JP2019045953A - Synonym processing apparatus and program - Google Patents

Abstract

To provide a synonym processing apparatus and a program that enable search processing of synonyms relatively exhaustively even for terms which might be omitted in a technical-term-like process.SOLUTION: A synonym processing apparatus accepts input of a target character string of a first language to be processed, translates the target character string into words of a second language different from the first language, refers to a phrase table in which the word of the second language is associated with at least one of words of an output language different from the second language corresponding to the word of the second language, obtains a list of the words of the output language corresponding to the words of the second language obtained by translation by translation means, and outputs the obtained list of the words of the output language as synonyms of the words represented by the target character string.SELECTED DRAWING: Figure 5

Description

  The present invention relates to a synonym processing device and program.

  Patent Document 1 discloses a technique for searching for synonyms based on the frequency of appearance of the modified word text of each document data tabulated for each evaluation expression and for each category.

  In addition, the technical terms of kanji only are extracted from the Japanese document, and each of the extracted technical terms is converted into a simplified character string by replacing each of the kanji constituting the technical term with the corresponding simplified character. Each of the simplified character strings is collated with the Chinese document in the patent parallel translation document set, and if it exists in the Chinese document, the simplified character strings and the original Japanese technical terms are collected and translated. Patent Document 2 discloses a technique of outputting as a set.

JP, 2012-48291, A JP, 2014-164739, A

  However, in the technology disclosed in Patent Document 1, when there is no expression of evaluation, the appearance frequency can not be summed up, and the coverage of synonyms is lacking. Moreover, in the technology of Patent Document 2, although the coverage of synonyms is excellent, "Sakura" ("Hiragana" (for example, "Solder" etc.) not expressed in kanji or due to morphological analysis used in the present invention There is a problem that terms including "... feed" and the like in "paper feed" are leaked from processing.

  The present invention has been made in view of the above circumstances, and is a synonym processing device that enables search processing of synonyms even if there is a possibility of leakage in exhaustive and jargon processing. Providing a program is one of its purposes.

  The present invention for solving the problems of the above-mentioned conventional example is a synonym processing device, which is an accepting means for accepting an input of a target character string of a first language to be processed, a target character string in the first language And at least one word of a second language different from the second language, and translation means for translating into a second language word different from the second language, and a word of the second language and a word of the second language corresponding to the second language. A list acquiring unit for acquiring a list of words of the output language corresponding to the words of the second language obtained by translating by the translation unit with reference to the associated phrase table; and the words of the acquired output language And outputting means for outputting the list of words as synonyms of the word represented by the target character string.

However, although preparing a large amount of parallel corpus and preparing the trained phrase table is important for the completeness of synonyms, preparing a large amount of parallel corpus is a problem in itself.
Therefore, in the present invention, a phrase table using an English-Japanese corpus according to the most easily obtainable patent sentence is fixed, and translation means are combined to generate synonyms corresponding to multiple languages.

  According to the present invention, there is provided a synonym processing apparatus capable of searching for synonyms and presenting synonyms even with terms that may be leaked in a relatively exhaustive and jargon-like process. .

It is a block diagram showing the example of composition of the synonym processing device concerning an embodiment of the invention. It is an explanatory view showing an example of contents of a phrase table which a synonym processing device concerning an embodiment of the invention uses. It is a functional block diagram showing the example of the synonym processing apparatus which concerns on embodiment of this invention. It is a block diagram showing the other structural example of the synonym processing apparatus which concerns on embodiment of this invention. It is a flowchart figure showing the operation example of the synonym processing apparatus which concerns on embodiment of this invention. It is explanatory drawing showing the example of an output of the synonym presentation screen by the synonym processing apparatus which concerns on embodiment of this invention. It is a flowchart showing the operation example of the synonym processing apparatus which concerns on embodiment of this invention. It is a flowchart showing another operation example of the synonym processing device concerning an embodiment of the invention. It is a flowchart showing the further another operation example of the synonym processing apparatus based on embodiment of this invention.

  Embodiments of the present invention will be described with reference to the drawings. The synonym processing apparatus 1 according to the embodiment of the present invention includes the control unit 11, the storage unit 12, the operation unit 13, the display unit 14, and the network interface 15, as illustrated in FIG. Be done. The control unit 11 is a program control device such as a CPU, and operates in accordance with a program stored in the storage unit 12.

  The control unit 11 of the present embodiment accepts an input of a target character string to be processed. Here, it is assumed that the target character string is expressed in a first language, such as Japanese. Further, the control unit 11 translates the received target character string into words of a second language different from the first language. Here, the second language is, for example, English.

  The control unit 11 refers to a phrase table in which at least one word of the second language and at least one word of an output language different from the second language corresponding to the word of the second language are associated. The phrase table is stored, for example, in the storage unit 12 and, as illustrated in FIG. 2, a word A of a second language, a word B of a predetermined output language that can be selected as a translation of the second language A, and The word B of the output language is associated with the probability p selected as a translation of the word A of the second language.

  Such a phrase table is described, for example, in Uzero et al., "Estimated Term Translation using the Japanese-Chinese Patent Family," JAPIO YEAR BOOK 2014, 236-241, etc. A phrase table generated using (Moses: http://www.statmt.org/moses/) can be used.

  Further, the control unit 11 refers to this phrase table to obtain a list of words of the output language corresponding to the words of the second language obtained by translation. Then, the control unit 11 outputs the list of words of the output language acquired here as a synonym of the word represented by the target character string. Here, the output language may be the same as the first language, or may be a third language different from the first language (also different from the second language). The more detailed contents of the operation of the control unit 11 will be described later.

  The storage unit 12 is a memory device or the like, and holds a program executed by the control unit 11. This program may be stored in a computer readable non-transitory recording medium and provided, and may be stored in the storage unit 12. The storage unit 12 also operates as a work memory of the control unit 11. Furthermore, in the present embodiment, the storage unit 12 may store the phrase table described above.

  The operation unit 13 is a keyboard or the like, receives an operation of the user, and outputs information representing the content of the operation to the control unit 11. Specifically, the control unit 11 accepts an input of a word of the first language from the user and outputs the input to the control unit 11. The display unit 14 is a display or the like, and displays and outputs information in accordance with an instruction input from the control unit 11.

  The network interface 15 receives the information transmitted to the synonym processing apparatus 1 via the network (such as the Internet), and outputs the information to the control unit 11. Further, the network interface 15 transmits information to a predetermined server on the network, another PC, or the like in accordance with an instruction input from the control unit 11.

  Next, the operation of the control unit 11 of the present embodiment will be described. Functionally, the control unit 11 of the present embodiment is configured to include the receiving unit 21, the translation processing unit 22, the list acquisition unit 23, and the output processing unit 24 as illustrated in FIG. 3. Ru.

  The receiving unit 21 receives an input of a target character string of a first language to be processed from the user, and outputs the character string to the translation processing unit 22. In an example of the present embodiment, the receiving unit 21 receives an input of a target character string expressed in Japanese as the first language.

  The translation processing unit 22 translates the target character string output by the receiving unit 21 into words of a second language different from the first language. In an example of the present embodiment, the translation processing unit 22 translates a target character string expressed in Japanese as a first language into English words as a second language. In an example of the present embodiment, this translation processing unit 22 performs a machine translation process using statistical machine translation or a neural net machine translation engine.

  The translation processing unit 22 does not necessarily perform machine translation processing by itself. For example, the target character string output from the receiving unit 21 that has been the target of translation is May be sent to a server that performs Google translation (Google translation processing server), and the translation processing may be requested from the Google translation processing server. In this case, the translation processing unit 22 receives, from the Google translation processing server, a response of the translation result of the target character string, and outputs the translation result included in the received response to the list acquisition unit 23.

  The list acquisition unit 23 refers to, for example, the phrase table stored in the storage unit 12 and lists the words of the output language corresponding to the words of the second language obtained by the translation processing unit 22 by translating. get. Specifically, the list acquisition unit 23 refers to the phrase table, and a list of output language words associated with the phrase including the second language word obtained by the translation processing unit 22 by translation, Obtain as a list of output language words that correspond to the second language words.

  Specifically, when the word of the second language obtained by translating in the translation processing unit 22 is “automatic document feed”, the list obtaining unit 23 partially matches the word “automatic document feed”. Search from the phrase table. As an example, as such a phrase, a phrase such as an automatic document feeder, an automatic document feeding device, an original automatic document feeder or the like will be found by a search. Then, the list acquisition unit 23 acquires a list of words of the output language associated with each of the plurality of phrases found by the search. The output language may be Japanese, for example, the same as the first language.

  The output processing unit 24 outputs the list of words in the output language acquired by the list acquisition unit 23 as a synonym of the word represented by the target character string. In an example of the present embodiment, the output processing unit 24 performs predetermined expression unification processing on the words included in the acquired list, and a word representing a list in which the target word removes the duplicate words after the expression unification processing. It may be output as a synonym of An example of this expression unification process will be described later.

  The present embodiment has the above configuration and operates as follows. In the following example, in forming the phrase table, the second language is English and the output language is Japanese. Specifically, in order to form a phrase table in which the second language is English and the output language is Japanese, it is disclosed in published patent publications of Japanese patent publications, published patent publications and US patent applications that are a family thereof. A parallel extracted from a set of translated texts prepared in paragraph units of the invention title, abstract, claims and detailed description from the English publication, the English publication related to the European patent application, or the English publication related to the international patent application The corpus was used as a machine learning source. More specifically, of the bilingual sentences of Japanese and English as the machine learning source, Japanese sentences are subjected to morphological analysis to be decomposed into words, and further, words consisting only of "Hiragana" are subjected to pattern matching. Excluded and extracted the words. Then, based on the phrase table created by Moses (Moses: http://www.statmt.org/moses/), which is a translation model toolkit for statistical machine translation, using bilingual sentences in Japanese and English. I decided to get a phrase table.

  In Moses, the phrase table can be generated in two modes: phrase-based phrase table and tree-based phrase table. If the same entry exists in the phrase table obtained in each mode, the entry common to both phrase tables is extracted and the phrase table is constructed, assuming that the parallel translation related to the entry has a relatively high probability of being correct. Do. In this way, it is possible to extract a more accurate (relatively correct) word relationship than a phrase table based on a phrase table created in a single mode. In this example of the present embodiment, this phrase table is stored in the storage unit 12.

  Then, first, the control unit 11 functions as the receiving unit 21 and receives the input of the target character string of the first language to be processed from the user. Here, as an example, it is assumed that the word "automatic document feeding" expressed in Japanese as the first language is input.

  The control unit 11 translates the received target character string into a word of a second language different from the first language as processing of the translation processing unit 22. Here, the translation process may use, as a machine translation engine, the Google translation API of Google Inc., which is a machine translation of neural network method. That is, the control unit 11 requests the Google translation processing server to process the translation target word as a target character string using the Google translation API as a target character string. And the control part 11 receives the response of the translation result of the said object character string from the server of a Google translation process.

  The control unit 11 refers to the phrase table based on the translation result received here. As an example, when the target character string is "automatic document feeding" as described above, the English translation "automatic document feed" as the corresponding second language is transmitted from the Google translation processing server.

The control unit 11 refers to the phrase table stored in the storage unit 12 as the process of the list acquisition unit 23, and the output language corresponding to the word "automatic document feed" in the second language acquired here. Get a list of words. In addition, the control part 11 searches the word which partially corresponds to the word "automatic document feed" of the acquired 2nd language, in referring a phrase table. Referring to the phrase table generated by the method described above, from the phrase table as a word partially matching the word "automatic document feed",
automatic document feeder
automatic document feeding device
original automatic document feeder
automatic document feeding portion
automatic document feeder serving
automatic document feeders
reversing automatic document feeder
Seven entries of are found.

Therefore, the control unit 11 acquires a list of Japanese words that are output languages, which are recorded in the phrase table in association with each of the seven entries. Here, the control unit 11
Automatic Document Feeder Automatic Document Feeder
Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic document feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Service Although a list is acquired like automatic document feeder reverse, the control unit 11 performs predetermined expression unification processing on these. In particular,
・ Unification of long sound symbols (if the end of a katakana word of 3 or more characters ends with a long sound, delete the long sound, etc.),
・ Unification of full-width half-width (replace half-width characters with corresponding full-width characters),
・ Unification of upper case and lower case notation of English letters (for example, all English letters are in lower case),
・ Unify multiple expressions into singular expressions (filters in English are filters),
・ Unify the notation for ending probation (processing such as replacing “It is a device” with “a device, etc., deleting verbs etc. at the end of a phrase, etc.),
・ Punctuation mark deletion (Punctuation mark in the phrase is deleted)
・ Delete the top unnecessary part such as "Said", "Said", etc.
・ Delete the unnecessary part of the end such as "Et,""Every",
And so on. These processes are exemplification, and a widely known expression normalization process may be employed. Also, these do not need to be all processed, and some of them may be selectively performed.

In this example of the present embodiment, the control unit 11 applies all the above-described processing to the acquired character string of the list, and excludes those that are identical among the phrases after application (sorting , Can be performed by a process of comparing adjacent phrases) and post-processing is performed to exclude candidates for the phrase "original document feeder" whose head part is an expression including a particle, and as a result,
Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Service Reverse List of automatic document feeders.

  The control unit 11 outputs the list to the display unit 14 to display the list, and presents it to the user as a synonym of "automatic document feeding" input.

  Also, here, when the control unit 11 acquires a list of Japanese words that are the output language from the phrase table, the translated words respectively associated with the Japanese words corresponding to the original English words in the phrase table It is also possible to refer to the probability p selected as and to include in the list of Japanese words that are the output language when the probability p exceeds a predetermined threshold.

  Furthermore, although the phrase table is stored in the storage unit 12 in the above description, the present embodiment is not limited to this, and the phrase table is stored in a server that can communicate via a network. It may be In this case, the control unit 11 transmits the word of the second language obtained by the translation process to the server via the network interface 15 as the process of the list acquisition unit 23 and outputs the word of the output language. You may request a list. In this example, the control unit 11 acquires, from the server via the network interface 15, a list of words of the output language corresponding to the words of the second language as the translation result.

[Configuration to service via network]
Here, although the input of the target character string of the first language to be processed is accepted from the operation unit 13 and output to the display unit 14, the present embodiment is not limited to this example, and via the network The synonym information may be presented via the network in a manner such as accepting input of a target character string of the first language to be processed, providing a web page, or the like.

  The synonym processing apparatus 1 'according to this example of the present embodiment includes at least a control unit 11, a storage unit 12, and a network interface 15, as illustrated in FIG. The same reference numerals are given to those performing the same operations as those described above. The synonym processing apparatus 1 'is connected to the user's PC 2 and a server apparatus 3 that provides various information services via a network (the Internet or the like).

  The network interface 15 of the synonym processing device 1 ′ according to this example receives an input of a target character string of the first language to be processed from the PC 2 of the user and outputs the input to the control unit 11. Further, the network interface 15 outputs various information such as information on synonyms to the PC 2 of the user according to the instruction input from the control unit 11.

  Further, the control unit 11 of the synonym processing device 1 ′ of this example accepts the input of the target character string of the first language to be processed from the network interface 15, instead of the operation unit 13. Further, instead of the display unit 14, various information such as information on synonyms is transmitted to the PC 2 of the user via the network interface 15. The processing of the other control unit 11 is the same as that of the example already described, and thus the description thereof will not be repeated.

[Example of outputting synonyms in a language different from the first language]
In the above example, the first language is Japanese and the second language is English, and Japanese synonyms are output as synonyms, but the present embodiment is not limited to this. . For example, in the processing as the output processing unit 24, the control unit 11 further calls the list of words of the output language acquired by the list acquisition unit 23 as another language (hereinafter referred to as a synonym presentation language. This synonym presentation language is For example, it may be translated into a second language or another word different from the first and second languages), and the result of the translation may be output as a synonym of the word represented by the target character string. . Also in this example, a predetermined expression unification process is performed on the words included in the acquired list, and the list excluding the overlapping words after the expression unification process is translated into the above synonym presentation language, and the result of the translation is It may be output as a synonym of the word represented by the target character string.

  Further, at this time, when the synonym presentation language is the second language (the language included in the phrase table), as described above, in the present embodiment, when referring to the phrase table, the translation result from the first language In order to search the phrase table including the phrase from the phrase table, a list of words (synonyms) corresponding to the target character string in the second language which is a synonym presentation language is obtained at this time as well.

  Therefore, in an example of the present embodiment, a word corresponding to a target character string in a second language which is a synonym presentation language obtained by searching a phrase table including a translation result from a first language from a phrase table The list of words and the list of words in the second language which is a synonym presentation language obtained by translating each word included in the list in the output language in the processing as the output processing unit 24 as described above After these are connected, predetermined expression unification processing may be performed, and a list may be presented in which duplicate words have been removed after the expression unification processing.

According to this example, when the word “automatic document feeding” described above becomes a processing target character string, the control unit 11 translates the processing target character string into a second language, and “automatic document feed”. Get Then, the control unit 11 searches the phrase table for a phrase that partially matches this "automatic document feed", and the list A:
automatic document feeder
automatic document feeding device
original automatic document feeder
automatic document feeding portion
automatic document feeder serving
automatic document feeders
reversing automatic document feeder
Get

The control unit 11 is a word in an output language (here, Japanese) associated with each of the phrases contained in the list A, and the probability p of being selected as a translated word on the phrase table is previously determined. Obtain a list of words that have a probability greater than or equal to the defined threshold, and from this, apply the processing for uniform expression and processing for duplicate elimination from this list:
Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Reverse Automatic Document Feeder Automatic Document Feeder Automatic Document Feeder Service is obtained.

The control unit 11 further translates these (for example, by a Google translation server etc.) into a synonym presentation language (here, the second language) different from the output language, and the list B:
Automatic document feeder
Automatic document feeder
Automatic document feeder
Auto Document Feeder
Automatic Document Feeder
Automatic document feeder
Automatic document feeder
Automatic Document Feeder
Invert automatic document feeder
Automatic document feeder service
Get

The control unit 11 concatenates the list A previously obtained by searching from the phrase table and the list B obtained here, and performs predetermined expression unification processing (converts plural nouns into singular, upper case is A list to be output after eliminating duplicates after doing lowercase letters etc.):
auto document feeder
automatic document feeder
automatic document feeding device
automatic document feeding portion
automatic document feeder service
And present this list to the user.

  According to this example, it is possible to obtain a list of synonym expressions corresponding to the processing target character string in a predetermined (for example, designated by the user) synonym presentation language.

[Output language of phrase table]
Furthermore, the output language of the phrase table may not match the first language. For example, in the above example, the first language may be Chinese, the second language may be English, the output language of the phrase table, and the synonym presentation language may be Japanese. Also in this case, the synonym processor 1 according to the present embodiment accepts, for example, a word meaning glass transition temperature as a Chinese processing target character string, and uses a Google translation server etc. for the second language English. Translate. As a result, the synonym processor 1 obtains an English word, for example, "glass transition temperature".

Then, the synonym processing device 1 refers to the phrase table and, as a list of Japanese words corresponding to this word,
Glass transition temperature Glass transition temperature Glass transition temperature Temperature glass temperature A list of glass transition temperatures is obtained and presented to the user.

  According to this example, a synonym corresponding to the processing target character string input in Chinese can be obtained in Japanese via the second language, English.

  Furthermore, here, the output language of the phrase table may be different from the synonym presentation language, and the output language of the phrase table may be Japanese, the synonym presentation language may be German, or the like. In this case, each of the words included in the list of words corresponding to the processing target character string in the output language obtained by referring to the phrase table as described above (and further through expression unification processing etc.) Translate to a synonym presentation language and output using.

[Flow of processing]
As described above, in the present embodiment, the synonym processing device 1 accepts the input of the target character string of the first language to be processed as illustrated in FIG. 5 (S1), and the received target character string Are translated into words of a second language different from the first language (S2).

  In this process S2, a machine translation process may be performed, but when performing this machine translation process, a process based on a technical term, for example, when the word at the end is determined to be a verb by morphological analysis, the end It controls not to execute the process of excluding the word of and the process of removing the character string consisting only of "Hiragana". That is, a machine translation engine controlled so as not to execute these processes is selected and used.

  Thus, for example, "sending" of "automatic document feeding" is interpreted as a verb, so that it is converted into "automatic document" and a synonym search process is performed, or from a word such as "solder riding", "soldering" The part is removed as a hiragana-only word, and the synonym search processing based only on "ride" is not performed.

  The synonym processing device 1 refers to a phrase table in which at least one word of the second language thus obtained and an output language different from the second language corresponding to the word of the second language are associated. A list of words of the output language corresponding to the words of the second language obtained by translating in process S2 is acquired (S3).

  Then, the synonym processing device 1 subjects the list to a uniform expression process or a post-processing to eliminate duplication (S4), and outputs the post-processed word list as a synonym of the word represented by the target character string ((4) S5).

  When outputting a synonym as a web service as in the example already described, as illustrated in FIG. 6, a column A for inputting a processing target character string and a column B for outputting a list of synonyms are arranged side by side. It is also possible to distribute, create a web page for displaying the list obtained in the processing S4 in this column B, and transmit it to the user's PC2.

  In this case, the user may be able to specify the language (that is, the first language) of the processing object character string input to the column A by the user (pull-down menu X in FIG. 6). In this case, the synonym processor 1 sets the first language to be input as the word designated by the user when performing the machine translation process of the process S2. For example, when using a Google translation server as a machine translation process, the first language specified by the user is set as the translation source language.

[Input of other languages]
Further, in the present embodiment, in the processing as the translation processing unit 22, the control unit 11 is different from the first language to be an object of the machine translation (input language) (also different from the second language after translation) It is also possible to receive the input of the target character string in the third language. In this case, the control unit 11 translates the word of the target character string into the word of the first language to be processed as the processing of the receiving unit 21 and the translation result is the first language to be processed. Accept as input for target string of. In this case, the receiving unit 21 may perform machine translation processing different from the machine translation used by the translation processing unit 22 or may use the same machine translation processing as the machine translation used by the translation processing unit 22 (input and output) Different languages will be translated into second language words in two steps (or more).

  Hereinafter, as an example, an example of a synonym search result using the synonym processing device 1 according to the above-described embodiment of the present invention will be described. First, as a first embodiment, an example in which the first language is Chinese, the output language, and the synonym presentation language are Japanese will be described.

  As exemplified in FIG. 7, it is assumed that Chinese representing "glass transition temperature" is input as a processing target word (S11). The synonym processor 1 translates this into English as a second language to obtain "glass transition temperature" (S12). Furthermore, the synonym processing device 1 includes a word A of a first language, a word B of an output language that can be selected as a translation of the first language A, and a word B of the output language of a first language The output language word B associated with the word partially matching “glass transition temperature” as the language A word with reference to the phrase table associated with the probability p selected as a translation of A list of words B in the output language is obtained which has a probability that the word B is selected as a translation of the word A in the first language exceeds a predetermined threshold (S13).

  Next, the synonym processor 1 performs a predetermined expression unification process and a duplicate word deletion process on the words included in the list, and a list obtained by the synonym process corresponding to the processing target character string input in step S11 (S14).

  Next, as a second embodiment, an example in which the first language is Chinese, the output language is Japanese, and the synonym presentation language is Chinese will be described. In this case, after the synonym processing device 1 obtains the Japanese synonym list corresponding to the processing target character string input in processing S11 by the same processing as illustrated in FIG. 7 (the list presented in S14 is obtained After that, as illustrated in FIG. 8, the Japanese words in the list are subjected to machine translation processing into Chinese words (S15). Then, the synonym processing device 1 presents, as a synonym corresponding to the processing target character string, a list obtained by performing predetermined expression unification processing and deletion processing of duplicate words on the words included in the list (S16).

  Further, as a third embodiment, an example in which the first language is German, the second language is English, the output language is Japanese, and the synonym presentation language is Chinese will be described.

  In this case, it is assumed that the German word representing "glass transition temperature" is input as the processing target word (S21), and the synonym processing device 1 translates this into English as a second language. "transition temperature" is obtained (S22). Furthermore, the synonym processing device 1 includes a word A of a first language, a word B of an output language that can be selected as a translation of the first language A, and a word B of the output language of a first language The output language word B associated with the word partially matching “glass transition temperature” as the language A word with reference to the phrase table associated with the probability p selected as a translation of A list of words B in the output language is obtained which has a probability that the word B is selected as a translation of the word A in the first language exceeds a predetermined threshold (S23).

  Next, the synonym processing device 1 obtains a list obtained by performing predetermined expression unification processing and duplicate word deletion processing on the words included in the list (S24). Then, the synonym processing device 1 machine-translates the Japanese words in the list into Chinese words (S25), and further performs predetermined expression unification processing on the words included in the word list after the machine translation processing, The list of words obtained by performing the duplicate word deletion process is presented as a synonym corresponding to the processing target character string (S26).

[Effect of the embodiment]
According to an embodiment of the present invention, combining a single language pair phrase table (phrase table database) with a large number of patent equivalent translations (typically more than 100 million sentence pairs) and a machine translation engine It is possible to comprehensively find Japanese synonyms from Japanese technical terms that do not depend on character types consisting of hiragana, katakana and compound words. Also, if it is a language that can be translated into a second language, for example, the first language is Vietnamese, and from Vietnamese technical terms to Japanese synonyms or from Japanese technical terms English, Chinese, Korean It is possible to obtain synonyms corresponding to multiple languages, such as Thai and Vietnamese, from synonyms of the word, and furthermore, from German technical terms. That is, one or more synonyms of the language or a language different from the language corresponding to one technical term in any language can be presented.

  In addition, in the translation process to the second language, by controlling not to process in technical terms, when it contains terms or Hiragana only in hiragana that may be leaked in the technical terms processing Even in the case where (for example, “... Feed” in “paper feed” is included), a synonym search process is possible.

1, 1 'Synonym processor, 2 PC, 11 control unit, 12 storage unit, 13 operation unit, 14 display unit, 15 network interface, 21 reception unit, 22 translation processing unit, 23 list acquisition unit, 24 output processing unit.

Claims (8)

Accepting means for accepting input of a target character string of a first language to be processed;
Translation means for translating the target character string into words of a second language different from the first language;
Referring to a phrase table associated with at least one word of a second language and a word of an output language different from the second language corresponding to the word of the second language, the translation means translates List obtaining means for obtaining a list of words of the output language corresponding to the obtained second language words;
An output unit that outputs the acquired list of output language words as synonyms of the words represented by the target character string;
Synonym processing device with. The synonym processor according to claim 1, wherein
The output unit performs a predetermined expression unification process on the words included in the acquired list, and outputs a list in which duplicate words are excluded after the expression unification process as a synonym of a word represented by the target character string Processing unit. The synonym processing device according to claim 1 or 2, wherein
The phrase table further associates and holds probability information selected as a bilingual term for each of the output language words corresponding to the second language words,
The list acquisition unit acquires a list of words of the acquired output language with reference to the probability information. The synonym processing apparatus according to any one of claims 1 to 3, wherein
The synonym processing device wherein the output language is the first language. The synonym processing apparatus according to any one of claims 1 to 3, wherein
The synonym processing device, wherein the output language is a third language different from the first language. The synonym processing apparatus according to any one of claims 1 to 5, wherein
The synonym processing device further translates the word of the output language included in the acquired list into the corresponding word of another language different from the output language and outputs the word. The synonym processing apparatus according to any one of claims 1 to 6, wherein
The receiving means further receives an input of a target character string of a third language, translates the word of the target character string into a word of a first language to be processed, and the target translation result is to be processed. A synonym processor that accepts as input a target string of a first language. Computer,
Accepting means for accepting input of a target character string of a first language to be processed;
Translation means for translating the target character string into words of a second language different from the first language;
Referring to a phrase table associated with at least one word of a second language and a word of an output language different from the second language corresponding to the word of the second language, the translation means translates List obtaining means for obtaining a list of words of the output language corresponding to the obtained second language words;
An output unit that outputs the acquired list of output language words as synonyms of the words represented by the target character string;
A program to function as