Method for automatic translation between natural languages

• US 4,641,264 A
• Filed: 09/07/1982
• Issued: 02/03/1987
• Est. Priority Date: 09/04/1981
• Status: Expired due to Term

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Nitta, Yoshihiko, Okajima, Atsushi, Yamano, Fumiyuki
• Primary Examiner(s): Chan, Eddie P.
• Application Number:
  US06/415,601
• Time in Patent Office: 1,610 Days
• Field of Search: 364/200 MS File, 364/900 MS File, 364/419, 434/157, 434/167, 434/169, 434/176
• US Class Current: 704/4
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/53 Processing of non-Latin text kana-to-kanji conversion G06F40/129; vowelisation G06F40/232
• CPC Class Codes: G06F 40/55 Rule-based translation

Abstract

An automatic translation method between natural languages comprises assigning parts of speech to words/idioms of an input text sentence by looking up a lexicon storage. The input text sentence in the form of string of parts of speech is segmented into phrasal elements as minimum units having linguistic meaning to assign parts of speech to the respective elements. The sequence of phrasal parts of speech is converted to a string of syntatic roles to the respective phrasal elements and words/idioms. Patterns representing a sentence, a clause and a quasi-clause are detected from the sequence of syntatic roles to transform the input text sentence to a skeleton pattern represented by a combination of those patterns. The sequence of the simple sentence, clause and quasi-clause inherent to the input language which forms the skeleton pattern is transformed to a sequence inherent to the output language. The sequence of syntatic roles of the simple sentence, clause and quasi-clause inherent to the input language which forms the transformed skeleton pattern is transformed to a sequence inherent to the output language. Finally, target language equivalents are assigned to the transformed string of syntatic roles.

Claims

1. An automatic translation method between natural languages using a translation processing apparatus to which an input text sentence composed of a natural language is inputted in the form of an electrical signal from an input device and which carries out predetermined translation processes by referring to an electronic data storage device having a word/idiom lexicon storage area in which target language equivalent information and part of speech information on words/idioms are stored and a plurality of table storage areas in which tables defining predetermined translation rules are stored, said method comprising:

(a) a first step of assigning parts of speech to a string of words/idioms of the input text sentence inputted from said input device by referring to said word/idiom lexicon storage area of said electronic data storage device so as to provide a corresponding string of parts of speech;

(b) a second step of dividing the input text sentence transformed in the form of said string of parts of speech into phrasal elements which are minimum units having linguistic meanings and assigning phrasal parts of speech to the respective phrasal elements by referring to a first table storage area of said electronic data storage device in which a table indicating a relation between patterns of strings of predetermined parts of speech which form the phrasal elements and phrasal parts of speech of the phrasal elements is stored;

(c) a third step of comparing the sequence of phrasal parts of speech assigned to the phrasal elements and the words/idioms with predetermined strings of parts of speech to assign syntatic roles to the respective phrasal elements and words/idioms, said predetermined strings of parts of speech being stored in a second table storage area of said electronic data storage device together with the syntatic roles associated therewith;

(d) a fourth step of detecting patterns representing a simple sentence, a clause and a quasi-clause from the string of syntatic roles to transform the input text sentence to a skeleton pattern represented by a combination of said patterns;

(e) a fifth step of transforming the sequence of simple sentence, clause and quasi-clause inherent to the input language which forms the skeleton pattern to a sequence of simple sentence, clause and quasi-clause inherent to an output language by applying a predetermined transformation rule which is stored in a third table storage area of said electronic data storage device;

(f) a sixth step of transforming the sequence of syntatic roles inherent to the input language within each of the simple sentence, clause and quasi-clause which forms said transformed skeleton pattern to a sequence of syntatic roles inherent to the output language by applying a predetermined transformation rule; and

(g) a seventh step of assigning target language equivalents to the respective ones of the transformed string of syntatic roles to generate a sentence in the output language in the form of an electrical output signal.

2. An automatic translation method between natural languages according to claim 1, wherein said electronic data storage device has a fourth table storage area in which a table including predetermined patterns of prohibited sequences of parts of speech is stored, and wherein when there are a plurality of parts of speech to be assigned to the word/idiom in said first step, the table of said fourth table storage area is referred to in order to finally determine one of the plurality of parts of speech of the word/idiom.

3. An automatic translation method between natural languages according to claim 2, wherein the table of said fourth table storage area further includes data representing special conditions under which the prohibition is released, and the part of speech of the word/idiom is determined by referring to both the pattern of prohibited sequence of parts of speech and the prohibition release condition.

4. An automatic translation method between natural languages according to claim 2, wherein the pattern of prohibited sequence of parts of speech comprises a string of parts of speech of a plurality of words before and after the word whose part of speech is to be determined.

5. An automatic translation method between natural languages according to claim 1, wherein said electronic data storage device has a fourth table storage area in which a priority table indicating priorities among parts of speech is stored, and wherein when there are a plurality of parts of speech to be assigned to the word/idiom in said first step, said priority table is referred to in order to finally determine one of the parts of speech of the word/idiom.

6. An automatic translation method between natural languages according to claim 1, wherein said electronic data storage device has a fourth table storage area in which a prohibition table storing therein predetermined patterns of prohibited sequences of parts of speech and a priority table indicating priorities among parts of speech are stored, and wherein when there are a plurality of parts of speech to be assigned to the word/idiom in said first step, said prohibition table is referred to in order to eliminate the prohibited parts of speech from the plurality of parts of speech of the word/idiom, and if two or more parts of speech still remain, said priority table is referred to in order to finally determine one of the parts of speech of the word/idiom.

7. An automatic translation method between natural languages according to claim 1, wherein said word/idiom storage area of said electronic data storage device includes a word lexicon storage area in which an index word for the word, information on the presence or absence of idioms formed by the word, a target language equivalent of the word and part of speech information are stored, and an idiom lexicon memory storage area in which an index word of the idiom, a target language equivalent of the idiom and part of speech information are stored, and the lexicon look-up in said first step comprises:

a first phase for segmenting a word of the input text and transforming the word to a stem if the word has an inflectional ending;

a second phase for looking up said word lexicon storage area by using the stem of the word as an index word, and if no idiom which includes that word is retrieved, storing the target language equivalent of the word and the part of speech information in a buffer storage area of a working storage used by said translation processing unit; and

a third phase for looking up said idiom lexicon storage area if an idiom which corresponds to the input text sentence is retrieved and storing the target language equivalent of the idiom and the part of speech information in said buffer storage area.

8. An automatic translation method between natural language according to claim 7, wherein at said third phase, when the idiom which includes the word stored in said buffer storage area is retrieved, the target language equivalent and the part of speech information of that idiom are stored in said buffer storage area while the word information of that word stored in said buffer storage area at said second phase is invalidated.

9. An automatic translation method between natural language according to claim 7, wherein at said third phase, when there are a plurality of idioms which corresponds to the input text sentence, the target language equivalent and the part of speech information for the longest idiom are stored in said buffer storage area.

10. An automatic translation method between natural languages according to claim 1, wherein said second step comprises:

a first phase for comparing the string of parts of speech of the input text sentence with the pattern of the string of parts of speech in said first table storage area of said electronic data storage device; and

a second phase for regarding the word/idiom as a node if the strings of parts of speech are not equal in the comparison and regarding the phrasal element as a node if the strings of parts of speech are equal and storing the part of speech information of the node and the discrimination of the word/idiom and the phrasal element are stored in a buffer storage area of a working storage used by said translation processing apparatus.

11. An automatic translation method between natural languages according to claim 10, wherein at said second phase, when a plurality of patterns of strings of parts of speech correspond to the input text sentence, the phrasal element for the longest pattern of string of parts of speech is regarded as the node.

12. An automatic translation method between natural languages according to claim 1, wherein said word/idiom storage area of said electronic data storage device includes a word lexicon storage area, a fixed idiom lexicon storage area for the idioms each comprising a predetermined plurality of words and a split idiom lexicon storage area for the idioms each having a blank in a predetermined sequence of words so that an idiom is formed by inserting an appropriate word, phrase or clause in the blank, said word lexicon storage area and said fixed idiom lexicon storage area being looked upon in said first step and said split idiom lexicon storage area is looked up in said second step.

13. An automatic translation method between natural languages according to claim 1, wherein said second table storage area of said electronic data storage device includes a table indicating a relation between strings of parts of speech of the words and the phrasal elements before and after the verbal word and the strings of known syntatic roles which may assume those strings of parts of speech, and said third step comprises:

a first phase for detecting a verb in the input text sentence; and

a second phase for comparing the string of parts of speech of the words and phrasal elements before and after the verbal word in the input text sentence with the strings of parts of speech in said table of said second table storage area, and if the corresponding string is detected, assigning the string of syntatic roles corresponding to the string of parts of speech in the table to the string of words and phrasal elements.

14. An automatic translation method between natural languages according to claim 1, wherein said second table storage area of said electronic data storage device includes a table indicating a relation between strings of known syntatic roles which the string of words and phrasal elements before and after a conjunction word may assume and a syntatic role of that conjunction word, and said third step comprises:

a first phase for detecting the conjunction word in the input text sentence;

a second phase for comparing the string of syntatic roles of the words and phrasal elements before and after the conjunction word with the strings of syntatic roles in said table of said second table storage area; and

a third phase for assigning the syntatic role of the conjunction word corresponding to the string of syntatic roles in the table of said second table storage area to that word when both strings of syntatic roles are equal in said second phase.

15. An automatic translation method between natural languages according to claim 1, wherein said second table storage area of said electronic data storage device includes a clause pattern table indicating a relation between clauses or quasi-clauses patterns and strings of syntatic roles which they may assume, and said fourth step comprises:

a first phase for segmenting the input text sentence by a conjunction word and punctuation;

a second phase for comparing the string of syntatic roles in the segmented section with the string of syntatic roles in said clause pattern table; and

a third phase for assigning the clause or quasi-clause pattern corresponding to the string of syntatic roles to the input text in said segmented section if both strings of syntatic roles are equal in said second phase.

16. An automatic translation method between natural languages according to claim 15, wherein said second table storage area of said electronic data storage device further includes a skeleton pattern table indicating a relation between the sequence of clause or quasi-clause patterns and the skeleton patterns, and said fourth step comprises:

a fourth phase for comparing the sequence of clause or quasi-clause patterns assigned to the input text with the sequnce of clause or quasi-clause patterns in said skeleton pattern table; and

a fifth phase for assigning the corresponding skeleton pattern to the input text when both clause or quasi-clause patterns are euqal in said fourth phase.

17. An automatic translation method between natural languages according to claim 1, wherein said third table storage area of said electronic data storage device includes a tree/list transformation table indicating a relation between the skeleton patterns in the input language and the corresponding skeleton patterns in the output language, and said fifth step comprises:

a first phase for comparing the skeleton pattern of the input text with the skeleton patterns in the input language stored in said tree/list transformation table; and

a second phase for assigning the corresponding skeleton pattern in the output language to the input text when both patterns are equal in said first phase.

18. An automatic translation method between natural languages according to claim 1, wherein said third table storage area of said electronic data storage device includes a transformation table indicating a relation between the strings of syntatic roles in the input language and the corresponding strings of syntatic roles in the output language, and said sixth step comprises:

a first phase for comparing the string of syntatic roles of the sentence, clause or quasi-clause which forms the skeleton pattern in the output language assigned to the input text with the strings of syntatic roles in the input language stored in said transformation table; and

a second phase for transforming the string of syntatic roles of the sentence, clause or quasi-clause to the corresponding string of syntatic roles in the output language when both strings of syntatic roles are equal in said first phase;

said first and second phases being repeated until a category of a node having the syntatic role assigned thereto appears as the phrasal element on the word.