JPH01230177A - Translation processing system - Google Patents

Abstract

PURPOSE:To attain add-registration of a translating word by substituting the translating word based on the registering information of a learning table for a word and translating it with a translating means after storing and registering the learning table to register the output information of a post editing supporting part, a translating word selecting part, etc. on the word. CONSTITUTION:When an add-registering to a learning table by a translating word selecting part 9a ends, the effect is notified to a configuration element generating processing part 8a. As a result, at the configuration element generating processing part 8a, an output translating word is decided with referring to a learning table 9b to a parallel nodal point, and a configuration element generating processing is executed again. A translated result is displayed in the upper half area of the display scope of a display device 2 through a post editing supporting part 5 and an operational control part 3. Thus, a compromise dictionary register (add-registering) is attained without breaking a translation processing once.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an interactive translation processing system, and particularly to simple dictionary registration (additional registration) when editing translation results.

(Prior Art) Conventional interactive translation processing systems that allow manual intervention include one disclosed in Japanese Patent Application Laid-Open No. 61-260362, which performs translation processing at the time of translation based on the translation result of an input sentence. Corrections by replacing translated words are only possible within the scope of the dictionary held by the processing system.

For example, if we take a current English to Japanese translation processing system as an example, “That is the case for
The input sentence “thedefendant,” or “case” when translated as “That is the defendant’s argument.”
If you want to change the translation 1i5 for `` into ``saying'', use ``
The system displays all translated words that match the grammatical analysis in the content of ``C, a S e'', and redisplays the output sentence 11 times by selecting the translated word selected by the user. It has become.

That is, in the above example, for 'case', 1,
Rationale 2 is displayed, and the user selects the desired translated word "obate" from this display, and the system inputs the input h'? Based on the fti, it re-outputs that ``that is what the defendant is saying.''

(Problem to be Solved by the Invention) However, in the translation processing system having the above configuration, if the translation desired by the user is not registered in the dictionary that the system has in advance, -11 translation work is interrupted and the dictionary update work is performed. There was a problem that it had to be done.

SUMMARY OF THE INVENTION An object of the present invention is to provide a translation processing system that eliminates the problems listed in item L and allows the user to additionally register desired translated words without interrupting the user's translation work.

(Lower Stage for Solving the Problems) In order to solve the above-mentioned problems, the present invention provides a translation system comprising translation means for translating an input original text and converting it into a translated text, and a display stage for displaying the translation results. In this step, for the word selected based on the first input selection Mt information among the translation results, the display means causes the display means to select one of the candidate translation word and the additional registration, and the second word obtained as a result of the display is displayed.
a first means for outputting information indicating the candidate word when a candidate translation word is selected based on the input selection information of , and information indicating the input translation word when additional registration is selected; and a second means for storing a learning table in which output information of the means is registered for IF words, and after the registration, the translation means replaces the word with a translation obtained based on the registered information of the learning table. The translation is performed by

(Function) According to the present invention, since the translation processing system is configured as described above, the technical means functions as follows. The first means (for example, the host editor support section and the translation word selection section, etc., which will be described later) selects the relevant information from among the translations of the translation results displayed on the display means based on the first input selection information (for example, cursor information). For the word (translation word) indicated by , the display means displays a display for selecting one of a similar complementary translation word (obtained from the translation means, for example) and additional registration.

Further, the first means is based on the second input selection information obtained as a result of the above display, and when a candidate translation word is selected, information indicating the candidate translation word (for example, a number assigned in advance to each candidate translation word) is displayed. When the second means additional registration is selected, information indicating the input translated word (for example, a code indicating the translated word) is output to the second means. As a result, the output information from the first means for the selected word is registered in the learning table stored in the second means. After being registered in this way, the translation means (for example, a morphological analysis processing unit to be described later) refers to a learning table and replaces the selected word with a translated word obtained from the registered information, thereby translating it again. . In this way, if the desired translation is not among the candidate translations, by selecting additional registration,
Since the input translation words can be registered in the learning table and translated again with reference to this, the problems of the prior art can be solved.

(Embodiment) FIG. 1 is a block diagram of a translation processing system showing an embodiment of the present invention. In the figure, 1 is an input device such as a keyboard, 2 is a display device that displays translation results, editing results, etc. on the display screen, 3 is an operation management unit that controls the entire system, and 4 is a 1 that converts into a tree structure. - a translation processing unit that controls the entire translation process; 5 a post-edit support unit that controls editing after the translation result (post-editing); and 6a an analysis dictionary 6b.
7a is a grammar analysis processing unit that performs morphological analysis processing of tree structure data with reference to the grammar rule dictionary 7b; 8a is a generation dictionary 8
9a is a translation selection unit that performs translation selection based on input information and simple registration (additional registration) to learning table 9b during post-editing. .

Next, the operation of this embodiment will be explained with reference to the flowchart and process diagram of FIG.

First, a sentence input through the operation management unit 3 and translation processing unit 4 by key operations on the input device 1 is decomposed into morphemes by reference to the analysis dictionary 6b in the morphological analysis processing unit 6a (S1 ~S2). The input sentence is “Th” as in the conventional example.
at, is t, he case forthe d
'That-is-
the-case-for-jhe-dcfcndan
It is decomposed as t-,4. Next, the decomposed data is sent to the translation processing unit 4, where it is converted into tree-structured data with each morpheme as a node, and then sent to the grammar analysis processing unit 7a. In the grammar analysis processing unit 7a, as shown in the figure,
Referring to the grammar rule dictionary 7b, a grammar analysis process is performed in which each node is converted into a translated word corresponding to the original word, and into tree-structured data to which particles and the like are added (S4). In addition,
Examples of methods for realizing a tree structure include A, V, Eiho et al.
Translated by Yoshio Ohno [Data Structures and Algorithms] (1982-
3-10) Baifukan, P. 66-92 discloses that each node is associated with each morpheme, and grammatical information (headwords, translations, parts of speech, etc. ). Further, here, the specific translation processing methods (S2 and S4) are disclosed in, for example, Japanese Patent Laid-Open No. 58-40684.

In this way, the grammatically analyzed data is sent from the grammatical analysis processing unit 7a to the morpheme generation processing unit 8a, where morpheme generation processing is performed, thereby translating it into “or that is the defendant's argument.” (S5). This translation result is displayed on the display screen of the display device 2 via the translation processing section 4 and the operation management section 4. An example of the display is shown in FIG. 3(a). In the example shown in the figure, the original text is displayed in the upper half area 31 of the display screen, and the translated text is displayed in the lower half area 32. As a result of this display, if the user (operator) wishes to change the desired translated word in the translated text, the user (operator) can press a predetermined key on the input device 1 to change the desired translated word. Postedit Branch 5 will be notified. In response to this fear, the post-edit support unit 5 displays markers in a 7-space area 31 of the display screen at words whose translations can be changed from sentences that the user can select. An example of this display is shown in Figure 3(b).
It is shown in 31a. The sentences that can be selected here are sentences for which the tree structure as a result of the grammar analysis process (S4) is stored in the morpheme generation processing unit 8a. In the case of a system configuration with a small memory capacity, the last translated sentence is used. In addition, among the words at each node in step S4, only independent words that have little influence on the change result or the grammar analysis result are used as words whose translation can be changed. In other words, “ka” → “
It is not possible to change attached words (particles, auxiliary verbs) such as "and".

Next, the user moves a marker to a desired word using a cursor key or the like on the input device 1, and notifies the post-edit support unit 5 that the translation is to be changed using a key input or the like. The post-edit support section 5 notifies the translation word selection section 9a of the specified words. As a result, the translation selection unit 9a selects the candidate translation information for the word through the editing support unit 5 and the operation management unit 3.
Display on 2. An example of this display is shown at 32a in FIG. 3(b).

Here, as candidate translation information, the morpheme generation processing unit 8a
The translated words passed in parallel from the parallel nodes of the grammar analysis processing results held in
``point''). Alternatively, you may read the dictionary again.

If the desired translation Jb does not appear in the translation selection display (32a) displayed on the display screen of the display device 2 as described above,
The user gives an instruction to register a paraphrase translation to the translation word selection section 9a via the post-edit support section 5 by operating a key on the input device 1 or the like. In the example of 328 in FIG. 3(b), by inputting "3", "3. Paraphrase registration" is selected. Furthermore, when the user selects d-translation registration (additional registration), the user inputs the translation word to be additionally registered by operating keys on the input device 1, etc., and the translation word selection section 9
Send to a (S6, S7). As a result, the translation word selection section 9a registers the translation word taken by Uke in the learning table 9b.

Hereinafter, this simple registration will be explained in detail.

To realize simple registration, grammatical information (part of speech, usage)
A table (learning table 9b) is provided that holds only additional information (in this case, paraphrased translations) from the viewpoint of adding translations to the group, with the same words being one group. The format of this learning table 9b is shown in FIG. The learning table 9b in the figure is [Original language J (41)
, ``Part of speech'' (42), ``Usage J (43), and [number or paraphrase translation word J (44). It shall be defined so that a grammatical group can be uniquely determined based on this, the original language, and the part of speech.

For example, in the case of a noun, it is information on whether it is countable or uncountable, and in the case of a verb, it is information on which basic sentence pattern (sv, 5voo, 5voc, etc.) the verb takes. In addition, ``Number J (44)'' indicates each translated word, and when creating a dictionary, it is determined that ``Argument'' number = 1 ``Argument'' number = 2. As a result, in column 44, in 32a of FIG. 3(b), "argument" or "argument"
“1” if selected.

If "2" is entered and "3. Paraphrase registration" is selected. The paraphrase and translation word input by the user, such as "claim", is entered.

In this way, the translation word selection unit 9a selects the learning table 9b.
When the additional registration is completed, the morpheme generation processing unit 8a is notified of this fact. As a result, the morpheme generation processing unit 8
In a, by referring to the learning table 9a to determine output translation words for parallel nodes, morpheme generation processing is performed again, and the translation results are displayed via the post-edit editing section 5 and the operation management section 3. It is displayed in the upper half area 31 of the display screen of the device 1 (that is, the "argument" in 31a of FIG. 3(b) becomes "allegation" in the case of paraphrase registration). Note that for words that are not registered in the learning table 9b, the leftmost translation below the parallel nodes (i.e.,
may be output as "it" and "argument" respectively.

(Effects of the Invention) As explained in detail above, the present invention enables the user to update the dictionary (additional registration) without temporarily interrupting the translation process, so the operation mode is fast and simple, and dictionary addition is possible. This reduces the user's burden on the user, and as a result, improves the efficiency of the entire translation process. Originally, translation processing required a great deal of time to register specialized terms that differed from general terms, and each time required a means of shifting to dictionary maintenance work that was operated separately from translation processing. In the present invention, great practical effects can be expected by providing the user with a simple dictionary update function without going through dictionary maintenance processing.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a translation processing system showing an embodiment of the present invention, FIG. 2 is a flowchart and process content diagram showing the processing operation of the embodiment of FIG. 1, and FIGS. 3(a) and (b)
) is a display example of the display device, and FIG. 4 is a configuration diagram of a learning table. 1... Input device, 2... Display device, 3...
Operation management department, 4...Translation processing unit, 5...Host editor support unit, 6a---Morphological analysis processing unit, 6b...Analysis dictionary, 7a...Grammar analysis processing unit,
7 b ---grammar rule dictionary, 8 a = morpheme generation processing unit, 8b... generation dictionary, 9
a-...Translation word selection section, 9b...Learning table.

Claims (1)

[Scope of Claims] A translation system comprising a translation means for translating an input original text and converting it into a translated text, and a display means for displaying the translation result, wherein the translation result is selected based on first input selection information. The display means causes the display means to display a display for selecting one of the candidate translation and additional registration for the word that has been selected, and when the candidate translation is selected based on the second input selection information obtained as a result of the display, the candidate word is selected. and a second means for storing a learning table in which the output information of the first means is registered for the word. A translation processing system, wherein after the registration, the translation means performs translation by replacing the word with a translated word obtained based on the registration information of the learning table.