and underlined characters. In sentences, for example, in the case of superscripts, HTML indicates that normal characters are superscripts by being surrounded by <sup> ~ </ sup> tags, and character codes such as Unicode. In some cases, superscript characters are directly expressed in. If these are mixed and used for machine learning, or if the handling of character decorations and styles in the document at the time of translation is different from the handling at the time of learning, an incorrect translation result may be output.

Subscripts and subscripts are used to describe chemical symbols, chemical formulas, mathematical formulas, variables, etc., but in neural machine translation, chemical symbols, chemical formulas, mathematical formulas, variables that are seemingly similar but completely different due to the effects of Word Embedding May be output as a translation result. Chemical symbols, chemical formulas, mathematical formulas, and variables are difficult to find at first glance, so they often remain as mistranslations in translation products.

Further, in neural machine translation, there is a problem that the same word in the original text is translated into a plurality of different kinds of words. For example, although the word "address" contained in the original document has the same meaning, it may be translated as "address", "address", or "address" depending on where it appears. It may be done. Also, for example, the word "display unit" contained in the original document may be translated as "display unit" or "display portion" depending on where it appears, even though it indicates the same thing. It may be translated as "display means". Such fluctuations in notation and output of words with different meanings are not particularly preferable in the translation of legal documents and patent documents that require strictness.

Furthermore, in neural machine translation, it is necessary to prepare a huge amount of bilingual sentences for learning. In addition, since the amount of calculation for learning is large, there is a problem that a large amount of time and computer resources (GPU, etc.) are required for learning.

In order to reduce the cost of neural machine translation, it is also possible to perform neural machine translation using a learned external computer (an internet-connected site, server, etc. operated by another company or another company) that is not dedicated to the company. Be done. One business operator provides a learned neural machine translation system on the Internet, and a plurality of users (unspecified number of users) use the system on the Internet. The use of an unspecified number of people means that the site (server) is open to the public and can be accessed from various places.

When using such a system, the user of the system sends document data to a trained neural machine translation computer operated by another company via the Internet, and the result of machine translation performed on that server is transmitted via the Internet. To receive. Transmission and reception can be performed using an API (Application Programming Interface) provided by the system provider, or can be performed using a general communication protocol such as HTTP or HTTPS.

In addition, transmission and reception can also be performed via a WEB browser which is a user interface. In this case, the user pastes the sentence (text) to be translated into the form displayed on the WEB browser by copy and paste or the like, and sends it to an external neural machine translation computer. The translation result, which is the response from the neural machine translation computer, is displayed on the WEB browser. As a data exchange format in these communications, a JSON (JavaScript Object Notation) format or the like is used.

In the above-mentioned jointly usable neural machine translation system, the system user does not need to prepare a huge amount of bilingual sentences for learning, a large amount of time for learning, and maintenance. In addition, the system user does not need to own computer resources (server, GPU, etc.) for system construction. That is, the user can conveniently reduce the time required for learning the neural network, the hardware cost, the maintenance cost, and the labor while using the neural machine translation service.

On the other hand, when translating highly confidential documents (for example, legal documents such as contracts, especially unpublished patent documents, internal confidential documents, research and development documents, etc.), these secrets are externally transmitted via the Internet. High-quality documents should be avoided.

Even if the communication path is encrypted by using HTTPS communication or the like, the transmitted document is naturally converted into plain text at the destination of communication. Therefore, if you use services provided by other companies, you should avoid sending highly confidential documents over the Internet after all. Ensuring security by encrypting the communication path and security risk by passing information that should be kept secret to the outside (such as other companies) are different dimensions and need to be dealt with independently. There is.

In this regard, it is conceivable to conclude a confidentiality contract with the destination of the document (a business operator that provides a neural machine translation service on the Internet). However, since information is easily leaked in the Internet society and it is difficult for the leak to be revealed, it is not feasible and dangerous to conceal information by a non-disclosure agreement. In addition, the reliable performance of such non-disclosure agreements is not guaranteed. Moreover, it is difficult to regularly supervise and control the ongoing performance of nondisclosure agreements.

Today, personal computers and smartphones have become familiar tools, and the Internet and WEB browsers have become daily necessities that anyone can easily use. On the other hand, there are many cases where highly confidential documents are easily sent to machine translation services on the Internet, with the help of lack of understanding of Internet systems, communication protocols, and information transmission routes. In addition, the risk of information leakage due to the use of machine translation services on the Internet is often overlooked.

In the translation service and dictionary service via WEB, the user inputs a sentence or word to be translated from a WEB browser or the like (this includes input methods such as typing, voice input, and copy and paste). The input sentences and words are transmitted to the server owned (or managed) by the provider of the translation service and the dictionary service, and are accumulated together with the sender's information.

Providers of translation services and dictionary services can set usage policies (terms of use) for accumulated information, but some businesses do not. In addition, depending on the terms of use, "When the user uploads, provides, sends, etc., the user uses the content to the service provider (and a third party who cooperates with the service provider). , Retain, copy, modify, create derivatives, (publicly) transmit, publish, publicly display, and grant worldwide licenses for distribution. "

Confidentiality is achieved by a translator of a translation company or an individual translator who has received an order from a translation company by easily inputting highly confidential sentences (or some sentences or words) into a web browser of a personal computer or smartphone. There are many cases where high-quality sentences are leaked to the outside (especially overseas). There is also the problem that such leaks have continued for many years without being noticed by the translator himself, the translator's manager, the translation company, or the person who ordered the translation.

In particular, leaking highly confidential documents such as contracts, R & D documents, and patents before publication to the outside (especially overseas) is not only from the perspective of protecting the interests of a company, but also from the perspective of protecting national interests. You absolutely need to avoid it.

For example, the following items are described in the material of the press release of the Information-technology Promotion Agency, Japan on February 20, 2015, "Beware of unintentional information leakage of the contents entered in the" alert "cloud service." Has been done. (Quoted from https://www.ipa.go.jp/about/press/20150220.html.)

"IPA (Information-technology Promotion Agency, Japan, Chairman: Kazumasa Fujie) responded to the fact that the text entered in the translation service on the Internet was published as it was on the Internet. For the purpose of raising the awareness of users when using cloud services, we have decided to issue a new warning about the attitude of using cloud services. In recent years, various cloud services have been enhanced, and the use of services not only for businesses but also for individuals is increasing. For example, by saving data such as photos and materials on the Internet, there are services that can be used anytime, anywhere, and can be easily translated. However, while cloud services have spread rapidly due to their convenience, the problem of unintended information leakage has been repeatedly pointed out because users used the service without correctly recognizing the content and risks of the service (" * 1).

In 2013, users of the Google Group, one of the services provided by Google, did not correctly recognize the information disclosure range setting, so even non-related parties could view the exchange. I'm having a problem. In addition, in the IPA survey (* 2), "I don't care at all" or "I don't care much" about "the information entered in the browser, search history, etc. are collected by the company that provided the browser". Since the number of people was 31.3% of the total, it is necessary to improve the awareness and knowledge of users regarding cloud services.

On the other hand, the service provider is also required to explain the content of the service and the handling of information on the service side in an easy-to-understand manner in order to prevent information leakage due to lack of explanation to the user and lack of recognition of the user. 』\

In addition, the following items are stated in the "Call for February 2014" by the Information-technology Promotion Agency, Japan. (Quoted from https://www.ipa.go.jp/security/txt/2014/02outline.html.)

"Are you leaking information to the outside without knowing it?" ~ Tips for using cloud services ~ (Omitted)

While cloud services are convenient to use, it is inevitable to pass some information to the service provider.

The three cloud services in Table 1 below are services that you casually use in your daily work. In particular, the "online translation service" allows you to easily translate sentences without installing translation software on your computer, but it can also be said that the original "text itself is handed over to the business operator".

Please be aware of this and use it with caution when handling confidential information in your business.

Table 1: How to use cloud services that may lead to information leakage

Online translation service

Overview Copy and paste the document you want to translate onto a web page, and the translation program on the service provider will automatically translate it.

Risk when using

The contents of the original translation document copied and pasted will be sent to the outside, and if the contents correspond to confidential information, it will violate the company regulations without knowing it.

Possible damage example

An employee was exchanging contract negotiations with an overseas company via encrypted email. The decrypted email was written in English, so I translated it into Japanese using an online translation service. As a result, the contract information was leaked. I used to use encrypted e-mail to prevent information leakage, but the information was leaked through another route called online translation. 』\

Despite the above alerts, the security awareness of those who use external translation systems (such as the cloud) remains low due to their high convenience.

An object of the present invention is to provide a translation evaluation device capable of easily evaluating a translated sentence.

In order to achieve the above object, according to a certain aspect of the present invention, in the translation evaluation device for evaluating the result of translating a sentence written in a first language into a sentence written in a second language, the first The sentence described in the language contains a plurality of character strings, and the translation evaluation device includes the character string corresponding to the character string included in the first language in the sentence described in the second language. It is provided with a determination means for determining whether or not the determination means, and a presentation means for presenting the determination result of the determination means.

Preferably, the sentence described in the second language is a sentence obtained by machine-translating the sentence described in the first language.

Preferably, the character string is either a character string indicating a word, a character string indicating a word and a reference code immediately following the word, or a coded character string, or a character string indicating a word, a word immediately following the word. A character string including either a character string indicating a reference code or a coded character string.

Preferably, the translation evaluation device replaces a character string indicating a certain word or a character string including a character string indicating a certain word in a sentence written in the first language with another character string corresponding thereto. The sentence described in the second language, which is the result of machine translation of the sentence, is acquired, and the determination means determines whether the replaced character string is included in the sentence described in the second language. To judge.

Preferably, the determination means makes a determination for each unit constituting the document, and the presentation means presents the determination result for each unit constituting the document.

Preferably, the determination means makes a determination for the entire document, and the presentation means presents a determination result for the entire document.

Preferably, the determination result is at least one of a character string not included in the sentence written in the second language and a character string not included in the sentence written in the second language. Including.

Preferably, the determination means determines whether or not the character string corresponding to the character string included in the second language is included in the sentence described in the first language.

According to another aspect of the present invention, in a control program of a translation evaluation device that evaluates the result of translating a sentence written in a first language into a sentence written in a second language, the translation evaluation device is a computer. The sentence described in the first language contains a plurality of character strings, and the control program of the translation evaluation device includes the character string corresponding to the character string included in the first language. A computer is made to execute a determination step of determining whether or not the text is included in a sentence written in a second language and a presentation step of presenting the determination result of the determination step.

According to still another aspect of the present invention, in a translation evaluation method using a translation evaluation device that evaluates the result of translating a sentence written in a first language into a sentence written in a second language, the translation evaluation is performed. The device includes a computer, the sentence written in the first language contains a plurality of character strings, and the translation evaluation method includes a character string corresponding to the character string included in the first language. It includes a determination step of determining whether or not the sentence is included in the sentence described in the second language, and a presentation step of presenting the determination result of the determination step.

According to the present invention, at least one of the above problems can be solved.

It is a block diagram which shows the structure of the translation system in 1st Embodiment of this invention. It is a functional block diagram which shows the structure of the translation system in 1st Embodiment of this invention. It is a figure which shows the specific example of the data structure stored in the database 207. It is a flowchart which shows the Japanese-English translation processing of the computer program included in the translation system in 1st Embodiment of this invention. It is a flowchart following FIG. It is a flowchart which shows the word of the computer program included in the translation system in 1st Embodiment of this invention, and the code registration process. It is a flowchart which shows the word of the computer program included in the translation system in the 2nd Embodiment of this invention, and the code registration process. It is a flowchart which shows the Japanese-English translation processing of the computer program included in the translation system in 3rd Embodiment of this invention. It is a flowchart which shows the Japanese-English translation processing of the computer program included in the translation system in 4th Embodiment of this invention. It is a figure which shows the specific example of the data structure stored in the 1st database in the database 207 in the translation system in 5th Embodiment of this invention. It is a figure which shows the specific example of the data structure stored in the 2nd database in the database 207 in the translation system in 5th Embodiment of this invention. It is a flowchart which shows the Japanese-English translation processing of the computer program included in the translation system in 5th Embodiment of this invention. It is a flowchart following FIG.

[First Embodiment]

FIG. 1 is a block diagram showing a configuration of a translation system according to the first embodiment of the present invention.

With reference to the figure, the translation system is composed of a computer 100 operated by a user (system user), a LAN (Local Area Network) 300, an Internet 400, another company's computer resource 200, and an in-house computer resource 500. Will be done.

The computer 100 operated by the user includes a CPU 101, a communication unit 103, an I / O 105, a ROM 107, a RAM 109, a storage device 111, an audio input / output unit 113, a display 115, a graphic unit 117, and the like. It is configured to include a keyboard 119 and a mouse 121.

The computer 100 is connected to the LAN 300, and the LAN 300 is connected to the Internet 400. The company's computer resource 500 is connected to the LAN 300, and another company's computer resource 200 is connected to the Internet 400.

Here, the computer resources include a server, a personal computer, a storage device (storage in a computer, NAS (Network Attached Storage), etc.), an information communication path (network, load balancer, switch, router, etc.), and a configuration thereof. CPU, memory, storage device (hard disk, optical or magnetic storage device, SSD and other semiconductor devices), ROM, motherboard, keyboard, mouse, input device such as microphone, output device such as display and speaker, and storage device. Indicates all or part of software such as computer programs, data, etc. that are read from and generally run on memory.

The in-house computer resource 500 is a computer resource owned by the user (or the organization to which the user belongs). The organization may be an organization consisting of one user or an organization composed of a plurality of personnel.

The company's computer resource 500 and the computer 100 are not directly connected to the Internet 400, and are connected to the Internet 400 via a router (not shown) of the LAN 300. Further, by providing a firewall function on the router, the company's computer resource 500, and the computer 100, it is prohibited to access the company's computer resource 500 and the computer 100 from the Internet 400 side. As a result, the company's computer resource 500 and the computer 100 are not operated from the outside via the Internet 400, and access to the data stored in the company's computer resource 500 and the computer 100 from the outside is prohibited. ..

The other company's computer resource 200 is a computer resource that is not owned by the user (or the organization to which the user belongs). The user can temporarily borrow the computing power and storage of another company's computer resource 200 by accessing from the computer 100. The usage pattern does not matter whether it is paid or free of charge.

The original text data of the translation, which is a highly confidential sentence, is described in HTML or plain text, and is recorded in the company's computer resource 500 or computer 100. When the original data of the translation is an image, it is converted into an HTML document or a text document by the software of OCR (Optical Character Recognition / Reader) recorded in the company's computer resource 500 or the computer 100, and is processed. To.

The communication protocol between the company's computer resource 500 and the computer 100 and the computer resource 200 of another company is performed by HTTPS (Hypertext Transfer Protocol Security) or the like. That is, since HTTP communication is performed on secure encryption (data encryption method using private key and public key) provided by SSL (Secure Sockets Layer) / TLS (Transport Layer Security) protocol, the company's computer The communication content between the resource 500 and the computer 100 and the computer resource 200 of another company can be concealed, and the communication content can be prevented from being leaked to a third party. Encrypted communication other than HTTPS may be performed as long as the communication content can be prevented from being leaked to a third party.

The communication content encrypted by HTTPS or the like is naturally decrypted by the computer resource 200 of another company. Therefore, even if a communication protocol such as HTTPS is used, it is not desirable to transmit the plaintext of the user or the document to be kept secret in the organization to which the user belongs to the computer resource 200 of another company. In addition, the operator of the computer resource 200 of another company may have agreed as a terms of use to reserve the right to disclose or use the sent data. In such a case, it is technically possible to send the plaintext of the document to be kept secret within the user or the organization to which the user belongs to the computer resource 200 of another company, but it is not business-wise. It is possible.

The user does not know the details of how the communication content returned to plain text by the computer resource 200 of the other company is actually stored and used by the other company. Further, in general, the user cannot know the configuration and connection of the computer resource 200 of the other company, and cannot know the specific handling of the received data inside the other company.

Considering that the leakage of confidential information occurs frequently in reality and it is difficult to prevent it, even if a confidentiality agreement is concluded with the operator of another company's computer resource 200, it is still kept secret. It can be said that it is not desirable to send the plain text of the document to be kept to the computer resource 200 of another company.

That is, even if encrypted communication such as HTTPS is used, the text to be translated, which is a highly confidential text, or a part thereof should not be transmitted to an external computer resource 200 of another company.

Therefore, the computer 100 in this embodiment encrypts a part of the text to be translated, which is a highly confidential text, by replacing it with a character string that is difficult to decipher (hereinafter, this is referred to as "first encryption"). After that, when the text is transmitted to an external computer resource 200 of another company, it is further encrypted by a protocol such as HTTPS (SSL encrypted communication) (hereinafter, this is referred to as "second encryption"). ..

The communication content encrypted by the second encryption in the computer resource 200 of another company is returned to plain text by a protocol such as HTTPS (since it corresponds to the "second encryption", this is referred to as "2" below. Second decryption "). However, the first encryption is not decrypted, and the document as encrypted by the first encryption is translated by the translation program in the computer resource 200 of the other company. The translated text (text document or HTTPS document) created by the translation is encrypted by a protocol such as HTTPS (that is, encrypted by the method of "second encryption") and transmitted to the computer 100 via the Internet 400. Will be done. The computer 100 returns the received text to plain text by a protocol such as HTTPS, which is a method of "second decoding". In this state, the first encryption has not yet been decrypted.

The computer 100 performs "first decryption", which is a decryption process for "first encryption", on the received translated data. In addition, subsequent processing is performed as necessary. As a result, the user of the computer 100 can obtain the completely translated sentence.

Even if the communication data is leaked to other than the company and other companies shown in FIG. 1, it is double-encrypted by the first encryption and the second encryption. This prevents information from being leaked to a third party other than the communicating party.

Further, the other company shown in FIG. 1 can obtain the decrypted information for the second encryption, but cannot decrypt the first encryption. Therefore, the leakage of information from the company shown in FIG. 1 to the other company shown in FIG. 1 is also prevented. That is, the other company (human being on the operator side of the computer resource 200) shown in FIG. 1 can only obtain the first encrypted text. Also, the translated text remains the first encrypted. For this reason, the other companies shown in FIG. 1 cannot obtain either the unencrypted complete untranslated data or the unencrypted fully translated data.

The first encryption needs to be such that the computer resource 200 of another company can perform machine translation of the encrypted text, and the encryption is strong enough that no one other than the company can understand the contents. There is.

FIG. 2 is a functional block diagram showing a configuration of a translation system according to the first embodiment of the present invention.

In the storage device 111 of FIG. 1, a plurality of computer-readable executable programs that operate the computer 100 as a translation device by being sequentially executed by the CPU 101 are recorded. The program is read from the storage device 111 and expanded on the RAM 109. The program is executed by the CPU 101 on the RAM 109.

By executing the program, as shown in FIG. 2, the computer 100 has a control unit 201, a communication unit 103, a search / replace unit 203, a document editing / word registration unit 205, a database 207, a memory 209, a storage device 111, and a display. / Exhibits functions as an output unit 211 and an input unit 213.

The control unit 201 is a functional block that performs various controls for the entire translation device. The communication unit 103 is a functional block for communicating with internal and external computer resources. The search / replace unit 203 searches for a specific element (word, phrase, etc.) in at least one of the document to be translated and the document after machine translation, and the searched element is a character string or code corresponding to the searched element. It is a functional block that replaces with. The document editing / word registration unit 205 is a functional block for editing at least one of a document to be translated and a document after machine translation, and registering a dictionary for word conversion.

The database 207 is a database that mainly registers a word dictionary. Database server software may be used as the database 207, or by recording a simple text file such as a CSV file or a table of spreadsheet software (EXCEL, etc.), searching for it, and reading the corresponding data. It may be a database.

The memory 209 is a working area for temporarily recording data. The storage device 111 is a non-volatile storage device, and stores data before translation, data after translation, data recorded in a database, data in memory 209, and the like. The display / output unit 211 is a user interface for a display, a speaker, or the like. The input unit 213 is a user interface such as a mouse, a keyboard, and a microphone.

FIG. 3 is a diagram showing a specific example of the data structure stored in the database 207.

Here, the database 207 is composed of a table having numbers (record numbers, serial numbers of data), words, codes, bilingual words, etc. as column names (column names, field names).

Nouns are registered as words, but verbs, adverbs, adjectives, etc. may be registered. The nouns also include compound nouns consisting of a plurality of nouns (such as "electric vehicle" which is an electric vehicle + an automobile).

The number column records the serial number of one record. In the word column, the words in the original translation are registered. The cipher column registers the cipher corresponding to the word. The parallel word column registers the translated word corresponding to the word (and cipher).

For example, in FIG. 3, translation data used for Japanese and English translation is registered. That is, a Japanese word, a corresponding English word (translation word), and a corresponding code are registered. By using the table in Fig. 3, Japanese words can be converted to the corresponding ciphers, Japanese words can be converted to the corresponding English words, and ciphers can be converted to the corresponding Japanese words. It is possible to convert the cipher to the corresponding English word, convert the English word to the corresponding cipher, and convert the English word to the corresponding Japanese word.

In addition, by adding a column and registering bilingual words in other languages in the table of FIG. 3, translation between three or more languages is possible. For example, by adding a column of bilingual Chinese words, it is possible to translate Japanese-English, English-Japanese, Japanese-Chinese, Chinese-Japanese, English-Chinese, and Chinese-English (and conversion of words in each language into ciphers).

As an example, as a record of the number "0", the database records the original word of "electric vehicle", the encrypted character string of "AAA", and the bilingual word of "electric vehicle" in association with each other.

Such a database may be created by the user in advance, or may be created during translation work or other than translation work by repeating the process shown in the flowchart of FIG. 6 or 7. In addition, the vendor can create a database and provide it to the user.

By using such a database, words can be converted into ciphers, and words can be converted into bilingual words. It is also possible to convert the code into a bilingual word. Furthermore, it is also possible to convert a bilingual word into the original word (reverse translation).

One basic word ("car", "means", "part", "vehicle", "device", "unit", etc.) may be registered in one record as a word or a bilingual word, or a compound. Words (“electric vehicle”, “signal input means”, “display unit”, “electric vehicle”, “signal input unit”, “display unit”, etc.) may be registered in one record.

The code here consists of three letters of the alphabet and functions as a virtual word during machine translation. Here, the cipher is expressed as a 26-ary number in which a record number (number) described in a decimal number starting from AAA is assigned to a numerical value of 0 to 25 in each of the alphabets A to Z.

That is, the last digit of the cipher is the 0th power of 26, the last 2nd digit of the cipher is the 1st power of 26, and the last 3rd digit (most significant) of the cipher is 26 squared. It is the place of.

For example, if the "number" column is 0, the code of "AAA" (A indicates zero in 26-ary) is applied, and if the column of "number" is 1, "AAB" (A is 26-ary). Indicates zero, and B indicates 1 in 26). Similarly, for example, if the "number" column is 3501, the code of "FER" (26-ary number 3501) is applied.

That is, the uppermost (last 3rd digit) of the cipher is assigned the alphabet corresponding to the integer part of the value obtained by dividing the number by the square of 26 (that is, 676). The middle part (last second digit) of the cipher is assigned the alphabet corresponding to the integer part of the value obtained by dividing the remainder of the above division by the first power of 26 (that is, 26). The lowest (last digit) of the cipher is assigned the alphabet corresponding to the remainder of the division. This is equivalent to converting from decimal to 26.

The number of digits in the cipher is not limited to three. Also, of the three-digit ciphers that have meaning as words (for example, CPU, RAM, USB, NOX, etc.), do not register them as reserved words together with the corresponding numbers in order to prevent misunderstandings and mistranslations. May be.

Further, here, the record number converted into a 26-ary number is used as the encryption, but the encryption may be randomly determined or may be determined by another rule.

The same word and translated word may be registered. For example, for chemical formulas and abbreviations (HCl, CPU, etc.), words and translated words may be registered as the same. In this case, the same word is registered in both Japanese and English. Further, the word (Japanese) may be a full-width character and the translated word (English) may be a half-width character.

Further, a word may be registered in which a part of the original word is described in the translated language, such as the word is "information display unit" and the translated word is "information display unit".

The cipher may be expressed in 26-ary notation following the usual hexadecimal notation (0-9, A to F are used), and may be expressed using 0-9, A to P. In addition, lowercase letters and symbols (! ”# $% &, Etc.) may be used. The cipher may be a random character string that does not overlap between a plurality of ciphers. The characters constituting the character string are alphabets. It is desirable that it consists of symbols or a combination of alphabets and symbols. Pictograms such as face letters may be used as encryption.

As in the present embodiment, the cipher is AAA, AAB, AAC, ..., ZZY, ZZZ, one digit is a 26-ary number from A to Z, and it is represented by a three-digit (three-character) alphabet. For example, 26 * 26 * 26 = 17576 words can be registered in the database. If there is a shortage, 26 power records (word, cipher, and bilingual word pairs) can be registered in the database by increasing the number of digits of the cipher.

As in the present embodiment, by expressing the code in a 26-ary number with A to Z as 0 to 25, there is an advantage that it is not possible to know at first glance what the word indicates. Further, in neural machine translation, since the character strings of alphabets and symbols are recognized as one word, there is an advantage that the machine translation can be performed while leaving the code as one word in the original text.

That is, when translating using neural machine translation, the translated word is not applied to the character strings of alphabets and symbols and is output as it is (however, there are exceptions).

For example, if the original translation of "the electric vehicle 100 includes the engine 101" is replaced by the table of FIG. 3, the ciphertext of "AAA 100 includes the AAB 101" is generated. This sentence is translated into the sentence "AAA 100 includes AAB 101." by neural machine translation. After the neural machine translation is performed, the translated sentence of "electric vehicle 100 includes engine 101." Is obtained by converting the code in the table into the translation of the original word using the table in FIG. Can be done.

It is possible that the cipher may be converted to the plural, for example, the original text of "plural AAA" may be translated into "a plural of AAAs". In this case as well, by replacing the coded part of "AAA" in "AAAs" with the English word corresponding to the word (for example, "book"), a translated word of "books" with "s" at the end can be obtained. Can be done.

However, in this method, words such as "bus", "leaf", and "city" (words that cannot be correctly pluralized just by adding "s") are misspelled (misspelling). For this reason, it is necessary to execute a conventional spell check routine after translation to ensure that these words are spelled correctly. Also, for example, prepare a table that associates "buss" with "buses", "leafs" with "leaves", and "city" with "cities" (a table that associates misspellings with correct spellings) after machine translation. It may be a batch conversion.

In addition, depending on the learning result of the neural network, the cipher in the input original text and the cipher in the corresponding translated text may be different (a kind of mistranslation).

To prevent this, when translating the original translation by neural machine translation, it is advisable to enclose the cipher part with a character indicating that it is a special character string (insert a special character before and after the cipher part). .. For example, the cipher is converted to another cipher by inserting a space (blank) character before and after the cipher part and transmitting it to the neural network, such as "AAA 100 includes AAA 101." This (a kind of mistranslation) can be prevented.

In addition to the character indicating the space, it is also effective to enclose the encrypted part in the document before translation with hook brackets (""). Similarly, quotation characters ('', "", "", ", etc.), parentheses, double parentheses, double brackets, square brackets, double square brackets, curly braces, braces, double brackets. It is also effective to enclose it in braces, angle brackets, double angle brackets, gume, or corner brackets. As a result, neural machine translation can be performed after setting the enclosed part to be a part (encryption) having a special meaning different from other parts.

Characters inserted (added) by these processes are deleted by post-processing after machine translation. For example, if the cipher is enclosed in brackets and machine translated as "AAA" 100, it will be translated into "AAA" 100. Since the double quotation mark was added during the above processing, which was not in the original text, it is deleted after machine translation and becomes AAA 100. Deletion is possible by using a regular expression.

If you use characters other than alphabets and symbols such as kanji and hiragana for the code, there is a problem that they will be translated into English and cannot be decrypted (for example, the character string of "Asai" assigned as a code character string is "". It is translated into English as "shallow" (English translation of "shallow"), or a character string that does not have a meaning like "ahii" is translated into English as "Aai"). Therefore, it is desirable that the cipher be a character string of alphabets and symbols.

In addition, when using the table of FIG. 3, the first letter of one sentence of the translated English sentence may be lowercase (in the case of English, the beginning of the sentence should start with an uppercase letter). Against this, it is advisable to capitalize the beginning of the sentence in a later spell check. For example, the process of searching the alphabet at the beginning of a sentence (or the character after a period) with a regular expression and making it all uppercase.

The data in the database of FIG. 3 may be registered by purchasing and obtaining it from a vendor, or may be registered by a user.

FIG. 4 is a flowchart showing a Japanese-English translation process of a computer program included in the translation system according to the first embodiment of the present invention. FIG. 5 is a flowchart following FIG.

The process shown in this flowchart is executed by reading the computer program recorded in the storage device 111 onto the RAM 109 and sequentially executing the computer program by the CPU 101. The program may be saved in a compiled machine language executable format, the source code may be sequentially executed by the interpreter, or the program may be written in an intermediate language.

Here, word processing software (Microsoft's WORD, OpenOffice published in an open source system, etc.) is used, and the macro (VBA: Visual Basic For Applications, etc.) is used to execute the program shown in the flowchart of FIG. To do. The database may be a dedicated database or a CSV file, but here the database is formed using spreadsheet software such as Microsoft's EXCEL, and the database is accessed using the word processing software VBA. And. The types of word processing software and program language used are not limited in the practice of the invention.

With reference to FIG. 4, in step S101, the user opens the document file in which the sentence to be translated is recorded by the word processing software (expands the document file from the auxiliary storage device to the RAM). Also, open the spreadsheet software.

The document file may be a text file described in plain text format, a document file of word processing software, or an HTML file. Further, the characters included in the image may be a file converted into text by OCR. An image file may be input, and a file in which the character portion is converted into text data by the OCR function in step S101 may be created. In addition, the format of the file is not limited to a specific one as long as it can describe sentences. The document may be received from another computer using a file transfer protocol or e-mail software, or may be input from a storage such as a USB memory. Further, it may be input by a keyboard 119 or a microphone. The document may be downloaded from the Internet.

Here, it is assumed that the patent specification describing the electric vehicle is translated, and the document file is

"[0023]

The electric vehicle 100 includes an engine 101 and a display unit 102, and the engine 101 is connected to a signal input means 103 and a display unit 102. When the input of the signal input means 103 is high, the display unit 102 displays a warning. It is assumed that the sentence "" was described. The translation process in the present embodiment will be described by taking the translation of this sentence as an example. [0023] is a paragraph number in the document.

Preprocessing is performed on the document to be translated in step S103. This is for the following purposes:

(1) Clarify the unit of sentences (sentence breaks) to be sent to the computer resources 200 of other companies at one time.

(2) In order to reduce mistranslation, the length of one sentence processed at one time in the computer resource 200 of another company is shortened.

(3) Clarify the part where the meaning in one sentence is separated.

For example, in the present embodiment, the unit of the sentence to be transmitted to the computer resource 200 of another company at one time is from the beginning of the sentence to the punctuation mark (.). Assuming that the paragraph number is also a unit of sentences, in the preprocessing, the paragraph number part is searched using a regular expression, and a punctuation mark is added after that (in the above example, a punctuation mark is added after [0023]). ..

In addition, a compound sentence (a sentence composed of two or more coordinated clauses (a set of words including a combination of a subject and a predicate)) is less likely to be mistranslated when it is decomposed into short sentences. Therefore, in the preprocessing, the compound sentence is decomposed into two or more sentences (in the above example, the electric vehicle 100 includes the engine 101 and the display unit 102, and the engine 101 is connected to the signal input means 103 and the display unit 102. The sentence "is." Is converted into two sentences "The electric vehicle 100 includes an engine 101 and a display unit 102. The engine 101 is connected to the signal input means 103 and the display unit 102.").

A compound sentence consisting of an adverbial clause and a main clause indicating a condition, time, cause, reason, etc. is related to the adverbial clause and the main clause, so it was processed as one sentence rather than being converted into two sentences and processed. More accurate translation is possible. On the other hand, there is an antinomy that mistranslation is likely to occur when one sentence becomes long. In the present embodiment, the adverb clause and the main clause are used as one processing unit, but by inserting a line feed code between them, machine translation is performed assuming that both clauses are different (complex sentence). Machine translation is performed after inserting a line feed code between the constituent clauses). This makes it possible to make the translation more accurate. (In the above example, a line feed code is inserted after "if high" in the sentence "when the input of the signal input means 103 is high, the display unit 102 displays a warning").

The same processing may be performed not only for adverbial clauses but also for adverbial phrases, adjective clauses, and adjective phrases.

By the process in step S103

"[0023]

The electric vehicle 100 includes an engine 101 and a display unit 102, and the engine 101 is connected to a signal input means 103 and a display unit 102. When the input of the signal input means 103 is high, the display unit 102 displays a warning. Is the sentence

"[0023].

The electric vehicle 100 includes an engine 101 and a display unit 102. The engine 101 is connected to the signal input means 103 and the display unit 102. When the input of the signal input means 103 is high,

The display unit 102 displays a warning. Is converted into a sentence. (A line feed code is inserted after "If high").

In step S105, the words in the sentence to be translated after the processing of step S103 are converted into bilingual words by using the database of FIG.

In this process, the records are read in order from the record number 0 in the database of FIG. 3, the full-text search of the document to be translated is performed using the words described in the records as the target words, and the found words are the same as those in FIG. It is a process of replacing with a bilingual word described in a record. It is possible to replace the word with the code described in the record here, but since the readability for humans in the later check will be reduced, here it will be replaced with a bilingual word that can be understood by humans (translators). Is desirable.

In the example of FIG. 3, first, "electric vehicle" with record number 0 is a search word in the document to be translated, and "electric vehicle" is a word to replace it (word after replacement). The processing of the CPU searches for the word (target word) of "electric vehicle" from the beginning of the translation target document, and if it exists, it is replaced with "electric vehicle". Before the replacement, display a dialog box such as "Do you want to replace XX with △△?" To the user, and decide whether to replace the word based on the user's YES / NO input. May be. In addition, full-text search and batch replacement (replacement of all target words in a document) may be performed mechanically without obtaining the consent of the user. When the process of record number 0 is completed, it is determined whether there is a next record, and the same process is performed until the final record.

It is desirable to insert a half-width or full-width space (blank symbol) before and after the word of the replacement word in order to improve readability and to clarify that it is a word delimiter at the time of machine translation. Also, the replacement word may be enclosed in brackets ("") to make it clear that it is a definite word. Quartet characters ('', "", "",'', etc.), parentheses, double parentheses, double brackets, square brackets, double square brackets, curly braces, braces, double braces, It may be enclosed in angle brackets, double angle brackets, gume, or brackets with corners, or other symbols. In this way, the enclosed portion may be enclosed by a symbol indicating that it is different from other portions.

If the data is not yet registered in the database (FIG. 3) (initial stage of system introduction, etc.), the process of step S105 in FIG. 4 is passed (it is not executed and the text to be translated does not change). ..

Assuming that the data shown in Fig. 3 is registered in the database,

"[0023].

The electric vehicle 100 includes an engine 101 and a display unit 102. The engine 101 is connected to the signal input means 103 and the display unit 102. When the input of the signal input means 103 is high,

The display unit 102 displays a warning. Is the sentence

"[0023].

The electric vehicle 100 includes an engine 101 and a display 102. The engine 101 is connected to the signal input unit 103 and the display 102. If the input of signal input unit 103 is high,

display 102 displays a warning. Is converted to the sentence. (Here, in order to improve readability, half-width spaces (blank symbols) are inserted before and after the replaced word.)

In step S107, the user visually confirms the document created in the processes up to step S105, and if necessary, edits the sentence, newly registers new words and ciphers, and batch-converts the newly registered words in the document. Do.

This is done for the following purposes:

(1) Change sentences that are still long to short sentences.

(2) Correct the erroneously converted words in the batch conversion in step S105.

(3) Words to be replaced with bilingual words and words to be encrypted are collectively converted and registered in the database.

That is, the user visually confirms the document created by the process up to step S105, and converts the sentence that is still left long into a short sentence. For example, a compound sentence consisting of two or more sentences is cut at a punctuation mark to obtain two or more sentences having the same meaning.

In addition, the user corrects the erroneously converted words in the batch conversion in step S105. There are cases where even the same word should be translated into different words due to differences in technical fields. In such a case, the user converts the word by batch conversion. Also, if necessary, update the database to the correct word, or prepare multiple databases to be used properly according to the technical field.

For words that were not converted in step S105 (particularly words before the reference code) because the words were not registered in the database, in step S107 the user newly registered new words, bilingual words and ciphers, and , Perform batch conversion in the document of newly registered words.

FIG. 6 is a flowchart showing a word and cryptographic registration process of a computer program included in the translation system according to the first embodiment of the present invention. This process is executed, for example, in step S107 of FIG. 4 in response to a user input.

With reference to FIG. 6, it is assumed that the user is inputting and editing the text in the document in step S203. If the user thinks that the new word needs to be registered in the database, the user selects the word in the document in step S205. This is done by moving the cursor to the beginning (or end) of the word to be registered when the document is displayed in word processing software, and pressing the shift key while pressing the direction key to the end (or beginning) of the word to be registered. By pressing it, the word to be registered is highlighted (or changed in color or underlined so that it can be distinguished from other parts). A word may be selected by dragging the word with the mouse.

When the shortcut key for word registration is pressed (or the display indicating word registration is selected from the displayed menu), a dialog box is displayed in step S207. The dialog box has a field for entering the translation of the selected word. The user confirms the translation of the selected word by entering the translation of the selected word in this field. In addition, a database different from that in FIG. 3 and a learned neural network are used to display candidates for bilingual words of the selected word in a dialog box and accept selection from the user to determine the bilingual words of the selected word. It may be that.

The word selected in step S205 may be a part of the word converted in step S105 of FIG. For example, when a bilingual translation with the "display unit" as the "display unit" is recorded in the database, the "information display unit" in the document is converted into the "information display unit" by the process of step S105. The user may select "information" and register the word "information" in the database as a bilingual word, but select "information display unit" and use the word "information display unit" as a bilingual word in the database. You may register.

In step S209, the same word as the word selected in step S205 is searched in full text, and in step S211 it is replaced with the bilingual word determined in step S207. Also in this replacement, as in step S105 of FIG. 4, a dialog box such as "Do you want to replace XX with △△?" Is displayed to the user before the replacement, and based on the user's YES / NO input. You may decide whether to replace the word. In addition, full-text search and batch replacement (replacement of all target words in a document) may be performed mechanically without obtaining the consent of the user.

In step S213, the bottom row in which the data of the database (FIG. 3) is registered is searched. The line immediately below it is used as the registration line for the word this time, and the code is created from the record number in step S215. In step S217, a number, a word, a code, a bilingual word, and the like are newly registered in the line immediately below the bottom line of the database. As a result, the database, which is a translation asset, is updated.

After the process in step S217, the process returns to the process from step S203. If the word is not registered in step S205, the process from step S203 is performed.

In the event of starting the word registration process in step S205, the shortcut key for word registration is pressed, the display indicating word registration is selected from the displayed menu, and ("word registration"). It may be voice input (such as "), but it is desirable to use a shortcut key (because word registration can be performed without using a mouse or voice input). It is desirable that the shortcut keys are assigned to, for example, pressing both the control key and a specific key so as not to interfere with the character input.

When the process in step S107 of FIG. 4 is completed, in step S109, it is checked whether or not there is a part in the document generated by the processes up to step S107 that is difficult to machine translate. Specifically, this is as follows.

(1) If the length of one sentence (from the beginning of the sentence to the punctuation mark) is longer than the predetermined length, a warning is output as it is difficult to machine translate the sentence.

(2) If the length from the beginning of one sentence or line feed code to the next line feed code or punctuation is longer than the specified length, a warning is output as it is difficult to machine translate the sentence. ..

(3) If there is a word that appears a predetermined number of times or more and has not been converted (not registered in the database) in the process of step S105 or S107, a warning is output to convert.

(4) If there is a word that appears before the reference code and has not been converted (not registered in the database) in the process of step S105 or S107, a warning is output to convert it.

(5) If the subject and predicate in one sentence do not correspond, a warning is output.

(6) If there is not one subject and one predicate in one sentence, a warning is output.

If the check is caught in step S109, the user continues the process in step S107. The process in step S109 may be omitted.

In step S111, a process of batch-converting English words into ciphers is performed on the documents obtained in the processes up to step S109.

This process is similar to the process in step S105, and records are read in order from the record number 0 in the database of FIG. 3, and the full-text search of the document to be translated is performed using the bilingual word described in the record as the target word. Is performed, and the found word is replaced with the code described in the same record of FIG.

This will

"[0023].

The electric vehicle 100 includes an engine 101 and a display 102. The engine 101 is connected to the signal input unit 103 and the display 102. If the input of signal input unit 103 is high,

display 102 displays a warning. Is the sentence

"[0023].

AAA 100 comprises AAB 101 and FER 102. AAB 101 is connected to AAC 103 and FER 102. If the input of AAC 103 is high

FER 102 displays a warning. Is converted to.

In this way, the word is replaced with the cipher described in the record, resulting in a document that is incomprehensible to humans (first cipher).

When reading records in order from the record number 0 in the database of FIG. 3, in the process in step S111, it is desirable to sort the data in the order of longest translation words (or in order of including more words).

For example, "display unit" and "information display unit" may be registered as bilingual words in the database. When there is a word "information display unit" in a sentence, if only the "display unit" part in it is searched first and encrypted, "information" remains unencrypted. Therefore, the long words can be encrypted first by performing the search in the order of the longest translation words (the order containing more words). Instead of sorting the data in the database, the search terms may be picked up from the database in the order of longest translation words (in order of inclusion of more words).

In step S113, it may be searched whether there is a word (particularly a noun) before the reference code that cannot be converted into a cipher. That is, if there is a word before the reference code that cannot be converted into a cipher, in step S115, the user is urged to register in the database as in step S109. By the processing of steps S113 and S115, the word before the reference code can be reliably encrypted. The processing of steps S113 and S115 may be omitted.

In step S117, the document data created by the word processor software by the processing up to step S115 is converted into text data (data such as UTF-8). Although it may be converted into HTML data, it is desirable to convert it into text data in order to reduce the amount of data to be communicated and the amount of data to be translated.

Next, in step S121 of FIG. 5, one sentence (a processing unit to be machine translated at one time) to be processed is specified in order from the beginning of the data obtained in step S117. In this process, the first processing data is from the beginning of the data to the next punctuation, the second processing data is from the next punctuation, and it is continued until the end of the sentence (the end of the nth processing data). Is. The end of the nth processed data (the end of the data) is a punctuation mark or the last character of the data.

"[0023].

AAA 100 comprises AAB 101 and FER 102. AAB 101 is connected to AAC 103 and FER 102. If the input of AAC 103 is high

FER 102 displays a warning. If it is a sentence of

The first sentence is "[0023]."

The second sentence is "(line feed code) AAA 100 includes AAB 101 and FER 102."

The third sentence is "AAB 101 is connected to AAC 103 and FER 102."

The fourth sentence is "If the input of AAC 103 is high (line feed code).

FER 102 displays a warning. ". These are processed in order from the first sentence, and are processed in step S121. The second and fourth sentences are one sentence to be processed, but the line feed code is included in the sentence.

In step S121, it is determined whether or not one sentence to be processed does not require translation. In the above example, the first sentence is the paragraph number "[0023].", Which contains only symbols and numbers. In this way, if it is one sentence consisting of only symbols, numbers, and alphabets, it is not necessary to translate it, so it is added as it is to the variable (queue) to which the translation result is assigned. Alternatively, it may be added after performing processing such as converting full-width characters to half-width characters. That is, if YES in step S121, the sentence is processed in step S123 as necessary and added to the end of the queue. As a result, communication costs and translation costs can be reduced. In addition, it is possible to prevent mistranslation from occurring.

If NO in step S121, the sentence is transmitted to an external computer resource 200 of another company by HTTPS communication in step S125. The encryption performed in HTTPS communication is the second encryption.

In step S127, the translation result of the one sentence is received from an external computer resource 200 of another company. This communication is also encrypted by HTTPS, and is decrypted by decryption (second decryption) by the computer 100. The translation result is added to the end of the queue.

When adding the translation result to the end of the queue, a space (or two spaces (two spaces)) may be inserted after the period. That is, the received documents may be combined by inserting a space between them. This is a process for improving the readability of the translated text.

In step S129, it is determined whether the processing has been completed up to the last sentence (nth sentence), and if NO, the processing from step S121 (processing of the next sentence) is repeated. As a result, the processing from the first sentence to the nth sentence is completed.

If YES in step S129, the queue contains the translation result. For example, in the above example

"[0023].

AAA 100 comprises AAB 101 and FER 102. AAB 101 is connected to AAC 103 and FER 102. If the input of AAC 103 is high

FER 102 displays a warning. Is the sentence

"[0023].

AAA 100 includes AAB 101 and FER 102. AAB 101 is connected to AAC 103 and FER 102. When an input of AAC 103 is high,

It will be converted to a statement like "FER 102 displays warning." And it will be queued. In this way, the translated text is also encrypted, so even if a person other than the company obtains the translated text, the content cannot be known.

Here, in step S131, a process of converting the code in the queue into English words is performed.

This process is similar to the process in step S105, and records are read in order from the record number 0 in the database of FIG. 3, and the full-text search of the document to be translated is performed using the code described in the record as the target word. This is a process of replacing the found code with a bilingual word described in the same record of FIG.

This will result in the above

"[0023].

AAA 100 includes AAB 101 and FER 102. AAB 101 is connected to AAC 103 and FER 102. When an input of AAC 103 is high,

FER 102 displays warning. "

The sentence is

"[0023].

electric vehicle 100 includes engine 101 and display 102. engine 101 is connected to signal input unit 103 and display 102. When an input of signal input unit 103 is high,

display 102 displays warning. "

Is converted into a sentence of. By performing this conversion in-house, the confidentiality of the document is guaranteed.

In this conversion, it is desirable to set the full-text search flag so that the cipher does not match a part of the word. For example, it prevents "ABS" in the word ABSTRACT from being erroneously converted as a cipher. Such processing is realized, for example, when Microsoft's VBA is used, a full-text search is performed by setting the MatchWoleWord property to True.

Further, in order to prevent words having meanings such as CPU and RAM from being translated as ciphers, it is desirable that words having such meanings are not registered in the database as ciphers (reserved words).

In step S133, post-processing is performed. Specifically, this is the following process.

(1) A process of deleting unnecessary spaces. For example, if two or more spaces are inserted at positions other than after the period, the process of unifying them. Also, the process of deleting the space between the numbers and the alphabet of the reference code (100A, etc.) consisting of the alphabet and the numbers.

(2) A process of deleting a line feed code existing after a comma or a line feed code other than the line feed code existing after a period. For example, in the above example, it is a process of deleting the line feed code after "When an input of signal input unit 103 is high,".

(3) Processing to convert plural errors into correct forms. For example, the process of converting boxes to boxes.

(4) The process of converting "said", which is the translation of "above", into the general word "the".

(5) When a special code is added, such as enclosing the code in brackets, and translated, the process is deleted. For example, if the code of AAA100 is machine-translated by adding brackets such as "AAA" 100, double quotation marks such as "AAA" 100 that are not in the original are added, so the process of deleting this.

(6) Processing to correct articles. The first word should have an indefinite article, and if it appears a second time or later, a definite article should be added.

(7) Processing to delete full-width punctuation marks. (This is a process to delete the punctuation marks of [0023]. Above.)

In step S135, the user checks the post-processed document and corrects any errors.

In step S137, the spell checker in the word processing software corrects any errors in the word or syntax, and at the same time corrects any errors in the plural form. The final translation is completed by the user correcting and editing the error.

In this way, machine translation using an external computer resource can be performed while maintaining the confidentiality of both the original translated text and the translated text.

Also, by registering words containing superscripts, subscripts, italics, bold, and underlined characters in the database, they are machine-translated in a cryptographically converted state. Machine translation allows superscripts, subscripts, italics, bold, and underlines because the code can be converted back into words that contain superscripts, subscripts, italics, bold, and underlined characters. Words containing subscripts are prevented from being mistranslated. For example, the word may be registered as "SiO ₂ ", the corresponding code may be "ABC", and the translated word may be registered as "SiO ₂ " in the database, or may be registered as in the record No. 3504 in FIG. Good. If the word before translation and the word of the translated word are the same, only the word and the code may be registered in the database, and the translated word may not be registered. Words may be registered in the database as full-width characters and translated words as half-width characters (or vice versa).

If the document has low confidentiality, the document partially converted into a bilingual word may be machine-translated without performing the encryption in steps S111 to S115 and the decryption in step S131. In this case, the first encryption and the first decryption are not performed, and only the second encryption and the second decryption are performed. This method is effective when it is not necessary to keep the content of the translation secret from the other companies in FIG. In this method, it is possible to prevent fluctuations in the words of the document after machine translation. In this method, machine translation may be performed after adding brackets or double quotation marks surrounding the bilingual words converted by batch conversion. As a result, machine translation can be performed in a state indicating that the word does not need to be translated, so that the fluctuation and mistranslation of the word in the document after machine translation are reduced.

There are multiple translations of "electric vehicle" such as "electric vehicle", "electrical vehicle", "electric-powered vehicle", "electronic vehicle", "battery car", and "battery vehicle". When machine translation is performed by a neural network, it is not possible to know which translated word is applied depending on the context, and the word expression after translation fluctuates. In addition, appropriate translations often differ depending on the field of the document to be translated. In the present embodiment, the words are converted into a unified translation in advance before the machine translation by the neural network. Therefore, there is no fluctuation in expression.

If Japanese-English machine translation is performed using the code as full-width characters, they will be half-width characters after machine translation. Therefore, in the conversion in step S131, it is necessary to perform full-text conversion in consideration of this. If the encryption is set as half-width characters from the beginning, such a problem does not occur.

[Second Embodiment]

FIG. 7 is a flowchart showing a word and cryptographic registration process of a computer program included in the translation system according to the second embodiment of the present invention.

The process shown in FIG. 6 is not limited to the process in the translation system of FIG. 1, and can be performed separately and independently as a word registration process (database creation process).

That is, it is also possible to perform the processing after step S203 (same as the processing after step S203 in FIG. 6) after opening the sentence with the word processor software in step S201 of FIG.

In the first and second embodiments, if the translation is performed without using the cipher (when translating a document that does not need to be kept secret), it is not necessary to register the cipher in the database. In this case, the process of step S215 becomes unnecessary.

Further, although the cipher is registered in the database, for example, a cipher uniquely determined from the record number may be used, and the cipher may not be recorded in the database and may be obtained by calculation from the record number on the computer 100.

The process shown in such a flowchart is executed by reading the computer program recorded in the storage device 111 onto the RAM 109 and sequentially executing the computer program by the CPU 101. The program may be saved in a compiled machine language executable format, the source code may be sequentially executed by the interpreter, or the program may be written in an intermediate language.

Even if the program shown in the flowcharts of FIGS. 6 and 7 is executed using word processing software (Microsoft's WORD, OpenOffice published in an open source system, etc.) and its macro (VBA: Visual Basic For Applications, etc.). Good. The database may be a dedicated database or a CSV file, but here the database is formed using spreadsheet software such as Microsoft's EXCEL, and the database is accessed using the word processing software VBA. And.

With reference to FIG. 7, in step S201, the user opens the document file in which the sentence to be translated is recorded by the word processing software (expands the document file from the auxiliary storage device to the RAM). Also, open the spreadsheet software.

The document file may be a text file described in plain text format, a document file of word processing software, or an HTML file. Further, the characters included in the image may be a file converted into text by OCR. An image file may be input, and a file in which the character portion is converted into text data by the OCR function in step S201 may be created. In addition, the format of the file is not limited to a specific one as long as it can describe sentences. The document may be received from an external computer using a file transfer protocol or e-mail software, or may be input from a storage such as a USB memory. Further, it may be input by a keyboard 119 or a microphone. The document may be downloaded from the Internet.

Here, it is assumed that the patent specification describing the electric vehicle is translated, and the document file is

"The electric vehicle 100 includes an engine 101 and a display unit 102, and the engine 101 is connected to the signal input means 103 and the display unit 102. When the input of the signal input means 103 is high, the display unit 102 warns. It is assumed that the sentence "Display." Was described.

In step S203 of FIG. 4, the user checks the contents of the opened document and corrects it if necessary. Also, add (input) sentences as needed.

In step S205 of FIG. 4, the user visually selects a word to be registered in the database by visually observing the sentence. This moves the cursor to the beginning or end of a word and highlights the target word (or changes the text and / or background color to other parts) by pressing the SHIFT + arrow keys or dragging with the mouse. It is a thing. The word registration is mainly performed on the word existing before the reference code, but the word without the reference code may also be registered. The reference code is a code for referring to a code (such as a number or a character indicating a component used together with a leader line) described in a drawing. For example, "100" of "electric vehicle 100" is a reference code, and the word existing before the reference code means "electric vehicle".

The word may be selected by human judgment (key or mouse input) as described above, but for example, the word existing before the reference code may be automatically detected and selected. Further, the word may be automatically detected and selected regardless of the presence or absence of the reference code.

If the registration button (or shortcut key for registration) displayed on the screen is pressed while the word is selected, in step S207, the CPU 101 determines the translation of the word selected in step S205. Accept input. This may display a dialog box on the screen and accept keyboard or voice input from the user, or use a UI (user interface) that displays multiple word candidates from the dictionary and allows the user to select a bilingual word from among them. You may. This determines the correspondence between the untranslated word and the translated word.

In step S209, the CPU 101 performs a full-text search of the document to be translated, searches for the untranslated word determined in the processes up to step S207, and converts it into the corresponding translated word in step S211. Full-text search may be performed without asking the user, and each time one word is searched, a dialog box such as "Do you want to replace it with (translated word)? (YES / NO)" May be displayed on the screen and the user may choose whether or not to replace the word. As a result, the word in the whole sentence including the word selected in step S205 is collectively converted into the translated word determined in step S207. For example, if the word "electric vehicle" is associated with "electric vehicle", the word "electric vehicle" in the document is collectively converted to "electric vehicle".

In step S213, the CPU 101 searches for the bottom line (bottom line in FIG. 3) in which the information in the database 207 is registered, and acquires the number (serial number) of the blank line that is the next line.

The number acquired in step S215 is converted into a 26-ary number, and the code of the target word is created. In step S217, the cipher is recorded as one record together with the untranslated word and the translated word in the line below the bottom line where the information in the database is registered. The information registered in the database will be an asset as a translation dictionary to be used in the future.

After the process of step S217, the process from step S203 is repeated. When the registration button is not pressed in step S205, the process from step S203 is repeated.

When the full text processing is complete, the user saves the document. This is the original data in which a part of the document is translated, and may be the target of subsequent neural machine translation.

The words registered in the database in the processes of FIGS. 6 and 7 are mainly the words described before the reference code. For example, the original text to be translated states, "The electric vehicle 100 includes an engine 101 and a display unit 102, and the engine 101 is connected to the signal input means 103 and the display unit 102. When the input of the signal input means 103 is high. , The display unit 102 displays a warning. ”, The“ electric vehicle ”,“ engine ”,“ display unit ”, and“ signal input means ”are subject to registration.

Words such as "input" and "warning" that do not have a reference code may also be registered in the database 207 of FIG. 4 together with the code and the bilingual word.

Further, at least a part of the word with the reference code may be excluded from the registration. That is, at least a part of the "electric vehicle", the "engine", the "display unit", and the "signal input means" need not be registered.

After the registration in step S213, the words in the original text may be sorted in descending order. The longer words are sorted so that they are closer to the number 0 (upper side) and the shorter words are located closer to the higher number side (lower side).

By the above processing, "the electric vehicle 100 includes an engine 101 and a display unit 102, and the engine 101 is connected to the signal input means 103 and the display unit 102. When the input of the signal input means 103 is high, the display is displayed. The sentence "Part 102 displays a warning."

"The electric vehicle 100 comprises engine 101 and display 102, and engine 101 is connected to signal input unit 103 and display 102. If the input of signal input unit 103 is high, display 102 displays a warning. Is converted into a sentence. The user checks this sentence to determine if there are any mistakes. If there is an error, edit it on the word processing software.

[Third Embodiment]

FIG. 8 is a flowchart showing a Japanese-English translation process of a computer program included in the translation system according to the third embodiment of the present invention.

The process shown in this flowchart is executed by reading the computer program recorded in the storage device 111 onto the RAM 109 and sequentially executing the computer program by the CPU 101. The program may be saved in a compiled machine language executable format, the source code may be sequentially executed by the interpreter, or the program may be written in an intermediate language.

With reference to the figure, in step S101, the CPU 101 inputs a document (sentence) to be translated according to a user's instruction. It is assumed that the document input here is a sentence after the conversion is performed by the process shown in FIG. Here, "the electric vehicle 100 includes engine 101 and display 102, and engine 101 is connected to signal input unit 103 and display 102. If the input of signal input unit 103 is high, display 102 warns. An example in which the sentence "Display." Is processed will be described.

In step S103, the CPU 101 searches for a part of the input document that has difficulty in machine translation. For example, when translating from Japanese, the document is decomposed into multiple sentences by separating the document with a delimiter such as "." (Period if translating from English), and each sentence is within the specified length. (That is, whether the length of each sentence is equal to or less than a predetermined threshold) (as a threshold, a value of, for example, about 80 to 120 characters is selected according to the performance of the computer resource 200 of another company. ).

If the length of one sentence is long, machine translation tends to be inaccurate. Therefore, the length of the sentence is checked in advance, and long sentences cannot be translated. For a long sentence, for example, in the case of a compound sentence, a sentence such as "A is B and C is D." And a plurality of sentences such as "A is B. C is D." Conversion / editing processing is performed to convert to sentences and shorten each sentence. Such conversion may be performed automatically, or the user may be shown a sentence judged to be long and the user may be manually prompted to correct it. Further, in the case of automatic conversion, it is desirable to show the user the text after the automatic conversion (and the text before the conversion if necessary) so that the user can check whether the conversion is performed correctly.

To check the part where machine translation is difficult, is there a subject in one sentence (for example, is there a phrase corresponding to "○○ wa" or "○○ wa") and the subject in one sentence? You may check whether the predicates correspond. If machine translation is performed more accurately without mistranslation, for example, it may be a condition of a translatable sentence that there is only one subject and only one predicate in one sentence.

These checks may be performed by matching sentence patterns using regular expressions. The check may be performed by morphological analysis. Also, for example, using a learned recurrent neural network (RNN), the sentence to be checked is input, and the sentence that is not difficult for machine translation is output as the output, or the signal with / without difficulty is output as the output. It is also possible to urge the user to make corrections in some cases.

In step S105, it is checked whether the sentence can be machine translated, and if there is a difficult part, the process in step S103 is repeated.

If it is determined in step S105 that machine translation is possible, the sentence is preprocessed in step S107. This is a process that facilitates machine translation so that there are no mistakes in machine translation. For example, a compound sentence has a structure of "when A is B, C performs D." However, in order to prevent translation errors, the delimiter part of the complex sentence (in the above sentence, immediately after "when,"). It is a process of embedding a code that serves as a mark in. The code is a line feed code, a space, a character that serves as a delimiter, etc., and a code that does not affect the translation and makes the delimiter easy for the machine to understand is selected. Further, for example, an important cultural property having a structure such as "A is B and C is D" is used as the mark at the delimiter portion of the important cultural property (in the case of the above sentence, immediately after "is"). You may embed the code. Since the unit of the sentence to be processed in the machine translation is reduced by the process of step S107, it is possible to prevent the wrong subject-predicate combination from being selected or other mistranslations in the machine translation.

By the processing in steps S103 to S107, for example, the above

"The electric vehicle 100 comprises engine 101 and display 102, and engine 101 is connected to signal input unit 103 and display 102. If the input of signal input unit 103 is high, display 102 displays a warning. The sentence is

"The electric vehicle 100 includes engine 101 and display 102. The engine 101 is connected to signal input unit 103 and display 102. If the input of signal input unit 103 is high (line feed code).

display 102 displays a warning. "

(A single culture is performed in step S103, and a line feed code is inserted at the break of the sentence that creates a compound sentence in step S107).

In step S109, the database in the storage device 111 is referred to, and the word (particularly noun) described before the reference code in the document to be translated is converted into a cipher (first encryption). The database contrasts the original text with the encrypted string and the translated string.

By batch-converting the words before the reference code into the cipher in step S109, for example, the above-mentioned

"The electric vehicle 100 includes engine 101 and display 102. The engine 101 is connected to signal input unit 103 and display 102. If the input of signal input unit 103 is high (line feed code).

display 102 displays a warning. Is the sentence

"AAA100 comprises AAB101 and FER102. AAB101 is connected to AAC103 and FER102. If the input of AAC103 is high (line feed code).

FER102 displays a warning. "

Is replaced by the ciphertext of.

In such a ciphertext, since a 26-ary number such as "AAA" is processed as a word, machine translation is possible with an external computer resource 200 of another company without leaking the true meaning to the outside.

The process in step S109 is performed as follows.

(1) From the beginning of the document, search for the component part of the sentence consisting of the combination of the word and the reference code.

(2) Search the database for the words of the searched components.

(3) Obtain the corresponding cipher from the database.

(4) The word of the constituent part searched in the above (2) is replaced with the code obtained in the above (3).

(5) Search for the component of the sentence that appears next in the sentence and consists of a combination of words and reference codes.

(6) The process from (1) above is repeated until all the components are processed. If all the components have been processed, exit.

Alternatively, the process in step S109 may be performed as follows.

(1) Read the first word registered in the database

(2) All the read words are searched from the beginning of the document and replaced with the corresponding ciphers registered in the database (full-text batch replacement).

(3) The next word registered in the database is read, and the process of (2) above is repeated. When all the registered words in the database are processed, the process ends.

In the processing after step S119, the data before machine translation obtained in the processing up to step S109 will be transmitted to the computer resource 200 of another company that has already learned the neural network by HTTPS transmission in order from the first sentence. ..

More specifically, the first sentence of the data before translation (for example, the sentence up to the punctuation mark of ".") Is transmitted by HTTPS to an external computer resource 200 of another company in step S119. This transmission is encrypted (second encryption). The external computer resource 200 of another company translates the received data into the target language and transmits it to the computer 100 in response to the request.

In step S121, the computer 100 obtains the data of one sentence after translation from the external computer resource 200 of another company as a response by HTTPS.

In step S123, it is determined whether the translation has been completed up to the last sentence, and if NO, the next sentence is processed in step S119. If the translation is completed up to the last sentence, the process proceeds to step S125.

In step S125, post-processing of the translated document is performed. Here, the translated sentences for each sentence obtained in the processes of steps S119 to S123 are put together in one document. Further, since the translated text obtained by the first encryption in steps S109 to S113 contains a cipher, a process of batch-converting the cipher into the corresponding translated word is performed. This is done, for example, as follows.

(1) Read the first cipher (such as "AAA") registered in the database.

(2) All the read ciphers are searched from the beginning of the document and replaced with the corresponding translations in the database (full-text batch replacement).

(3) The next cipher registered in the database is read, and the process from (2) above is repeated. When all the registered ciphers in the database are processed, the process ends.

In step S127, a spell check such as a check for plural words is performed. The user finally checks the translated text and corrects or edits any problems.

By performing the process in step S127,

"AAA 100 in AAB 101 and FER 102. AAB 101 is connected to AAC 103 and FER 102. When an input of AAC 103 is high, (line feed code)

FER 102 displays alert. "

From the text of

"Electric vehicle 100 engine 101 and display 102. engine 101 is connected to signal input unit 103 and display 102. When an input of signal input unit 103 is high, (line feed code)

display 102 displays alert. "

Sentence is obtained.

By performing the same processing as in steps S135 and S137 of FIG. 5 in steps S125 and S127, the same effect as that of the first embodiment can be obtained.

[Fourth Embodiment]

FIG. 9 is a flowchart showing a Japanese-English translation process of a computer program included in the translation system according to the fourth embodiment of the present invention.

The processing in this flowchart is obtained by adding the processing of steps S111 and S113 in addition to the processing of FIG.

After the process in step S109, in step S111, it is determined whether or not there is a word before the reference code that could not be converted because it is not registered in the database. This may be determined by using a regular expression whether or not a sentence component composed of a combination of a word and a reference code exists, or the user may visually confirm it.

If YES in step S111, the process proceeds to step S113. In step S113, the word that could not be converted in step S111, its cipher, and the translated word are registered in the database as a set. For example, after making the word that could not be converted stand out as a black-and-white inverted display, a dialog box containing the word "Please enter the translation of XX" and the word input field is displayed so that input from the user is accepted. You may. For example, highlight the word "electric vehicle" in the document, display a dialog box containing the word "please enter a translation of'electric vehicle'" and a word entry field, and the user will say "electric vehicle". It accepts input. A word, its code, and a translation are set and registered in the database. It is desirable to create the cipher automatically.

The process in step S113 is repeated until the result becomes NO in step S111.

If NO is obtained in step S111, the process proceeds to step S119.

[Other]

It is also possible to combine or replace the plurality of embodiments described above and the elements contained therein (partial configuration, partial processing) to obtain another new embodiment.

When performing full-sentence batch conversion (processing of converting words into bilingual words or translating words into ciphers), it is desirable that the words to be searched are read from the database in order from the longest word. For example, if there are two records in the database, one that converts "detection means" to "detection means" and the other that converts "information detection means" to "information detection means", the long word "information" After the process of batch-converting "detection means" into "information detection means", the process of batch-converting the shorter word "detection means" into "detection means" is performed. This prevents only the "detection means" part of the "information detection means" in the document from being converted, and can improve the efficiency of translation. Similarly, when converting a word into a cipher, it is possible to prevent only a part of a set of words having a series of meanings from being converted into a cipher by reading from the database in order from the longest word.

When reading words in ascending order of record number and performing batch conversion, the records may be sorted by the number of characters of the registered words.

In the case of a patent specification, there is an item "description of code" at the end, and generally important reference code and the component (word) indicated by the reference code are explained there. Only the words described in the item of the code description may be encrypted. Prior to machine translation, a warning may be issued if the word in the code description section is not encrypted in the document.

If it is a patent specification, a form for carrying out the invention, a brief explanation of drawings, an effect of the invention, a means for solving a problem, a document name specification, a title of the invention, a technical field, a background art, etc. The prior art document, the patent document, the outline of the invention, the description of the code, the scope of the document name claim, the document name abstract, etc. may be regarded as words and encrypted. This makes it difficult to understand that the text to be translated is a patent specification.

In documents such as patent specifications that are explained with reference to the reference code of the drawing, the word that is the noun before the reference code in the document has a particularly important meaning. By substituting at least a part of the word before the reference code with a code (or a bilingual word), the present invention can exert a high effect. Of course, at least a part of a word other than the word before the reference code in the document may be replaced with a code (or a bilingual word). Also, the word before the reference code in the document is not replaced with a code (or translation), and at least a part of the words other than the word before the reference code in the document is coded (or translation). It may be replaced.

Of the set of noun words and reference codes (set consisting of noun words and the reference code immediately following them) that appears in the document, only the part of the noun words is stored in the database as the target of batch conversion. It is desirable to register and not include the reference code in the target of batch conversion. By doing so, the same word with a different reference code can be processed using the database (in the future, the database can be used more effectively when translating other documents. Can be done).

The present invention is also applicable to documents that do not include reference numerals. In this case, at least some of the words in the document are replaced with ciphers (or bilingual words) and then machine translated.

The batch conversion process using the database in step S105 of FIG. 4 and step S109 of FIG. 8 may be continuously performed a plurality of times. For example, the "information detection means" in the translated document may be converted to "information detection means" by recording in the database a record in which the user converts "detection means" → "detection means" in advance. .. This is the case where some of the words have been converted. In this case, the user selects "information detection means" as a word and records "information detection means" as a new record in the database as a translation. When translating the next document using that database, if the word "information detection means" exists in the document, that word is first converted to "information detection means" by batch conversion using the database. To. If the search term "information detection means" appears after that in the search order of the database, "information detection means" is converted to "information detection means" without any problem. However, if the search order in the database is that "information detection means" is earlier than "detection means", "information detection means" can be obtained by performing batch conversion processing using the database only once. Not converted to "information detection means".

Therefore, the batch conversion process using the database is repeated twice (or more) to perform the batch conversion again. As a result, even a word such as "information detection means" that includes a partially translated word in the original text is converted into a correct word such as "information detection means".

Further, in the embodiment, an example of performing machine translation using a neural network has been given, but statistical machine translation, phrase-based and rule-based machine translation, artificial intelligence machine translation, deep learning machine translation, and the like are used. Also, the present invention is not limited to machine translation using a neural network.

The method of transmitting / receiving information to an external computer resource 200 of another company may be performed via a WEB page. Information may be sent and received by an automated method of browser operation. In addition, information may be transmitted and received using an API provided by an external business operator that operates computer resources of another company.

The post-processing may include deleting unnecessary spaces in the sentence, unifying words, converting uppercase and lowercase letters in the sentence, checking spelling, and making plural words correct.

[Effect of Embodiment]

According to the above embodiment, both the original translated text and the translated text are subjected to two-step encryption, that is, the first encryption and the second encryption. Both encryptions keep the document confidential to third parties. In addition, for the translation service provider, the confidentiality of the document is maintained by the first encryption. As a result, it is possible to avoid information leakage to the outside while using the external translation service.

Even if the computer resource that transmits the information is a computer resource used by an unspecified number of people (or a plurality of persons or others), the confidentiality of the information can be ensured. In addition, "use of an unspecified number of people" means that the site (server) is open to the public and can be accessed from various places.

In addition, the conventional neural machine translation has a problem that the information contained in the input sentence cannot be translated exactly without excess or deficiency (translation may be omitted). In addition, duplicate translations may be output. In the present embodiment, the original text can be compressed (short and simple) by encryption. In addition, the structure of the sentence can be simplified (short and simple). As a result, even if neural machine translation is adopted, the information contained in the input sentence can be translated in just proportion, and there is an effect that translation omissions are unlikely to occur. In addition, there is an effect that it is difficult to output duplicated parts of translation.

Further, there is an effect that the learning data (translation dictionary data, word order data, etc.) that have been used so far can be easily used as data to be registered in the database.

Furthermore, in the conventional neural machine translation, while Word Embedding enables flexible translation, a completely different word (sometimes a meaningless character string) may be selected to create a translated sentence having a completely different meaning. There were (especially infrequently occurring words, infrequently occurring proper nomenclature, etc.). In addition, in the case of long sentences and complicated sentences, translated sentences that do not make sense may be created. In the present embodiment, the words are batch-converted into bilingual words or ciphers in advance, and then machine-translated (preferably, machine-translated with hook brackets or the like added). This prevents machine translation from translating the original word into a bilingual word with a completely different meaning.

[Fifth Embodiment]

Since the hardware configuration of the translation system according to the fifth embodiment of the present invention is the same as that in the first to fourth embodiments, the description thereof will not be repeated. In the fifth embodiment, the sentence written in the first language to be translated contains, in part, a set consisting of a word that is a noun and a reference code that immediately follows it. The translation device includes a replacement unit that replaces a set consisting of a noun word and a reference code immediately following it with a character string corresponding to the set. The text replaced by the replacement part is transmitted to an external computer via the Internet. The translation device receives the translated text written in the second language from the computer via the Internet.

The difference between the fifth embodiment and the above embodiment will be described below.

FIG. 10 is a diagram showing a specific example of a data structure stored in the first database in the database 207 in the translation system according to the fifth embodiment of the present invention.

The data registered in this first database does not include a cryptographic column as compared with the data of FIG. Words in the first language and words in the second language, which are their translations, are recorded in association with each other (in pairs). The data in the first database may be registered by the same registration process as described in the first to fourth embodiments, or the past dictionary data and the external dictionary data may be registered in the first database. May be. Further, by registering the corresponding words in the third to Nth languages, translation into a plurality of languages may be possible.

FIG. 11 is a diagram showing a specific example of a data structure stored in the second database in the database 207 in the translation system according to the fifth embodiment of the present invention.

The data registered in the second database is created from the data recorded in the first database and the document to be translated containing a plurality of sets consisting of a noun word and a reference code immediately following it. Is. The second database contains columns for numbers, words, bilingual words, and ciphers. Of course, by registering the corresponding words in the third to Nth languages, translation into a plurality of languages may be possible. In the fifth embodiment, the second database serves as a codebook used when performing the first encryption and the first decryption.

In creating the second database, the contents of the first database are first copied (see lines 0-3504) and the code for identifying each word is recorded. The cipher is the same as that in the first to fourth embodiments.

Next, a set consisting of a word that is a noun and a reference code that immediately follows it is detected in a sentence written in the first language to be translated. In this detection, a part in which one or more nouns are continuous and a part in which a reference code such as a number or an alphabet is continued by one or more characters may be detected as a set by word-separation / morphological analysis of a sentence, or a regular expression may be detected. It may be used for detection. Further, the word registered in the first database of FIG. 10 may be searched from the document to be translated, and when the reference code follows the word, the word and the reference code may be detected as a set. For a set consisting of a word that is a noun and a reference code that immediately follows it, the word may be a word in the first language or a word in the second language (translational word).

In particular, if the sentence to be translated is written in the first language, a set consisting of a word in the first language and a reference code immediately following it is detected. If the text to be translated is written in the first language, but some words have already been replaced by the second language, then the words in the second language and the reference code that immediately follows them. A set consisting of is detected.

Here, it is assumed that the sentence to be translated includes a set consisting of the word "electric vehicle 100" and the reference code immediately following it. The "electric vehicle" in the set is replaced with the "electric vehicle" using the data from the first database of FIG. 10 (or the second database of FIG. 11). As a result, a correspondence relationship between the "electric vehicle 100" and the "electric vehicle 100" is established, a new code is given to the correspondence, and the correspondence is newly registered in the second database of FIG. 11 (No. 4012). See line).

At this time, the word "electric vehicle 100" may be registered as the word in the first language, or "" AAA "100" in which the "electric vehicle" part is encrypted with the line data of the number 0 is used. You may register. In FIG. 11, "" AAA "100" is registered.

If the cryptographic part is processed by enclosing it in special symbols such as double quotation marks, it can be clarified that the part is cryptographic, which is convenient for later processing. That is, when converting to a cipher, the cipher is enclosed in a special symbol such as double quotation marks and inserted into a sentence, or when being registered in the database of FIG. 11, the cipher is enclosed in special symbols such as double quotation marks and registered. Is what you do. Of course, it does not have to be surrounded by special symbols.

Similarly, a set of other words and reference codes included in the sentence to be translated is sequentially registered in the second database. In some documents to be translated, different reference codes are given even for the same word (for example, "gear 501a" and "gear 501b"), but in this case, they are different. It is coded and registered in a separate line (translated words are "gear 501a" and "gear 501b", respectively, and different codes are assigned to each).

By using the second database, bidirectional substitution can be performed between the word, the translated word, and the cipher, as in the first to fourth embodiments, and the word with the reference code and the word with the reference code can be used. Bidirectional substitution between the bilingual word and the cipher is also possible.

In addition, the first database is held as a database of translation knowledge (database for word conversion and word translation), and the second database is used as a database for translation of documents to be translated (so-called custom-made database). By doing so, even in a plurality of documents in which different reference codes are assigned to the same word, it is possible to associate a code or a bilingual word + reference code set with the word + reference code set.

In the present embodiment, the set of words + reference codes can be replaced with a cipher for translation. Therefore, the text to be translated (text to be sent to the server) can be shortened (the amount of data can be reduced), and the translation cost can be suppressed. In addition, since the machine-translated sentences are shortened, mistranslations are less likely to occur.

In translation, the document to be translated is encrypted using a second database. At this time, if the word in the above set in the document to be translated is the first language, the word described in the "word" column of the data in FIG. 11 is converted into the code of that line.

At this time, it is conceivable that a set of a word + a reference code exists in the document to be translated, and the word appears without a reference code. In order to shorten the encrypted sentence, first, the replacement of the word + reference code set with the cipher (in FIG. 11, the replacement using the line after the number 4012) is performed with priority, and then the cipher of the word is performed. It is desirable to perform the replacement with (in FIG. 11, the replacement using the line before the number 0 to the number 4012). As a result, a word with a reference code can be encrypted as a set of a word + a reference code, and a word without a reference code can encrypt only a word.

If a code + reference code is registered in the word column as shown in FIG. 11, the code is decoded by the same database, converted to the word + reference code, and then converted into a code. You may be asked.

If the word in the set in the document to be translated has been replaced with the second language, the word described in the "translation word" column of the data in FIG. 11 is converted into the code of that line. Will be done. At this time, it is conceivable that a set of a bilingual word + a reference code exists in the document to be translated, and the bilingual word appears without a reference code. In order to shorten the sentence after encryption, first, the replacement of the set of the translated word + the reference code with the cipher (in FIG. 11, the replacement using the line after the number 4012) is performed with priority, and then the translated word is replaced. It is desirable to replace the above with the cipher (in FIG. 11, the line after the number 4012 is used). As a result, the translated word with the reference code can be encrypted as a set of the translated word + the reference code, and the translated word without the reference code can encrypt only the translated word.

If the bilingual word + reference code is registered in the column of the bilingual word as shown in FIG. 11, it is searched in the sentence as it is and then converted into the code. The bilingual words may also be partially encrypted and registered in the database (such as "AAA" 100).

If a set of a word + a reference code is registered in the second database as a code + a reference code as shown in FIG. 11, the second database is registered in the correspondence part between the word and the code (number 0 in the case of FIG. 11). ~ 3504) and the association part between the word + reference code and the cipher (number 4012 or later in FIG. 11) are managed separately, and first, the word is encrypted using the association part between the word and the cipher. If the step of converting is executed, and then the step of converting the word into the cipher using the associating part between the word + reference code and the cipher is executed, the word with the reference code is eventually changed to the word + reference code. It can be encrypted as a set, and for words without a reference code, only bilingual words can be encrypted. Such processing can be appropriately changed depending on the algorithm. For example, in the second database, the correspondence part between the word and the cipher (numbers 0 to 3504 in FIG. 11) and the correspondence part between the word + reference code and the cipher (number 4012 or later in FIG. 11). And may be processed as a separate database.

When encrypting a word or a translated word, it is desirable to give priority to the longer word or the translated word (with a large number of characters). For example, when both "input means" (corresponding cipher is JKL) and "signal input means" (corresponding cipher is YSH) are recorded together with different ciphers in the database, the "input means" with a small number of characters comes first. This is because when the text is processed, the "signal input means" included in the text to be processed is converted into the "signal JKL", which is not preferable from the viewpoint of shortening the text and reducing mistranslation.

By preferentially encrypting longer (more characters) words and bilingual words, the "signal input means" included in the text to be processed is correctly converted to the "YSH" cipher, and the text can be shortened and mistranslated. Can be reduced.

In this way, by using the second database of FIG. 11, the set of words + reference codes contained in the sentence to be translated described in the first language is replaced with the corresponding cipher, and also. Words contained in the text to be translated written in the first language are replaced with the corresponding ciphers (first cipher). By sending the first encrypted document to an external server, the processing (translation) is performed while ensuring the confidentiality of the document.

Since the document in the second language translated by the external server contains the cipher, the decryption is performed using the second database (encryption table, codebook) in FIG. That is, the code in the sentence is replaced with the corresponding data of the translated word in FIG.

For example, suppose that a patent specification describing an electric vehicle is translated, and the document file is

"[0023]

The electric vehicle 100 includes an engine 101 and a display unit 102, and the engine 101 is connected to a signal input means 103 and a display unit 102. When the input of the signal input means 103 is high, the display unit 102 displays a warning. If the text of "" is described, in the present embodiment, the text after the first encryption is

"[0023]

The "HSD" has "HSE" and "HEF", and the "HSE" is connected to the "HEG" and "HEF". If the "HEG" input is high, "HEF" will display a warning. "

(For the sake of simplicity of explanation, preprocessing is not performed). Since the encryption can be performed as one unit including the reference code in this way, the sentence can be made shorter than that of the first to fourth embodiments, and the possibility of mistranslation is reduced. In addition, words without a reference code can also be encrypted.

If this sentence is translated by an external computer resource, the result will be

"[0023].

"HSD" includes "HSE" and "HEF". "HSE" is connected to "HEG" and "HEF". When an input of "HEG" is high, "HEF" displays warning. " To. By decoding this using the second database, the same as in the first to fourth embodiments,

"[0023].

electric vehicle 100 includes engine 101 and display 102. engine 101 is connected to signal input unit 103 and display 102. When an input of signal input unit 103 is high, display 102 displays warning. "

(For the sake of simplicity of explanation, post-processing etc. is not performed).

12 and 13 are flowcharts showing Japanese-English translation processing of a computer program included in the translation system according to the fifth embodiment of the present invention.

The difference between this flowchart and the flowcharts of FIGS. 4 and 5 will be described.

In step S107, the selected word and its bilingual translation are registered in the first database (FIG. 10).

After the process in step S109, in step S110, a process for creating a second database (FIG. 11) is performed from the contents of the first database and the sentence to be translated. In step S111, a process of encrypting the word or the word + reference code is performed. A second database is used in this process.

In step S131 of FIG. 13, a process of converting a cipher into a word (translated word) or a cipher into a word (translated word) + reference code is performed using the second database (first decoding). ..

The first and second databases may operate with different types of programs. For example, the first database is a CSV file or an EXCEL file, and the second database uses software such as SQLite or MySQL.

In the present embodiment, the word + reference code is registered in the database together with the cipher as one unit, but the prefix word + word or the prefix word + word + reference code is registered in the database together with the cipher as one unit. May be. For example, if "Si" of "Si board 300" is used as a prefix word and there is a prefix word before the word "board", "Si board" can be registered as one unit together with encryption, or "Si" can be registered. The "board 300" is registered together with the code as one unit.

Further, after performing the full-text encryption processing of the translated original text using the database of the combination of the word and the cipher in FIG. 11 (numbers 0 to 3504), the cipher + reference code portion in the document is searched for the word + reference. A database (numbers 4012-) of a combination of code and cipher may be created. After that, the encryption is performed again by the database of the combination of the word + reference code and the encryption.

It is desirable to perform processing to prevent duplicate registration of words and bilingual words in the database. This is a process of preventing the same pair of words and bilingual words from being registered, and a process of preventing the same pair of words from being registered.

[Other]

The flowchart in this embodiment may be executed by a plurality of software. For example, some may be executed by word processing software (macro, etc.), and some may be executed by a scripting language or a compiled executable file. Some may be performed by humans.

In addition, a process of converting full-width characters to half-width, a process of deleting spaces, a process of reducing the number of spaces, and the like may be performed as appropriate (either of the processes).

In the above-described embodiment, the word of the first language is replaced with the word of the second language before the encryption is performed (steps S105, S107, etc.), but this is not performed and the database is used. A word in the first language may be encrypted directly and sent to an external computer resource.

All words (and / or words + reference codes) in the document may be replaced with ciphers, or only some of them may be replaced.

[Sixth Embodiment]

Since the hardware configuration of the translation system according to the sixth embodiment of the present invention is the same as that in the first to fifth embodiments, the description thereof will not be repeated. In the sixth embodiment, the translation system compares the original text of the text translated by the method (or other method) according to the above-described embodiment with the translated text, and evaluates the translation result.

The evaluation of translation is based on the determination that the information existing in the original text is left in the translated text and / or the information not present in the original text is not present in the translated text. The former judgment determines whether or not the information in the original text is missing, and the latter judgment determines whether or not unnecessary information is added to the translated text (whether or not there is a so-called "information gushing"). It is a judgment.

For example, the first encrypted text illustrated in the fifth embodiment is

"[0023]

The "HSD" has "HSE" and "HEF", and the "HSE" is connected to the "HEG" and "HEF". If the "HEG" input is high, "HEF" will display a warning. "

And when this sentence is translated by an external computer resource, the result is:

"[0023].

"HSD" includes "HSE" and "HEF". "HSE" is connected to "HEG" and "HEF". When an input of "HEG" is high, "HEF" displays warning. "

Here, the first encrypted text contains one "HSD", two "HSE", three "HEF", and two "HEG".

The translated text also contains one "HSD", two "HSE", three "HEF" and two "HEG".

This indicates that the information of all the words before translation remains in the sentence after translation without being lost (the word loss rate is determined to be 0%). It also indicates that the translated text does not contain words that were not in the pre-translated text (the word spout rate is determined to be 0%).

When a word is missing, for example, "One" HSE "is missing.", "Both" HSE "are missing.", "There are two" HSE "in the original text. Present a message to the user (displayed on the display) such as "One is missing.", "The" HSD "in the original text is missing.", "The missing rate of the word in this sentence is XX%." , Voice guidance, etc.) Further, when there is no defect, a message such as "No defect" or "0% defect rate" may be presented to the user.

The above warning message may be displayed only when there is a defect, and the message may not be displayed when there is no defect.

When there is a word sprout (unnecessary information is added to the translated text), for example, "" HSE "is increased by one.", "" HSE "is increased by two.", "" HSD "that is not in the original text has been added.", "There is one" HSE "in the original text, but there are two in the translated text.", "The rate of words in this sentence is XX. You can present a message such as "%." To the user. Further, if there is no defect, a message such as "There is no springing out" or "Spouting rate 0%" may be presented to the user.

It is not necessary to display the above warning message only when there is a spring, and not to display the message when there is no spring.

The warning message may be inserted in the translated text so that the warning is for which sentence (for example, it may be inserted before the beginning of the sentence or after the end of the sentence). The text to be warned and the content of the warning may be clearly displayed to the user by underlining, shading, adding a comment to the text, or the like.

Also, instead of presenting with a code such as "HSE", the corresponding word "electric vehicle 100" (or "electric vehicle" without a reference code) may be read from the database and presented. , "Electric vehicle 100" (or "electric vehicle" not including a reference code) may be read from the database and presented.

"The translation does not include" electric vehicle 100, "which is the translation of the original" electric vehicle 100. " ], Etc., a message containing both pre-translation and post-translation words may be presented to the user as a warning.

Further, the unencrypted "electric vehicle 100 includes an engine 101 and a display unit 102, and the engine 101 is connected to the signal input means 103 and the display unit 102. When the input of the signal input means 103 is high. When the sentence "electric vehicle 100 includes engine 101 and display 102. engine 101 is connected to signal input unit 103 and display 102. When an input of signal input unit." Assuming that the translated text of "103 is high, display 102 displays warning." Is obtained, the translated word ("electric vehicle") corresponding to the untranslated word ("electric vehicle 100", or "electric vehicle", etc.) Translation evaluation may be performed by determining whether there are the same number of "100" or "electric vehicle") or different ones.

For such a judgment, it is necessary to determine the word after translation from the word before translation (or conversely, it is necessary to determine the word before translation from the word after translation). This can be done by determining the corresponding word using a database such as 10. Alternatively, the translation may be determined from the word by using machine translation or another dictionary.

If you want to find out if a code such as "HSE" is included in the pre-translation text and the post-translation text, you can do it without using a database. That is, the number of occurrences of each word (each character string) can be checked by finding three uppercase letters of the alphabet with a regular expression and counting the number of occurrences for each cipher.

In this way, the result of translating a sentence written in the first language into a sentence written in the second language can be evaluated. That is, the sentence written in the first language contains a plurality of character strings. It is determined whether or not the character string corresponding to the character string included in the sentence described in the first language is included in the sentence described in the second language, and the determination result is displayed.

For example, in Japanese, the distinction between the singular and plural forms of nouns is not clear, and in English, the distinction between the singular and plural forms of nouns is clear. Therefore, when the correct translation is performed, the singular and plural nouns may differ between the original translation and the translation. Due to these singular and plural differences, the word (eg "apparatuses") corresponding to the word (eg "device") contained in the sentence written in the first language is correctly written in the second language. When the word corresponding to "device" is registered as "apparatus" in the database, it is judged that "apparatus" and "apparatuses" do not match (there is a missing word). It is not preferable.

So, for example, a plural word is converted to a singular form (or singular word is converted to a plural form), a singular form word and a plural form word are treated as the same, and so on. Matching (determination of existence in both) between the word contained in the previous sentence and the word contained in the translated sentence may be performed.

Similarly, when dealing with a language in which the verb changes depending on the person of the subject, whether the word of the verb contained in the sentence described in the first language is included in the sentence described in the second language. When making a judgment, for example, the verb is converted to the original form, the verb is treated as the same even if the word form changes, etc., so that the word contained in the sentence before translation and the word contained in the sentence after translation are matched. May be done.

As mentioned above, in the current machine translation, some information in the original text may be lost (missing) after translation, or information not found in the original text may appear in the translated text (information springing out). It takes a lot of effort to check such information deficiencies and springs with the human eye (it is necessary to check both the original text and the machine translated text).

When evaluating a machine-translated sentence, the units that make up the document (sentence, clause, phrase, paragraph, line, page, multiple sentences, multiple clauses, multiple phrases, multiple paragraphs, multiple lines, multiple pages, or When the above judgment is made for each unit (such as the entire document) and the judgment result is presented for each unit that constitutes the document, the user can omit translation (loss of information) or unnecessary information for that unit. You can see if there is an addition of, and if so, which word. This reduces the effort of the user to check the translated text.

That is, since it is possible to present the user with missing words or additions in association with the translation result of one unit, the user can easily determine whether or not the one unit should be checked in detail. Can be done. It calls attention to the user only for the unit part where there is omission of translation (loss of information) or addition of unnecessary information (the user checks the unit part in detail and refines the translated text. It can be carried out). For units for which no warning has been issued, the labor of checking and elaboration by the user will be reduced.

Further, when evaluating the translated text, the judgment may be made for the entire document and the judgment result for the entire document may be presented. As a result, it is possible to determine whether the translation of the entire document is missing (loss of information) or addition of unnecessary information (evaluation of the translated text). This determination result can be used as an evaluation of the quality of translation. For example, it is possible to evaluate machine translation, machine translation engine, and manual translation.

For example, the process for evaluating each sentence (the unit of a sentence from the beginning to the time when it is separated by a punctuation mark (or a period, etc.)) can be performed as follows.

(1) The translation result of one sentence transmitted in step S125 of FIG. 5 and the translation result of one sentence received in step S127 are compared and evaluated, and if there is any omission of translation (loss of information) or addition of unnecessary information, Information (character string, its number) about the missing character string (encryption, word, etc.) and the added character string (encryption, word, etc.) is presented to the user.

(2) Prepare the original document that has been completely translated and the document of the translated sentence, and divide both from the beginning by one sentence (from the beginning of one sentence to a punctuation mark (or period, etc.)). ) By reading, comparing and evaluating both, and if there is a translation omission (loss of information) or unnecessary information added, the information about the character string (character string, its number) is presented to the user.

In the above embodiment, the translation may be performed using an external server, or may be performed in a stand-alone PC or a PC in the LAN.

It is also possible to combine or replace the plurality of embodiments described above and the elements contained therein (partial configuration, partial processing) to obtain another new embodiment.

The processing in the above-described embodiment may be performed by software or by using a hardware circuit. It is also possible to provide a program that executes the processing according to the above-described embodiment, and record the program on a recording medium such as a CD-ROM, a flexible disk, a hard disk, a ROM, a RAM, or a memory card and provide the program to the user. You may decide to do it. The program is executed by a computer such as a CPU. Further, the program may be downloaded to the device via a communication line such as the Internet.

It should be considered that the above embodiments are exemplary in all respects and not restrictive. The scope of the present invention is shown by the scope of claims rather than the above description, and is intended to include all modifications within the meaning and scope equivalent to the scope of claims.

100 computer 101 CPU
103 Communication unit 109 RAM
111 Storage device 200 Computer resources of other companies 203 Search / replace unit 205 Document editing / word registration unit 207 Database 209 Memory 211 Display / output unit 213 Input unit 400 Internet 500 Own computer resources

Claims (10)

It is a translation evaluation device that evaluates the result of translating a sentence written in a first language into a sentence written in a second language.
The sentence written in the first language contains a plurality of character strings and contains a plurality of character strings.
A determination means for determining whether or not the character string corresponding to the character string included in the first language is included in the sentence described in the second language.
A translation evaluation device including a presentation means for presenting a determination result of the determination means. The translation evaluation device according to claim 1, wherein the sentence described in the second language is a sentence obtained by machine-translating the sentence described in the first language. The character string is either a character string indicating a word, a character string indicating a word and a reference code immediately following the word, or a coded character string, or a character string indicating a word, a word and a reference code immediately following the word. The translation evaluation device according to claim 1 or 2, which is a character string including either a character string indicating the above and a coded character string. A machine translation result of a sentence in which a character string indicating a certain word or a character string including a character string indicating a certain word in a sentence written in the first language is replaced with another corresponding character string. Get a sentence written in the second language
The translation evaluation device according to any one of claims 1 to 3, wherein the determination means determines whether or not the replaced character string is included in a sentence described in the second language. The determination means makes a determination for each unit constituting the document.
The translation evaluation device according to any one of claims 1 to 4, wherein the presentation means presents a determination result for each unit constituting the document. The determination means determines the entire document and determines.
The translation evaluation device according to any one of claims 1 to 5, wherein the presentation means presents a determination result for the entire document. The determination result includes at least one of a character string not included in the sentence written in the second language and a number of character strings not included in the sentence written in the second language. The translation evaluation device according to any one of claims 1 to 6. The determination means according to any one of claims 1 to 7, wherein the determination means determines whether or not the character string corresponding to the character string included in the second language is included in the sentence described in the first language. The described translation evaluation device. In the control program of the translation evaluation device that evaluates the result of translating a sentence written in the first language into a sentence written in the second language.
The translation evaluation device includes a computer.
The sentence written in the first language contains a plurality of character strings and contains a plurality of character strings.
The control program of the translation evaluation device is
A determination step for determining whether or not the character string corresponding to the character string included in the first language is included in the sentence described in the second language, and
A control program for a translation device that causes a computer to execute a presentation step that presents a determination result of the determination step. In a translation evaluation method using a translation evaluation device that evaluates the result of translating a sentence written in a first language into a sentence written in a second language.
The translation evaluation device includes a computer.
The sentence written in the first language contains a plurality of character strings and contains a plurality of character strings.
The translation evaluation method is
A determination step for determining whether or not the character string corresponding to the character string included in the first language is included in the sentence described in the second language, and
A translation evaluation method using a translation evaluation device, which includes a presentation step for presenting a determination result of the determination step.

Images