JPH086950A - Machine translation apparatus with keyword translation function - Google Patents

Abstract

PURPOSE:To obtain a translated sentence for an object sentence by translating only sentences including previously registered keywords out of plural Japanese sentences. CONSTITUTION:This machine translation apparatus is provided with 'quick reading mode' for translating only sentences including specific keywords previously registered in a keyword table storing means 9 out of Japanese sentences inputted from an input means 2 and 'overall translation mode' for translating all input sentences in a certain unit, and when the 'quick reading mode' is selected by a translation mode specifying means 11, only sentences registered in the table 9 are translated by a translation module 5 and an operation main memory 6 as objects to be translated and a translated result is outputted by an output means 3, so that an object output can quickly be obtained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a machine translation device with a keyword translation function in the field of natural language processing.

[0002]

2. Description of the Related Art At present, most machine translation devices perform translation using a "sentence" as a single processing range. Such a machine translation device analyzes a structure inside one “sentence” and translates by constructing one independent structure from one sentence. In a general machine translation device, the following analysis processing is performed.

Dictionary Lookup / Morpheme Analysis Referring to the dictionary of the translation dictionary storage means 8 shown in FIG. 9, the input source language sentence is divided into morpheme strings (word strings), and the part of speech for each word is divided. Obtain grammatical information such as information and translated words, and further analyze the time, person, and number.

[Syntax analysis] The structure of a sentence (syntax analysis tree) such as dependency between words is determined by receiving the result of morphological analysis.

Among natural language processing devices used in conventional machine translation devices and the like, ones that perform full-text translation of all sentences contained in one or a plurality of original text documents designated as the target of the translation device, area designated range Some have a translation function such as partially translating only. However, in translating news, etc., proper nouns, dates and times, "who" and "when"
By translating “where”, “who”, etc., you can grasp the approximate content. However, the conventional machine translation device has a problem that a considerable amount of time is required to translate news and the like, and the user has to wait a long time until the translation end time. On the other hand, a machine translation device that supports "speed reading" (= quickly grasp the outline of a document) by quickly outputting a translation result by translating only a specific word or date and time registered in a table by a user. Exists (Japanese Patent Laid-Open No. 5-158971). The technique will be described in detail below with reference to FIG.

The configuration from 1 to 8 in FIG. 10 is the same as that of a general machine translation device. In addition to them, the user dictionary (table) storage means and either one of the "speed reading mode" for translating only limited words for speed reading or the "full translation mode" for translating all words is selected. When the "speed reading mode" is selected by the translation mode selection means 10 and the selection means, only the words stored in the table (FIG. 11) are translated, and the other words are translated as the original text, Further, with respect to the date, it has a date pattern table as shown in FIG. 12, and is provided with a speed-reading translation means 11 which judges whether or not the pattern matches, and translates the matched pattern. An example of translation is as follows.

[0007] Midway through Secretary of St
ate James A. Baker 3rd's tour of Europe and the Mid
dle East, it has become clear that Washington's pa
rtners in the coalition against iraq ・ ・ ・ ・ (Translation results for speed reading) Midway through Secretary of State James Baker tour of Europe and the Middle East it has become c
lear that Washington partners in the coalition agai
nst Iraq ・ ・ ・ ・ Translate only the words stored in the table like this,
It was possible to support speed reading by translating other words as the original text.

[0008]

In the "full-text translation" technique in the prior art, when a large amount of translation is performed, a huge amount of translation processing time is required, and it takes time and labor for the user to read the translation result. However, it was not suitable for "speed reading".

Further, in the "partial translation" technique, it is necessary for a person to specify a translation target area, and it is also unsuitable for "speed reading".

Therefore, there is the above-mentioned "speed reading technique", but the above technique translates only specific words and dates registered in the table. For translation of news, etc., if you have registered proper nouns etc. in the table, "who"
Since "when", "where" and "who" are translated, it was possible to know the outline of the sentence.

However, when translating a document or the like for introducing a new technology, the important part to be translated is "what", so the keyword that seems to be effective for "speed reading" in advance. Is almost impossible to register in the table, and it is difficult for such a translation apparatus to grasp the outline of the document.

Therefore, in the present invention, when the "speed reading mode" is selected, it is determined whether or not a specific word registered in the table in advance is in the sentence, and if it is detected, the sentence is translated. It is an object of the present invention to provide a translation device that

[0013]

According to the invention of claim 1, at least one keyword table storage means for storing a table in which a specific keyword is registered, and keyword table designation means for selectively designating at least one table, "Speed reading mode" for translating sentences containing keywords stored in the table and normal "translation mode"
A translation mode selecting means for selecting any one of them, and a speed-reading translation means for translating the entire sentence including the keyword registered in the table when the "speed-reading mode" is selected by the selecting means. To have.

Further, in the invention of claim 2, there is provided a variation form extracting means for extracting the variation form of the keyword from the translation dictionary by referring to the keyword and part-of-speech pair registered in the table, Further, in the invention of claim 3, with respect to the keyword registered in the table and the translated word of the variation of the keyword, the translated word corresponding to the keyword is retrieved from the table and output.

[0015]

According to the first aspect of the present invention, the keyword table storage means stores a speed-reading translation table in which a specific keyword is registered. The keyword table can be rewritten (customized) by the user. The translation mode selection means is a "speed reading mode" for translating the entire sentence including the keywords registered in the table.
Alternatively, the user can select any one of “all translation modes” for translating all sentences of a certain unit through the input / output device.

The speed-reading translation means makes it possible to translate the entire sentence including the keyword registered in the table when the speed-reading mode is selected. The keyword table designating means allows the user to designate one or a plurality of the tables through an input / output device according to a translation target field. According to the invention described in claim 2, the dictionary storage means can store a translation dictionary in which at least a word (or a word string), a variation and a translation corresponding to the word are registered as a set. . In addition to the above functions, the keyword table can also store a speed-reading translation table in which a specific keyword and a part of speech corresponding to the keyword are registered as a pair. The variation extracting means refers to the keyword-part-of-speech pair registered in the table and extracts the variation of the keyword from the translation dictionary. The speed-reading translation means makes it possible to translate the entire sentence including the keyword and the variation of the keyword registered in the table when the speed-reading mode is selected. According to the invention described in claim 3, in addition to the above-mentioned function, the speed-reading translation means, regarding the keyword registered in the table and the translated word of the keyword, converts the translated word corresponding to the keyword from the table. It is searched, and if necessary, it is output after changing the inflection of the translated word.

[0017]

The present invention will be described in more detail below with reference to the drawings. FIG. 1 is a block diagram of an embodiment of a machine translation device of the present invention. Reference numeral 1 denotes an operation control means, which is realized mainly by a CPU (= central processing unit) in an actual device configuration. 2 is an input means,
In an actual device configuration, a keyboard, a scanner, a character recognition device, a mouse, etc. correspond to this. Reference numeral 3 is an output means, and in an actual device configuration, a display device such as a CRT (= cathode ray tube), an LCD (= liquid crystal display device), a printing device such as a printer, or a communication device connected to a communication line, etc. Corresponds to this. Reference numeral 4 is a bus line, 5 is a translation module, 6 is a main memory for operation, 7 is various translation rule storage means required for translation processing, and 8 is translation dictionary storage means.

When the source language is input, the translation module translates the source language and outputs the target language (language to be translated). That is, the source language input from the input means 2 is sent to the translation module 5 under the control of the operation control means 1. The translation module 5 translates the input source language into a target language using the translation rules stored in the translation rule storage unit 7 and the dictionary stored in the translation dictionary storage unit 8. The translation result is temporarily stored in the operation main memory 6 and displayed on the output means 3.

The operation main memory 6 has a part of the original text buffer 6a, the dictionary lookup buffer 6b, the pre-conversion syntax buffer 6c, the post-conversion syntax buffer 6d, and the translation buffer 6e according to the progress of the translation operation. , Respectively.

First, a source sentence (English sentence) is read from the input means 1 and stored in the original sentence buffer 6a. Next, under the control of the control means 1 based on the translation module 5, the dictionary stored in the translation dictionary storage means 8 is used by the dictionary lookup / morpheme analysis means 5a to store the original text buffer 6a.
Information such as a translated word of each word is obtained from the source sentence stored in, and is stored in the dictionary lookup buffer 6b.

In the syntax analysis means 5b, according to the analysis grammar rules stored in the translation rule storage means 7 and the dictionary stored in the translation dictionary storage means 8, a syntax analysis tree showing a dependency relation between each word is created. Determined, syntax buffer 6
It is stored in c. This syntactic parse tree is created by applying rules from the parsing grammar rules stored in the translation rule storage means 7.

In the syntactic conversion rule 5c, similarly to the syntactic analysis unit 5b, the structure of the syntactic parse tree corresponding to the input source sentence is used by using the conversion grammar rule stored in the translation rule storage unit 7. Converted into a parse tree corresponding to a sentence (Japanese sentence). The result obtained by this tree structure conversion is stored in the syntax buffer 6d.

The target language generation means 5d adds appropriate particles and auxiliary verbs to the target sentence stored in the converted syntax buffer 6d and stores it in the translated sentence buffer 6e. This is output from the translation module 5 and temporarily stored in the operation main memory 6. Reference numeral 9 denotes a keyword table storage means, which can store one or a plurality of keyword tables. Reference numeral 10 is a translation mode designating means. The user specifies either the "translation mode" or the "speed reading mode" by selecting the mode with the input means 2. 11 is a keyword table designating means,
The user specifies one or a plurality of keyword tables by the input means 2. Reference numeral 12 is a keyword variation extraction means, and 13 is a speed reading translation means.

A feature of the present invention is that the speed-reading translation process using the "keyword table", that is, the "keyword table" is used to sort the sentences to be translated and those not to be translated.

[Embodiment of Claim 1] FIG. 2 is an example of a flowchart for realizing the speed-reading translation process according to Claim 1 of the present invention. Further, FIG. 3 is a selected keyword table used for the speed-reading translation process in the embodiment of claim 1 of the present invention. Now, in the "translation mode selection" of step 0 in FIG. 2, the "translation mode selection means 10" (FIG. 1) selects the "speed reading mode", and in step S1, the "keyword table designation means 11" (FIG. 1). It is assumed that the "keyword table" shown in FIG. 5 is specified by. In reality, it is possible to specify one or more keyword tables. First, FIG. 2 step S1
Is set to "sentence pointer = 0", and the sentence of "sentence number = 0" in the original document shown in FIG. 7 is pointed to. That is (Original 1) Intel Corp. and Unisys Corp. said Tuesda
y they would work together to build powerful comput
ers for business that run by linking the microproce
ssing chips found in most desktop machines.

Next, in step S2 of FIG. 2, "word pointer =
“0” is set, and “Intel” at “word position = 0” in the word string shown in FIG. 8 is pointed to. Subsequently, referring to the "keyword table" shown in FIG. 5 stored in the "keyword table storage means 9" (FIG. 1) in step S3 of FIG. 4,
Register the first word "Intel" in the original sentence.

As a result, it is determined in step S4 in FIG. 2 that there is no registration in the "keyword table".
In step S5, it is determined whether there is a "next word". If there is a next word, the process proceeds to step S6 in FIG. 2 and the word pointer is advanced by one, so that the word “Corp.” at “word position = 1” in FIG. 8 is pointed to. In this way, steps S3 to S6 in FIG. 2 are repeated, and the word “build” at “word position = 12” in FIG. 8 is pointed to. Then, step S in FIG.
Since it is determined in 4 that the word is registered in the "keyword table", the process proceeds to step S7 in FIG. 2 and the "speed reading translation process" is performed. That is, FIG.
The sentence of “sentence number = 0” in is regarded as the target of the translation process, and the translation process is performed. Therefore, the (original sentence 1) is translated into Japanese by the functions 5 to 8 in FIG.

Then, in step S8 of FIG. 2, it is determined whether or not there is a "next sentence". If there is a next sentence, the process proceeds to step S9 of FIG. 2 to advance the sentence pointer by one. Then, returning to step S2 in FIG. 2, "word pointer = 0" is set. Even when it is determined in step S5 of FIG. 2 that "there is no next word", the process directly proceeds to step S8 of FIG. After that, in step S8 of FIG. 2, steps S2 to S9 are repeated until it is determined that "the next sentence does not exist", and when the sentence "sentence number = 17" is finally processed, the speed-reading translation process is completed. Ends.

[Embodiment of Claim 2] FIG. 4 is an example of a flow chart for realizing the speed-reading translation process according to Claim 2 of the present invention. FIG. 5 shows an example of a keyword table used in the speed-reading translation process according to claim 2 of the present invention. The difference from the embodiment of claim 1 is as follows.

In step S0 of FIG. 4, the translation dictionary stored in the "translation dictionary storage means 8" (FIG. 1) is stored in the "keyword table storage means 9" (FIG. 1) shown in FIG. The variations of each keyword registered in the “keyword table” are extracted and stored in the “keyword variation buffer 6f” (FIG. 1).

For example, the word "abilit" at the beginning of FIG.
For “y”, there are multiple variations of “abilities”
For the verb "allow", the past tense "allowed" and the like are automatically extracted from the translation dictionary and stored in the keyword variation buffer 6f. In step S4 of FIG. 4, the first word “Intel” in the original sentence is searched with reference to the “keyword table” shown in FIG. Since it is determined in step S5 in FIG. 4 that there is no registration in the “keyword table”, the process proceeds to step S6 in FIG. 4 and each keyword variation stored in the “keyword variation buffer” and the word currently pointed to (here Then, "Intel" is determined). As a result, since it is determined that they do not match, the process proceeds to the next step. After referring to the keyword table,
If they do not match, the keyword variation buffer is referred to in step S6, and if they match, the process proceeds to step S9 to perform the translation process, which is the difference from the embodiment of claim 1.

[Embodiment of Claim 3] FIG. 6 shows an example of a keyword table used in the speed-reading translation process according to Claim 3 of the present invention. It differs from claims 1 and 2 in the following points. In the speed-reading translation process of claim 3, the following process is performed on the sentence to be translated.

Each keyword in the "keyword table" as shown in FIG. 6 stored in the "keyword table storage means 9" (FIG. 1) or stored in the "translation dictionary storage means 8" (FIG. 1) Each "translated word" registered in the "keyword table" extracted from the existing translation dictionary
Or "translated word part-of-speech / inflection" information in "dictionary lookup buffer 6b"
(FIG. 1), and the translated word is finally adopted as the translation result.

For example, taking the keyword table of FIG. 6 as an example, the word "feature" originally means 1 "feature of a face", 2 "features", 3 "features and characteristics", 4 "special features", etc. Although significant, the keyword table in Fig. 6 shows "fe
Since the translated word “feature” is registered for the English word heading “ature”, this “feature” is the translated word.
To adopt.

In the embodiments of claims 1 to 3 of the present invention, although there is no particular reference to the use of an index or the like for improving the efficiency of the search of the "keyword table", the keywords registered in the keyword table are not mentioned. Of course, it is also possible to sort with the key and use the corresponding index.

When comparing a keyword consisting of a plurality of words registered in the keyword table with a word in the original text, the keyword and a plurality of words starting with the word in the original text currently pointed to It is of course also possible to be able to compare columns.

Although the sentence to be translated and the sentence not to be translated are sorted and the sentence actually regarded as the translation target is sequentially translated, first, only the sorting is performed and a marker or the like is added. A method of collectively translating only the sentence with the marker may be considered.

Furthermore, in order to accurately judge the sentence to be translated, not only the "keyword" and the "word / word string" in the original sentence are compared and matched, but also the "keyword" and the "English part of speech". It is also conceivable that a pair of "" and a pair of "word / word string" and "English part-of-speech", which correspond to the terminal node of the parse tree after the original sentence is parsed, are compared and matched.

Further, in the present embodiment, the description of the keyword variation extraction is made by directly searching the translation dictionary, but the dictionary lookup buffer 6b (FIG. 1) storing the dictionary lookup morphological analysis result is used. A method of referring to and extracting the variation information may be considered.

Further, in the above embodiment, there is a step of extracting the variation form of the keyword, but when actually using the present invention, the order of performing this processing needs to be determined in consideration of efficiency and the like. There is. In this embodiment, for convenience of explanation,
As for all the cases that can be extracted in the variation form, only once at the beginning, the method is extracted and stored in the keyword variation form buffer 6f.

[0041]

【The invention's effect】

[Effect of claim 1] To promptly present a translation result by regarding the entire sentence including a specific keyword (= specific word or specific word string) registered in a table for speed-reading translation as a target of translation processing. You can

Further, since it is highly possible that the sentence (= the periphery of the keyword) containing the keyword registered in the table contains important information, the outline of the document can be presented. Therefore, according to the present invention, it becomes possible to assist the user in quickly grasping the outline of the document.

Furthermore, by making the table rewritable, it is possible to give the user room for customization.

Furthermore, by providing a framework in which one or a plurality of tables can be selected from the table for fast-reading translation, it becomes possible for the user to customize the fast-reading translation process according to the field to be translated. .

[Effect of Claim 2] A machine is referred to by referring to a specific keyword (= specific word or specific word string) registered in a table for speed-reading translation or a pair of the keyword and part of speech registered in the table. The translation result can be quickly obtained by targeting the entire sentence including the variation of the keyword extracted from the translation dictionary.

As described above, since the variation of the specific keyword registered in the table can be extracted from the machine translation dictionary, it is not necessary to additionally register the variation of the keyword in the table. Labor saving of table creation work and reduction of table size can be achieved.

[Effect of Claim 3] A specific keyword registered in the table for fast-reading translation or a pair of the keyword and part of speech registered in the table is used as a clue of the keyword extracted from the machine translation device. By considering the entire sentence including the variant as the target of the translation process, the translation result can be obtained quickly, and the keyword or the translated word corresponding to the variant of the keyword can be obtained from the table and desired by the user. The translation can be guaranteed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a functional configuration of an embodiment of a machine translation device of the present invention.

FIG. 2 is an example of a flowchart for realizing a speed reading process according to claim 1 of the present invention.

FIG. 3 is a diagram showing an example of a keyword table used in the speed-reading translation process according to claim 1 of the present invention.

FIG. 4 is an example of a flowchart for realizing a speed reading process according to claim 2 of the present invention.

FIG. 5 is a diagram showing an example of a keyword table used in a speed-reading translation process according to claim 2 of the present invention.

FIG. 6 is a diagram showing an example of a keyword table used for the process of the present invention according to claim 3 of the present invention.

FIG. 7 is a diagram showing an example of an original document used in an embodiment of the present invention.

FIG. 8 is a diagram showing, in order from the first word, each word and word position of the first sentence of the original document used in the embodiment of the present invention.

FIG. 9 is a block diagram showing a functional configuration of an example of a general machine translation apparatus.

FIG. 10 is a block diagram showing a functional configuration of an example of a conventional speed reading support machine translation device.

FIG. 11 is a diagram showing an example of a user dictionary (table) used in a conventional machine translation device for speed reading support.

FIG. 12 is a diagram showing an example of a date pattern table used in a conventional machine translation device for speed reading support.

[Explanation of symbols]

 1 Operation Control Means 2 Input Means 3 Output Means 4 Pipeline 5 Translation Module 5a Dictionary Lookup / Morpheme Analysis Means 5b Syntax Parsing Means 5c Syntax Converting Means 5d Target Language Generating Means 6 Operation Main Memory 6a Original Text Buffer 6b Dictionary Lookup Buffer 6c Conversion Previous syntax buffer 6d Converted syntax buffer 6e Translated text buffer 6f Keyword variation buffer 7 Translation rule storage means 8 Translation certificate dictionary storage means 9 Keyword table storage means 10 Translation mode selection means 11 Keyword table designation means 12 Keyword variation extraction means 13 Speed reading translation means

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takeshi Kitsumi 22-22 Nagaike-cho, Abeno-ku, Osaka-shi, Osaka Prefecture

Claims (3)

[Claims] 1. A keyword table storing means for storing at least one rewritable table in which a specific keyword is registered, a speed reading mode for translating a sentence including the keyword registered in the table, or a unit of a certain unit. In the case where the speed reading mode is selected by the translation mode selection means for selecting any one of all translation modes for translating all sentences, the keyword table specification means for designating at least one of the tables, and the translation means. A machine translation device with a keyword translation function, comprising: a translation unit that translates an entire sentence including a keyword registered in the table. 2. A dictionary storage means for storing a translation dictionary in which a word or a word string and a variation or a translation corresponding to the word or the word string are set, and a keyword and a part of speech registered in the table are referred to. 2. The machine translation device with a keyword translation function according to claim 1, further comprising a variation extraction unit for extracting the variation of the keyword from the translation dictionary. 3. Regarding a keyword registered in the table and a translation of the variation of the keyword, a translation corresponding to the keyword is retrieved from the table, and the inflection of the translation is utilized and changed as necessary. 3. The machine translation device with a keyword translation function according to claim 1, further comprising a translation means for outputting the item.