Báo cáo khoa học: "Report on Research"

Chia sẻ: Nghetay_1 Nghetay_1 | Ngày: | Loại File: PDF | Số trang:0

Thêm vào BST

Báo xấu

45
lượt xem 2
download

Download Vui lòng tải xuống để xem tài liệu đầy đủ

THE CAMBRIDGE Language Research Unit is primarily concerned with analytic investigation of language, and in particular with a correlative study of the descriptive-linguistic, logical, algebraic and other notational characteristics of natural languages and of translation between natural languages. Much of this work is relevant to machine translation and the following four sections by members of the unit illustrate some of the applications that are being made.

Chủ đề:

Bình luận(0) Đăng nhập để gửi bình luận!

Lưu

Nội dung Text: Báo cáo khoa học: "Report on Research"

[Mechanical Translation, vol.3, no.2, November 1956; pp. 36-37] Report on Research Cambridge Language Research Unit, Cambridge, England THE CAMBRIDGE Language Research Unit is is represented by a numeral which would be primarily concerned with analytic investigation calculated from information to hand respecting of language, and in particular with a correla- the current context-situation and the code- tive study of the descriptive-linguistic, logical, number given for the input word under consider- algebraic and other notational characteristics ation, then we could dispense with any actual of natural languages and of translation between t hesaurus entries in the computer's storage, natural languages. Much of this work is rele- all the relevant information being contained in vant to machine translation and the following the input and output dictionaries which respec- four sections by members of the unit illustrate tively provide the code-number of the input some of the applications that are being made. words and decode the numbers, calculated from The first three are concerned with the possibi- these and the context, into the target language. lities of using a mechanical thesaurus; the Mathematically, the problem is completely fourth deals with mechanical translation via an soluble, provided no limit is placed on the length interlingua. of numerical symbols. If we limit ourselves to a practicable length of symbol, the question of Potentialities of a Mechanical Thesaurus adapting the general mathematical solution to (M. Masterman) actual use becomes one of ingenuity which can The unit of a mechanical dictionary is the probably be solved, but which can only be as- semantically significant "chunk", not the free sessed by practical effort. The mathematical word. From a logical point of view, uncoded procedure consists in finding a set of Boolean MT dictionary entries form "trees", the paths operations having certain prescribed properties of which can be determined, to a significant which can be deduced from the conditions of the extent, by objective criteria. However, these, problem. These operations are few in number as they stand, are too complicated to be used and could be built into a computer; being directly for MT. Boolean, they can be performed with very great Attempts to construct multilingual MT dic- speed. tionary entries show that the entries form, not A model solution, substantially simpler than trees, but algebraic lattices, with translation that recommended for actual trial, will be de- points at the meets of the sublattices. It also scribed, and an example worked in it will be emerges that the complexity of the entries need demonstrated. This will show up the sort of not increase greatly with the number of lan- crossword-puzzle ingenuity required to devise guages, since translation points can, and do, a suitable context classification. The attrac- fall on one another. tion of the method, despite this inevitable lack Such a multilingual MT dictionary is analogous, of elegance, is that it makes the computer in various respects, to a thesaurus. A method actually calculate instead of merely looking of using such a thesaurus to refine the mecha- things up in lists, and thus makes the whole pro- nical pidgin output of a bilingual mechanical cedure capable of sufficient speed to be feasible dictionary has been devised. for a mechanical-translation program. Mechanical Translation Program Utilizing an Linguistic Basis of the Thesaurus-Type Mecha- Interlingual Thesaurus nical Dictionary and its Application to English- (A.F. Parker-Rhodes) Preposition Classification. The problem of setting the information con- (M.A.K. Halliday) tained in a fully general interlingual thesaurus The thesaurus method of mechanical lexico- into coded form for the use of an electronic graphy is an attempt to systematize the lexis computer would be formidable if not impossible in such a way that the "one-to-one word equi- of practicable solution, if it were necessary to valence" principle can be maintained as the include every entry as such. But if we could first stage in the dictionary, since the mechani- devise a system of coding such that each entry cal application of the concept of "primary
Report on Research 37 denotes some basic idea such as plurality. meaning" implicit in this principle requires the animal, or negation. A word in Nude may con- arrangement of secondary translation equiva- sist of one letter only; the more complex a lents into contextually determined systems. notion, the more letters are required. Each Each entry, consisting of a "key word" and its word in Nude is regarded as a relation, either associates, constitutes one such system. 0-ad, 1-ad, or 2-ad; 1-ads are preceded by a Multiple translation equivalence requires the point, 2-ads by a colon. Punctuation in Nude specification of the conditions under which one is used to indicate the concatenation of the of the terms in the closed system of a thesaurus words. The words linked by a 2-ad relation entry is to be selected, these conditions being precede it and are separated by a comma, e.g., contextual features of the target language. This A, B:C; coordinate conjunction is expressed by is illustrated by a "context-continuum" showing a hyphen, e.g., A-B.C. some word equivalence in non-technical rail- way terminology in four languages. The translation program involves the follow- The thesaurus exploits the redundancy of the ing operations: target language by handling its word classes without comparative identification. The (1) Matching semantically significant "chunks" autonomous treatment of the target language reduces the loss of determination involved in of the base passage against the Base-Nude the translation process. dictionary. Among the word classes established for En- (2) Reorganization of the syntax into Nude glish as a target language, prepositions are syntax by the method of cyclical reduction particularly suited by their relatively low en- described at the 1955 Symposium of the tropy to non-comparative treatment. Preposi- Cambridge Language Research Unit, utili- tions are classified as "determined" and zing the word-class sequence entries of "commutative". The former are listed as sub- the Base-Nude dictionary (cf. MT III, 1). entries of the determining word, having a single (3) Treatment of chunk-chunk, chunk conju- or multiple sub-entry according as they are gation and chunk-semantic interactions wholly or partially determined. The latter con- by comparison with the appropriate inter- stitute separate headings and are placed in action entries in the Base-Nude dictionary. closed commutation systems which differ from (4) Repetition of the above stages, using the those set up for e.g. nouns in that they are in Nude-Target mechanical dictionary. The the first instance grammatically, not contextu- potentialities of this method are to be a lly, restricted. illustrated by translation from a Japanese passage into English, German, Latin and Welsh. General Program for Mechanical Translation between Any Two Languages via an Algebraic Interlingua. (R.H. Richens) L ist of Publications of the C.L.R.U. I t has become clear that the amount of 1. Progress Report I (January, 1953), obtain- lexical and syntactical analysis required to able cyclostyled from C. L. R. U., 20 Milling- produce a smooth and idiomatic mechanical ton Road, Cambridge, England. translation from any base language into any 2. Progress Report II, MT Vol. 3, No.l. target language is very great. It is interesting, 3. Annexe V to same is obtainable cyclostyled therefore, to examine the possibilities of me- from Editors (MT). chanical translation via a notational interlingua. 4. The Potentialities of a Mechanical Thesaurus With this approach, only one program is en- by Margaret Masterman. visaged for translation between any two lan- 5. A General Program for Mechanical Trans- guages, with the addition of specific mechanical dictionaries for each input and output language. lation between any Two Languages via an Algebraic Interlingua, by R.H. Richens. The notational interlingua being studied is 6. The Linguistic Basis of the Thesaurus-Type ideographic and constructed so as to represent Mechanical Dictionary and its Application to the ideas of any base passage divested of all English Preposition Classification, by M. A. lexical and syntactical peculiarities; for which K. Halliday. reason it is called Nude. The words in Nude are constructed of some fifty elements (Roman 7. An Algebraic Thesaurus, by A.F. Parker- letters, capitals and lower case letters being Rhodes . regarded as different symbols), each of which