JPS63108461A - Machine translator - Google Patents

Abstract

PURPOSE:To efficiently translate and edit a natural language by constituting a proofreading/editing part, a translating part, a translation dictionary, etc., in the same concept for two languages of a translation language pair, and providing a displaying part to display in contradistinction to two languages. CONSTITUTION:English sentences and Japanese sentences inputted from an input part 1 are stored into a sentence storing part 2 (2a). A translating part 4 refers to a translation dictionary and grammar 5 by the control of an editing control part 3, and executes the machine translation processing of the sentence stored into the storing part 2 successively at a prescribed processing unit. Next, the translated sentence of a processing result is corresponded to an input sentence, controlled and stored into a storing part 2 (2b). In such a case, for the knowledge information stored in the dictionary and grammar 5, two bidirectional languages to be translated are prepared for, changed over and used in the translating direction and the English sentence and Japanese sentence stored into the storing part 2 are mutually corresponded, and simultaneously, displayed to the CRT of a displaying part 8. Thus, an editing part 3, etc., is constituted of the same concept to two languages of the translation language pair, are displayed in contradistinction to two languages, and the translation editing to the natural language can be made efficient.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a machine translation device that allows bidirectional translation processing.

(Prior Art) Recently, a machine translation service has been attracting attention, which uses a computer to automatically translate an input original text and obtain the translated text. For example, when you enter Japanese and get the English translation,
On the other hand, various attempts have been made to develop a natural language machine translation device that manually reads an English sentence and obtains its Japanese translation.

Basically, this machine translation device performs morphological analysis, syntactic analysis, and semantic analysis of the input original text and converts it into words (words and phrases).
Divide the human original text into predetermined language processing units such as (1), and (1) find the translation word (translated phrase) for each language processing unit by referring to the 1-1 blue report necessary for translation processing and the stored translation dictionary 2. seek,
(2) A clock is configured to combine these translated words (save words) according to predetermined translation rules to generate a translated sentence 2 for the above-mentioned translated sentence.

However, in the prior art, since there is no machine to judge the language of human text, translation can only be performed in one direction, and when changing the translation direction, it is necessary to replace the system. For example, when the English-Japanese translation is finished and you want to continue the Japanese-English translation, ! ? The operability was extremely poor, with the system starting up as a Japanese-to-English translation system.

(Problem to be Solved by the Invention) In this way, conventional translation systems do not have a function to determine the language of the input text, so translation is only performed in one direction, and it is difficult for the user to input text while always being aware of the language of the original text. The present invention was made in consideration of these circumstances2, and its purpose is to provide a method for translating a translation language pair in both directions using exactly the same concept. It is an object of the present invention to provide a highly practical machine translation device that can be operated and that can improve the efficiency of natural language translation.

[Structure of the invention]

(Means for Solving the Problems) The present invention provides an input means for manually inputting a first M language sentence or a second language sentence, a translation process from the first language to a second language, and a second language sentence.
A translation dictionary storing knowledge information used for translation processing from one language to a first language, and identifying whether the human-powered original text from the input means is a first language or a second language, and using the corresponding knowledge information, Translation means for translating into a second language when the input original text is in the first language, and translating it into @11 words when it is in the second language;
and a memory system for storing the input original text and the translated text obtained by the translation means in correspondence with the input original text.

The present invention is characterized by comprising display means for comparing and displaying the input original text and the translated text corresponding thereto.

(Function) The present invention is constructed with the same concept for two languages: a translation dictionary section, a station section 1, a pair of translation languages, and an editorial section 2, and has a display section that displays the two languages in comparison. In this case, there is no obvious distinction between which is the original text and which is the translated text. It will be done. For example, if the Japanese language is transferred to the Japanese side by some means (OCR)
.

If it is input manually using a keyboard, floppy disk, etc., it will be translated into English, and vice versa, if English is input into the English side.

Translated into Japanese.

(Embodiment) An embodiment of the present invention will be described below with reference to one drawing.

This example inputs English/Japanese sentences and performs mutual machine translation into Japanese/English sentences.

gi is a schematic configuration diagram of the embodiment device.

In Fig. 1, 1 is an input section consisting of a keyboard, etc. V is an English sentence 7 consisting of a character string etc. input through this input section 1, and an input sentence is provided on a translation plate. It is stored in the sentence storage unit 2 (2a) as .

Under the control of the first editing section 3, the translation section 4 refers to a translation dictionary/grammar 5 that stores knowledge information necessary for translation processing in advance, and interprets the previously input data stored in the sentence storage section 2. Sentences (hereinafter referred to as original texts) are sequentially subjected to machine translation processing at a predetermined processing level. The translated text obtained by this translation W54 is associated with the original text obtained and stored in the sentence storage unit 2 (
2b).

The knowledge information stored in the dictionary/grammar 5 includes, for example, a rule/irregular variation dictionary 5a and a word dictionary 5b.

It consists of an analysis grammar 5c, a conversion grammar 5d, a generation grammar 5e, a morpheme generation grammar 5f, etc. This knowledge information is prepared for both bidirectional and Japanese languages to be translated, and is referred to by switching in the direction of translation.

The editing control section 3 drives the display control section 7 to associate the English sentences 7 and the Japanese sentences 2 stored in the sentence storage section 2 with each other, and displays them simultaneously on the display section 8 consisting of a CB, T display, etc. it's shown. By displaying the original text and the translated text on the display section 8, post-editing processing is performed on the translated text 'C1. this? The lkm collection process is performed in accordance with control information given from the human resources section 1 and by referring to various knowledge information stored in the translation dictionary/grammar 5.

The printing unit 6 is for outputting a hard copy of the completed English/Japanese text after the above-mentioned post-editing process.

By the way, what about the keyboard that constitutes the human power section 1?

For example, it is configured as shown in FIG. That is,
This keyboard is 1, for example, key details 1 for inputting character data.
In addition to a, key 1b for 8 translation instructions, key ff for editing
P1 c0 Function 1] Key details for the bell 1d.

and cursor control key 1 #1 in the display section 8
It is configured with 2 such as e.

Further, FIG. 3 shows an example of the configuration of a display screen for English sentences and Japanese sentences on the display section 8. As shown in FIG.

Here, a translation editing area 13a is shown above the display screen.

The display screen is composed of a text display area xi-sb on the left side of the screen, a text display area #, 2-sC on the right side of the screen, and a message display area 8d'9 at the bottom of the screen.

In this large display area 1-8b, the sentence part 1 (for example, an English sentence) stored in the sentence storage part 21f is displayed.
In the sentence display area 2-8C, the sentences (N, t, Japanese) stored in the sentence storage section 2b are displayed in correspondence with the original sentences from which the translated sentences have been obtained.

FIG. 4 shows the basic operation sequence of the embodiment device having this configuration.

The editing control unit 3 follows this operation sequence,
Translation completion information given from the translation section 4 and the input section 1
Various key information 2 inputted from the computer is determined, and the translation/editing process 2 is performed interactively.

That is, the first editing side unit 3 monitors the translation processing status in the translation @ 44. In step A), when the translation unit 4 detects the completion of the final translation for one original text, the translation unit 3 monitors the translation processing status in the translation @44. The translated text is stored in the sentence storage unit 2 and displayed on the display unit 8 in correspondence with the original text of the translated text 2 (step B).

In addition, when the translation process is in progress in the translation unit 4, or when the translation process is not being performed, the editing unit 3 determines the 1 blue report manually input from the input unit 1 (steps C and D). , g, F). If the input key information is a "translation instruction key" 8 (Step C) 1 Edit 1111J
The translation unit 3 supplies the input sentence Cf11 stored in the sentence storage unit 2 (for example, the sentence indicated by the cursor) 2 to the translation unit 4, and starts the translation unit m (step G). When taken out, the direction to be translated is also extracted and notified to the translation section 4 (step M).

If the manual key information is "r character key" (step D), the character code indicated by that character key is stored in the manual buffer. In step H), the character code and the character code are stored in the sentence storage section 2.

The character pattern is entered into the display section 8 (Step E). By displaying each character pattern of the character code stored in this input buffer, the sentence entered manually from the input section 1 is displayed.

Furthermore, if the manual key information is "r edit key" (step E), the editing process corresponding to the edit key is executed on the translation (step J).

Similarly, if the manual key information is "iso (junior key)" (step F), the process corresponding to the first function key is executed (step K).

If there is no key information, or if the input key information is something other than the above-mentioned "key", the process 1 for that pond, for example, outputting a hard copy of the original text and translated text obtained in the sentence storage section 2, etc. It will be done.

At this time, the original text and the translated text are distinguished by, for example, adding a mark.

According to such an operation sequence of the editing control unit 3, for example, an operator selects the character key group 1a on the keyboard.
When characters are input by operating , the character information is sequentially set in the input buffer and sequentially stored in the sentence storage unit 2 as the original text to be provided at the translation unit (steps D and H). For example, the display area 8b+Set, which is controlled by the cursor control key group 1e, is stored on the corresponding memory sections 2a, 2b+. The input sentence is then displayed in the large display areas 8b/8c of the display section 8 where the cursor is displayed (step I).

If the translation instruction key 1b2 is operated at any point in character input, for example, at the end of inputting one sentence.

According to the key human power information and translation direction, the translation process for the input sentence stored in the background is started (
Steps e, G, M). When the translation process is completed, the translated sentence thus obtained is displayed in the sentence display area on the opposite side of the input sentence of the display section 8 (steps A and B).

In addition, if editing such as correction of manual text is necessary.

1 in the middle of inputting a sentence by operating the character manual key details 1a.
For example, the editing process is executed by operating the cursor control key group 1eyP: to move the cursor to the correction point I, and operating the editing key details IC for correction, insertion, deletion, etc. (step E).

J).

In this manner, the basic operations of machine translation processing are controlled.

By the way, the mutual translation section 1 is executed as follows, for example.

FIG. 5 is a block diagram of the translation section 4, and FIG. 6 shows an example of its processing sequence.

The translation unit 4 first extracts the direction requested by the translation process (
Step a).

To extract the translation direction, if you input the sentence to be translated from the keyboard, the translation direction will naturally be the input sentence and 1.
For example, if you input a sentence to be translated in English -, the translation direction will be English → Japanese. Also, if the sentence to be translated is input from OCR etc.

For example, in mutual translation between English and Japanese, the input sentence can be determined from the character type (JIS-CODE for Japanese and ASCII-CODE for English). In addition, by comparing the input sentence with a dictionary, it is also possible to determine which dictionary contains more relevant words as the input source.

Furthermore, when retranslating, the translation direction may be determined in the opposite direction from the area where the cursor is displayed. If you expand this function, you can translate the result of one translation in the opposite direction, and the translation result will be 1, for example, 1 editing area 8b instead of the translation side.
It is also possible to easily display f. With this function, 1. For example, even if an unintended translation is made due to ambiguity caused by a supervisor or multi-product side, by comparing the original text with the original text that has been translated, it is possible to immediately determine that the translated text is not the intended translation. Become.

Once the translation direction has been extracted, the original text is morphologically analyzed (steps b, c, and d). This morphological analysis is based on 11J
It depends on the original text being translated. When the original text is in English, use the rule/irregular change dictionary 5a stored in the translation dictionary section 5#1 to perform a process F11 in which, for example, a word with a conjugation change or a change in ending is converted into its original form (basic form). , information such as part-of-speech information and translated words is obtained from this original form using the word dictionary 5b (step C).

On the other hand, when the original text is in Japanese, information 2 such as part-of-speech information and translated words is obtained in the same way using a word dictionary @5b and an adjunct word dictionary (not shown in FIG. 1) 2 (step d).

The original sentence that has been morphologically analyzed consists of multiple part-of-speech sequences, and there are many part-of-speech sequences that are incorrect. For this reason, syntactic filtering is performed to check whether the string is syntactically correct and to perform syntactic analysis on the part-of-speech string determined to be syntactically correct (step e). In syntactic filtering.

The filtering grammar (not shown in FIG. 1) is switched and referenced in the translation direction.

For the syntactically filtered correct part-of-speech sequence, the original sentence is parsed (step g) under a parsing constraint (step f). For syntactic analysis, the syntactic analysis grammar 5C is switched and referred to in the translation direction.

This parsing is repeated until the parsing of the original text is successful. This syntactic analysis determines the arrangement structure of the parts of speech of the original language that makes up the original sentence, and the mode of relational 1 tense in which the words are used.

If the original text is in Japanese, semantic analysis is performed on this parsed structure (step h).

i).

Thereafter, this parsed original sentence structure 2 is converted into a translated sentence structure using the conversion grammar 5d (step j).

In other words, using conversion grammar 5d, if the original text is English,
English syntactic structures are converted to Japanese structures, and if the original text is Japanese, the Japanese structures are converted to English structures using respective conversion grammars. still.

If this structural conversion fails, it is assumed that there is an error in the syntactic analysis of the original text, and the processing is restarted from the syntactic analysis process of the original text.

After that, translation word candidates are determined for each language of the original text based on the syntactic structure of the translation text.

The word order is rearranged according to the language form of the translated text to obtain translation candidates for the original text mentioned above (step k).

This syntax generation process is performed with reference to each generation grammar in the translation direction according to the generation grammar 5e.

Then, each translation word candidate (basic form) constituting the translation candidate is subjected to conjugation transformation and inflection transformation processing for each translation direction according to the syntactic analysis result of the original sentence and the morpheme generation grammar 5f, and the translation sentence 2 is completed (step l , m, n)
. Since morpheme generation also depends on the translation direction, it is separated in the example.

The translated text for the original text obtained in this way is stored in the text storage section 2, and is also displayed in correspondence with the original text. If a plurality of translations for one original text are obtained, for example, only one translation is displayed, and at the same time, the existence of Ike's translation candidates is displayed for identification.

By the way, after the translated text obtained in this way and the original text from which the translated text was obtained are displayed on the display section 8, if there is an instruction using the editing key, the translation editing process for the translated text is as follows. It is done by straining.

The translation editing process for this translation is! ! f Basically, for the word (original word, original word phrase, translated word, translated word phrase) indicated by the cursor on the screen of the above-mentioned roasting section 8.

This is realized by performing Process 2 corresponding to the operated edit key. in particular.

■ Insert 2 characters before the cursor position by operating the insert key, ■ Insert 2 characters before the cursor position by operating the delete key? delete.

■Move the characters within the range indicated by the cursor by operating the movement keys.

(2) Cancel key operation l This invalidates each editing function specified by each of the above keys.

■Searching consists of translation editing processing such as displaying other candidates for the word/phrase indicated by the cursor by key operation.

Further, when the above-mentioned function control key is operated, the following functions are provided, which are used for translation editing of the above-mentioned translated text.

That is.

- When the translated word display key is operated, the translated word 2 is displayed for the word in the translated text indicated by the cursor.

■When the translated word display key is operated, the contents of the translation dictionary with the word in the original text indicated by the cursor as the entry word will be displayed.

■When the dictionary registration key is operated, the character string indicated by the cursor is registered in the dictionary as a new word/idiom.

■When the dictionary deletion key is operated, the registered new words and phrases are deleted.

Functions such as f are thus realized.

In addition, instructions such as character strings (@) using the cursor mentioned above are
Does the operation of the cursor movement key move the cursor on the display screen? This is done by enlarging or reducing the cursor size using the cursor side leg keys while moving it.

Through these various functions, processing such as inputting the original text, giving instructions for translation, and modifying the translated text are performed interactively on the V original text, regardless of whether it is English or Japanese.

These are the basic translation functions that Honzue has.

The translation unit 4 mentioned above will be explained in more detail.

In FIG. 5, the translation direction extraction unit 401 has been described above, so its description will be omitted. When the original text is Japanese, the Japanese morphological analysis unit 402 needs to correctly recognize one clause because words are not expressed by separating them with spaces, unlike in Western original languages. For this reason, in the embodiment, the daily English rate dictionary 403 and the subsidiary word dictionary 4042 are searched.

The connection conditions here are being determined.

For example, in the case of ``need'', incorrect clause sequences are eliminated, such as necessitate (° verb) - saucer * (prefix) ten suru (verb) - rejection. Similarly, sticky words are rejected based on connection conditions to narrow down candidates for morphological analysis. In the example, the independent word dictionary and attached word dictionary are separated, but!
There is no problem if you stop.

In addition, if the phrase to be translated is not stored in the Japanese-English word dictionary 403, the original sentence is divided into smaller units E and the search is performed sequentially from the front. If the above connection condition is not satisfied, The words up to the point immediately before the next Peng word found are treated as unregistered words, and are then processed in the same way as normal phrases.

The English morphological analysis unit 405 performs word-by-word morphological analysis 2 when the original text is in English. In the example, English-Japanese Linguistic dictionary-1F4
07 is stored in its original form. For this reason.

After converting the words in the original text into their original forms while referring to the irregular change dictionary 406, a search is performed.

This English morphological analysis unit 405 is performed as follows. For example, when the word ``f-esJ'' appears in the original text,
What is the possibility of its original form?

■ flie (taking 32 plural) ■ fli
(take tt's es+2)■ fly(tt
(Change iea of y to y) ■ fly (3rd person, fathom, take current 3) ■ fli (take e3 of I) ■ fly (change y to iea%3NC'R).

In the next morphological analysis, for the bltQ notation jfliesJ, this number of possibilities 2 is extracted and a term f search is performed. In this example, j flieJjfliJ does not exist in the dictionary.

fty (courtesy; noun; fly; verb) Only that is required as the original form.

Since the original sentence subjected to form B elemental analysis consists of multiple part-of-speech sequences, the syntactic filtering unit 408 extracts the combination 2 of all part-of-speech sequences.
Since it takes a huge amount of time to parse the syntax, 11
It is checked to see if it is grammatically correct and prevents incorrect part-of-speech sequences from occurring. It also determines the order of the part-of-speech strings to be parsed. Since the determination criterion differs depending on whether the original text is English or Japanese, the Japanese-English filtering grammar 409 is referred to when the original text is Japanese, and the English-Japanese filtering grammar 410 is referred to when the original text is English.

The syntactic analysis leg 411 takes the part-of-speech sequence 2 that has been determined to be syntactically correct by the syntactic filtering unit 408.

I have control over syntactic analysis, semantic analysis, and structural transformation.

If the part of speech is not accepted during analysis, control is given to analyzing the part-of-speech string with the highest priority.

In addition, there is a safe restraint in the translation of parts such as inserted phrases.

The parsing unit 412 parses the part-of-speech sequence sent from the parsing control unit 411. This will be explained using an example where the original text is in Japanese. The derivation of the process is performed by one phrase derivation. In the implementation column.

(Prefix) An independent word (Suffix) (Adjunct word) There is always an independent word, followed by an attached word, which is not considered a clause, but only prefixes and suffixes are also interpreted as clauses. .

ATNG (
It is analyzed using Extended Transition Network Grammar) and converted into a conceptual structure.

This extended transition network grammar has an actor on the arc of the transition network, y
This was expanded by providing additional conditions and conditions.

Regarding the grammar to be parsed, when the original text is Japanese, the Japanese-English parsing grammar 413 is referred to, and when the original text is English, the English-Japanese parsing grammar 414 is referred to. The syntax analysis unit 412 is shared.

The semantic analysis unit 415 uses case grammar (@
(word case pattern), semantic analysis information.

Using heuristics such as position, determine the most likely and reliable dependency, and convert from the standard conceptual structure to the Japanese conceptual structure.

The implementation column is set. The processing contents of semantic analysis are shown below.

■Tracing upward from the lower-level nodes to the concept nodes and connecting nodes that may be dependent on each other with virtual arcs. Also,
The connection condition is that a rule is added to a lower node and the higher node matches that rule.

for example.

1f (a-noun ten-particle & b-verb) then (arc connection) Now, if the lower node is a noun with a case particle.

The rule is that if there is a verb at the top, it is connected with an arc.

■Uke prioritizes each arc seen from the side.

■From the above ■, each arc f attachment 71
The pending status is determined based on the priority level 2 which has been set to 0.

for example.

In ``Fat Be Raising Hanako and Tulips'', the standard conceptual structure after the morphological analysis and syntactic analysis is as shown in FIG. 7(a).

For this reference conceptual structure, the second line of processing in (■) above, and “
It turns out that there is a possibility that "Hanako" will "raise" l 5. The conceptual structure is changed as shown in FIG. 7(b).

Next, when processing (2) above is performed, the Raku pattern added to "Raise" and the semantic information of each node are used. For example, the Raku pattern of ``nurturing''.

Assuming that N (person) grows N (person% animals, plants 1 m) f grows N (person) and grows 2, do pattern matching with this Raku pattern and the three arcs connected to "grow". . In the embodiment, matching is achieved with all three arcs, and priority is given to the connection of each arc.

Next, we will evaluate ``r tulip'' and ``hanako'' with the request, but the combination with ``and'' (parallel particle).

When the meaning of each meaning is 1, it is used, and when it differs, points are deducted. This completes prioritizing all arcs.

Most importantly, perform the process described in (2) above. If there is only one arc emerging from a lower node, that arc is determined. Furthermore, something with multiple arcs coming out from a lower node to a higher one? Decide on one target. In the implementation column,
"Hanako" is the target, and the priority evaluation of the two arcs connected from "Hanako" in the upper order is performed. In this case, the above-mentioned step (1) is performed. In this process, the example sentence is changed to the Japanese conceptual structure shown in FIG.

The structure conversion unit 416 performs processing independently in three stages. First, local adjustment of the concept W structure is carried out, firstly, word conversion 2, if the original text is Japanese, Japanese-English conversion grammar 4
17. In the case of English, the conversion is performed with reference to the English-Japanese conversion grammar 418, and finally common processing for each part of speech is performed.

Local adjustment of conceptual structure 1 For example, the processing will be explained in the case where the original text is Japanese.

For verbs that take the ``to'' character in quotation, the subject may be omitted in Japanese in the embedded sentence of the quotation.In this case, the subject of the quoting verb is supplemented. For example, 1 Japanese-English Conversion Grammar 417.

1(ga-2tp-3(Iga)=l(ga-2to-
3(subj-4)/ is stored, and when the original sentence 2, "He said he wanted to go," is translated, sub j is supplemented as shown in FIG.

Word conversion determines the translated word based on the partial structure of the concept, and also converts the word usage as necessary.

The verb converts the Japanese case particle dependency into the English Raku pattern. For example, the grammar for "1 Translate" is.

1(ga-2'i'wo-3'i'n1-4)-1(s
ubj-2obj-3into-4)/, the structure is converted as shown in FIG.

The common processing for each part of speech is word transformation using the grammar specific to each word search.
conduct. The conversion of Raku that is necessary when determining the translation of the verb 2 is carried out by word conversion, but arbitrary dependencies that do not apply are also processed 2. Gabe.

Matching of the verb in the embedded sentence and the verb in the main sentence is performed.

The syntax generation unit 419 converts the conceptual structure converted by the 1m syntax conversion unit 416 into a surface structure. If the original text is Japanese, refer to the Japanese-Japanese Generative Grammar 420'5-, and if the original text is English, refer to the English-Japanese Generative Grammar 4212.

Since morpheme generation is different between Japanese and English, the English morpheme generation unit 422 uses English morpheme generation grammar 4
232 to generate a translated sentence. For Japanese, the Japanese morpheme generation unit 424 refers to the 1-Japanese morpheme generation grammar 425 to generate a translated sentence.

As described above regarding this apparatus, mutual translation processing can be realized very easily and efficiently using one apparatus. Also, as mentioned above, each processing section.

Since the hand-grinding part and the regular part are completely separated, it is purchased as a dog, so si1! In the example, I explained the basics of mutual translation between Japanese and English, but it is of course possible to easily achieve mutual translation between the original languages of Ike by replacing the rule part.

In addition, this healing is not limited to the above-mentioned implementation sequence,
It is a matter of course that it can be realized in four different ways without departing from the gist.

〔Effect of the invention〕

Thus, according to the present invention, the original text and the translated text can be distinguished.

There is no difference in the operator's operation for them, and the operation is a single line and there is no wasted force. Therefore, in the comprehensive translation process of translation and post-editing, great practical effects can be achieved, such as increasing the efficiency of invasive translation processing for natural languages.

[Brief explanation of the drawing]

The figures show one embodiment of the present invention. Fig. 1 is a schematic diagram of the structure of one embodiment, Fig. 2 shows the arrangement of the keyboard, and Fig. 3 shows the arrangement of the display screen. 4 is a diagram showing the processing flow of the editing control section, FIG. 5 is a diagram showing the configuration of the translation section, FIG. 6 is a diagram showing the basic flow of the translation section, and FIG. 7 is a diagram showing the basic flow of the translation section.
The figure shows the conceptual structure for the human source original text, Figure 8 shows the Japanese concept *structure 2 for the input original text, Figure 9 shows the local adjustment of the conceptual structure, and Figure 10 shows the concept t4. It is a figure which shows the situation of the word-form conversion of い. 1... Input section, 2... Sentence storage section, 3... Edit control section. 4... Translation section, 5... Dictionary/grammar, 6... Printing section, 7... Display device phloem, 8... Display section. Agent Patent Attorney Nori Chika Ren Introvert Kikuo Takehana Figure 3 Figure 4 Figure 6 Figure 6

Claims (1)

[Claims] An input means for inputting a first language sentence or a second language sentence, and a translation dictionary storing knowledge information used for translation processing from the first language to the second language and from the second language to the first language. and identifying whether the input original text from the input means is in a first language or a second language, and using knowledge information corresponding to this, when the input original text is in the first language, performs translation processing into the second language. In the case of a language, a translation means for translating into a first language, a storage means for storing the input original text and the translated text obtained by the translation means corresponding thereto, and the input original text and the corresponding translation text. A machine translation device comprising display means for comparing and displaying.