{ "paper_id": "C86-1023", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T13:15:30.922887Z" }, "title": "Idiosyncratic Gap: A Tough Prolem to Structure-bound Machine Translation", "authors": [ { "first": "Yoshihiko", "middle": [], "last": "Nitre", "suffix": "", "affiliation": { "laboratory": "Advanced Research Laboratory", "institution": "Hitachi Ltd. Kokubunji", "location": { "postCode": "185", "settlement": "Tokyo", "country": "Japan" } }, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "Current practical machine translation systems (MT, in short), which are designed to deal with a huge amount of document, are general]y structure-bound. That is, the translation process is done based on tile", "pdf_parse": { "paper_id": "C86-1023", "_pdf_hash": "", "abstract": [ { "text": "Current practical machine translation systems (MT, in short), which are designed to deal with a huge amount of document, are general]y structure-bound. That is, the translation process is done based on tile", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "analysis and transformation of the structure of source sentence, not on the understanding and paraphrasing of the meaning of that. But each language has its own :~yntactic and semantic idiosyncrasy, and on this account, without understanding the total meaning of source sentences it is often difficult for MT to bridge properly the idiosyncratic gap between source~ and target-language.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "A somewhat new method called \"Cross Translation Test (CTT, in short)\" is presented that reveals the detail of idiosyncratic gap (IG, in short) together with the so-so satisfiable possibility of MT. It is also mentioned the usefulness olf sublanguage approach to reducing the IC between source-and target-language.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "The majoJ:[ty of the current practical machine translation system (MT, in short) (See [Nagao 1985] and [Slocum 11.985 ] for a good survey.) are structurebound in tile sense that al] the target sentences (i.e. translated ,~entences) are composed only from the syntactic st:ructure of the source sentences, not from the meaning understanding of those. Though almost all tile MT are utilizing some semantic devices such as semantic feature agreement checkers, semantic filters antl preference semantics (See [Wilks 1975] for example.) which are serving as syntactic structural disambiguation, they still remain Jn structurehound approaches far from tile total[ meaning understanding approaches. ]?he structure-bound MT has a lot of advantageous features among which the easiness of formalizing translation process, that is, writing translation rules and the uniformity of lexicon description are vital from the practical standpoint that it must transact a huge vocabulary and ]numerable kinds of sentence patterns.", "cite_spans": [ { "start": 86, "end": 98, "text": "[Nagao 1985]", "ref_id": "BIBREF1" }, { "start": 103, "end": 117, "text": "[Slocum 11.985", "ref_id": null }, { "start": 505, "end": 517, "text": "[Wilks 1975]", "ref_id": "BIBREF9" } ], "ref_spans": [], "eq_spans": [], "section": "i. Introduct:ion", "sec_num": null }, { "text": "On the other hand, the structure-bound MT has the inewttable limitation on the treatment of lingu:istic idiosyncrasy originated from the different way of\" thinking.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Introduct:ion", "sec_num": null }, { "text": "In this paper, first of all, we will sketch out the typical language modeling techniques on which the structure-bound MT(= current practical machine translation systems) are constructed.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Introduct:ion", "sec_num": null }, { "text": "Secondly, we will examine the difference between the principal mechanism of machine translation and that of human translation irom the viewpoint of the language understanding abi]ity, l'hirdly, we will illustrate the structural idiosyncratic gap (IG, in short) by comparing the sample sentences in English and that in ,lapanese. These sentences are sharing the same reCalling. This comparison will be made by a somewhat new method which we call \"Cross Translation Test (CTT, in short)\", which will cventual]y reveal the various IGs that have origins in the differences of culture, i.e., the way of thinking or the way of representing concepts.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Introduct:ion", "sec_num": null }, { "text": "But at\" the same Lime, CTT wiJl give some encouraging evidence that the principal technologies of today's not-yet-completed structure-bound HTs have the potentia] for producing barely acceptable translation, if the source language sentences are taken from tile documents of less equivocations or are appropriately rewritten. Finally, we will briefly comment on the sub]anguage to control or normalize source sentences as the promising and practical approaches to overcoming the IGs.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "i. Introduct:ion", "sec_num": null }, { "text": "Modeling natural, language sentences is, needless to say, very essential to all kinds of natural language processing systems inclusive of machine translation systems. The aim of mode]ing :is to reduce the superficia] complexity and variety of the sentence form, so as to reveal the indwell:Lug structure which is indispensable for computer systems to analyze, to transform or to generate sententia] .representations.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "So far various modeling techniques are proposed (See for example [Winograd 1.983] .) among which the two, tile dependency structure modeling (Figure l) and the phrase structure modeling ( Figure 2 ) are important. The former associated with semantic cole labeling such as case marker assignment is indispensable to analyze and generate Japanese sentence strueture (See for example ]Nit,a, et all. 1984].), and the latter associated with syntactic rote labeling such as governor-dependent assignment, head-complement assignment, or mother-daughter assignment (See for example [Nitta, et el. 1982 \"To what extent should (or can) we treat semantics of sentences?\" is also very crucial to the decision for selecting nr designing tile linguistic model for machine translation.", "cite_spans": [ { "start": 65, "end": 81, "text": "[Winograd 1.983]", "ref_id": null }, { "start": 575, "end": 594, "text": "[Nitta, et el. 1982", "ref_id": null } ], "ref_spans": [ { "start": 188, "end": 196, "text": "Figure 2", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "But it might be fairly asserted that the majority of the current \"practical\" machine translation systems (MT, in short)are structure-bound or syntax-oriented, though almost all of them claim that they are semantics-directed.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "Semantics are used only for disambiguation and booster in various syntactic processes, but not used for the central engine for transformation, generation and of course not for paragraph understanding (See [Slocum 1985, pp. 14 ~16] for a good survey and discussion on this problem; and see also [Nitta, et al. 1982] for the discussion on a typical (classical) structure-bound translation mechanism,i.e, local rearrangement method).", "cite_spans": [ { "start": 205, "end": 222, "text": "[Slocum 1985, pp.", "ref_id": null }, { "start": 294, "end": 314, "text": "[Nitta, et al. 1982]", "ref_id": "BIBREF4" } ], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "Here \"practical\" means \"of very large scale commercial systems\" or \"of the daily usage by open users\", but neither \"of small scale laboratory systems\" nor \"of the theory-oriented experimental systems\".", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "For structure-bound machine translation systems, both the dependency structure modeling and the phrase structure modeling are very fundamental technical tools.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "\u00b0 [Lit. Co) ~1~", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "FHi~a) &l\u00a3-~/-lk\u00a29 I~ ~/l~tz ~.~C~,~To.] (J'l) Kono kusuri-wa itsft no ue-nl subayai kikime-wo mottedm. The semantic network medeling, which is recently regarded as an essential tool for semantic processing for natural languages (See for examples [SimmOns 1984] .), might also be viewed as a variation of dependency modeling.", "cite_spans": [ { "start": 247, "end": 261, "text": "[SimmOns 1984]", "ref_id": "BIBREF7" } ], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "However modeling problems are not discussed further here.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "Comparing Figure 1 and Figure 2 , note that the dependency structure modeling is more semantics-oriented, logical and abstract, in the sense of having some distance from surface word sequences.", "cite_spans": [], "ref_spans": [ { "start": 10, "end": 18, "text": "Figure 1", "ref_id": null }, { "start": 23, "end": 31, "text": "Figure 2", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "Modelin~ of Natural Lan~", "sec_num": "2." }, { "text": "Today's practical machine translation systems (MT, in short) (See for example [Nagao 1985] and [Slocum 1985] .) are essentially structure-bound ]iteral type. The reasons for this somewhat extreme judgement are as follows:", "cite_spans": [ { "start": 78, "end": 90, "text": "[Nagao 1985]", "ref_id": "BIBREF1" }, { "start": 95, "end": 108, "text": "[Slocum 1985]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Machine Translation vs. lluman Translation", "sec_num": "3." }, { "text": "(i) The process of MT is always under the strong control of the structural information extracted from source sentences; (2) In all the target sentences produced by MT, we can easily detect the traces of wording and phrasing of the source sentences; (3) MT is quite indifferent to whether or not the output translation is preserving the proper meaning of the original sentence, and what is worse, MT is incapable of judging whether or not;", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Machine Translation vs. lluman Translation", "sec_num": "3." }, { "text": "(4) MT is quite poor at the extra-sentential information such as situational information, world knowledge and common sense which give a very powerful command of language comprehension. Now let us see Figure 3 . This rather oversimplified figure illustrates the typical process of Japanese-English structure-bound machine translation. Here the analysis and transformation phase are based on the dependency structure modeling (cf. Figure i ) and the generation phase is based on the phrase structure modeling (cf. Figure 2 ) (For further details, see for example [Nitta, et al. 1984] .). This figure reveals that all the process is bound by the grammatical structure of the source sentence, but not by the meaning of that.", "cite_spans": [ { "start": 561, "end": 581, "text": "[Nitta, et al. 1984]", "ref_id": null } ], "ref_spans": [ { "start": 200, "end": 208, "text": "Figure 3", "ref_id": null }, { "start": 429, "end": 437, "text": "Figure i", "ref_id": null }, { "start": 512, "end": 520, "text": "Figure 2", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "Machine Translation vs. lluman Translation", "sec_num": "3." }, { "text": "Kono kusuri-wa itsu-ni sugu kiku. (El): This medicine has an immediate effect on stomachache.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Source Sentence:", "sec_num": null }, { "text": "Thus, the MT can easily perform the literal syntax-directed translation such as 'from (Jl) into (E'I)' (cf. Figure i) . But it is very very difficult for MT to produce natural translation which reflects the idiosyncrasy of target language) pre-serving the original meaning.", "cite_spans": [], "ref_spans": [ { "start": 108, "end": 117, "text": "Figure i)", "ref_id": null } ], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "(El) is an example of a natural translation of (J1).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "In order for MT to produce this (El) from (Jl), it may have to invoke a somewhat sophisticated heuristic rule.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "In Figure 3 , the heuristic rule, HR (KK, KS, ...), can sucessfully indicate the change of predicate which may improve the treatment for the idiosyncrasy of target sentence.", "cite_spans": [], "ref_spans": [ { "start": 3, "end": 12, "text": "Figure 3", "ref_id": null } ], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "But generally. , the treatment of idiosyncratic gap (IG, in short) such as 'that between (Jl) and (El)' is very difficult for MT. It might' be almost impossible to find universal grarmnatical rules to manipulate this kind of gaps, .and what is worse, the appropriake heuristic rules are not always found successfully.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "On the other hand, tile human translation (HT, in short) is essentially semantics-oriented type or meaning understanding type. 3?he reasons for this judgement are as follows:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "(i) HT is free from the structure, wording and phrasing of a source sentence; (2) liT can \"create\" (rather than \"translate\") freely a target sentence from something like an image (Hagram obtained froln a source sentence ( The reason for comparing tile two sentences is that we cannot: examine the linguistic idiosyncrasy itself.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "Because, currently, we cannot fix the one abstract neutral meaning without using something like the image diagram (cf. Figure 4 ) which is not yet elucidated.", "cite_spans": [], "ref_spans": [ { "start": 119, "end": 128, "text": "Figure 4", "ref_id": null } ], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "In order to examine tlm idiosyncratic gap, we have devised the practical method named \"Cross Translation Test (CTT, in short).\"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "The outline of CTT is as follows: ]first, take an appropriate well-written sample sentence written in one language, say English; Let E denote this sample sentence; Secondly, select or make the proper free translation of E in the other ].anguage, say Japanese; ],et J denote this proper free translation; J must preserve tile orlginal meaning of E properly; At the same time, make a literal Lrans].tion of F, in the same language that ,] is written in; Let J' denote this literal translation; Lastly, make a Literal translation of J in the same language.that E is written in; Let E' denote tM.s literal translation.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "iIere, the \"literal\" translation means the translation that :[.s preserving the wording, phras:ing and various senteatial structure of the original (source) sentence as much as possible.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "Then, eventually we may be able to define (a~d examine) the idiosyncratic gap, ].C, by Figure 5 .", "cite_spans": [], "ref_spans": [ { "start": 87, "end": 95, "text": "Figure 5", "ref_id": null } ], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "Iu. other words, we may be able to exam:ine and grasp tile i.di.osyncratie gap by comparl.ng ]:he structure of 1)' , and that ot! 1,',', or by comparing that of ji and that of J.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "[n order {:o s:imp]ify the arguments, let us assume Lhat some kind of diagram is to be :[nvokezd item ]:he understaadiing of the original scntence.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "'I/his d:iagram may (or should) be completely free from the sopcrficta] struclnre such as wording, phras~nF,, subjectobject relation and so on, and may be strengthened and rood:irked l)y varleus exCra--]iuguistlc knowledge. ] t may be early for hnmalTl to compose the sentetlces such as (J2) arm (E2) from tlu{s kind of :]aaage din-gram .invoked from (J1).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Figure 3. Simplified Sketch of Machine Translation Process", "sec_num": null }, { "text": "But the sentences suitst~ (.....)sugu (.....)[medicine][stomachache][immediately]kono [this]Transformation --[\"Maybe some heuristic l'ule, HR (KK, KS, ...-) suggests ]the change m the predicate-]L argument relation.3motsu (,...)[have lkusuri (_...)itsu (.....)k6ka (...-)[medicinel[stomachache][effect]M!M lkonosugu (...,.)[this][immediate]~ GenelationPhrase Structure Formation:Target Sentence:", "type_str": "table", "text": "(TNS: PRESENT ....... SEM: KK, .....) [take effect] kusuri ( ...... SEM: KS, .....)", "num": null, "html": null }, "TABREF2": { "content": "
o ~o)~.'&f,.ktT~'\u00a2IcOf, ft'~/~ sJ \u00a2\u00a3tl./a.,(J2)
Ko,,oktlsu it.we[iomll.to1 lio-ita,,i,.gasugutore-ru.
[this]~nlellicme I [if(you) take I[stomadlacne] [soon /[delmved l
o [l.it. If you Iake this medicine :,ou will sonn be dep[ived of a stomachache ]# (E'2)
* This medicine will so.n cure you ol lhe stmnachache.(E2]
o [l,a. c_v] Kono kusun-wa :ma{a.wo saga-hi )~la &(~t:{-J \u00a21c~'It,llb~G ~;)t~4~ d ilsu-kara suku/i d~ro.#(r2)
{tills] [medicine][>m:][so(ml [of tile stonlache] [will cure]
Now, note that there are b:[g structural, gaps
between (gl) and (],;]), and between (,]2) and (E2),
whic.h are tile natura] ref]eetJons o] ]:ingu:istic
idiosyncrasy orginated in thc culture, i.c, lhe
difference of the way of thJnki.ng. So far we have
seen that MT is poor at tile idiesyncrasy treatment
", "type_str": "table", "text": "huma[] unc[(lr tile nolTnla] eond].tlons.", "num": null, "html": null }, "TABREF4": { "content": "
\u2022 tiemay]lave savedfi~eflightfrom a tragic
[kare] [kamo-~hire-nail[ kyfljo-shi-ta ] [sono] [teiki-bin] [kara] [higeki-teki)
repeat [hanpukul I jikk6 I [nol performance of tile American Airlines DC-IO crash that Killed [tsuirakul [l~oroshi-tal 1275 nin-nol 275(Ed)
peoplein Chicago in 1979.
[hito-bitol [Cllicagc-de] [1979 hen-nil
Kare~'a sono teiki-bin-',~o, 1979 nen-m Chicago-de 275 nin-no hito-bito-wo koroshi-ta
#(J'4)
American-K6ktl-no DC-I(]-no tsuiraku-no higeki-teki hanpuku-no jikk6-kara
$~J~btc9)6 b~tX~0
kyfljo-shi-ta kamo-shirenai.
kore-ni-yotte kono ki-wa, shisha 275 m\u00a2i-wo dashi-ta 1979 nen-no Chicago-kf~k6-de-no(g4)
tsuintku-jiko-no higeki-no ni-no-mai-wo sake-eta-to ie-y6.
this airplanecould eseapefrom
[to-ie-you][kore-m-yottel Ikono hikouki] [sake-etal [kara[
tragic [higeki-tekil [hanpuku, ni-no~mail [nol Itsuirakul Ijikol [nol repetition of crash accident of American Airlinesa(E'4)
DC-10 in Chicago Airport in 1979that produced 275 dead persons.
[Chicago-KflkO-de-no][ 1979 nen-nil [daslfi-tal[silishal
\u2022 The soldiers fired at the women and we saw several of them fail.
Heishi*taehi-ws on-na.tachi-nihappo-shi-ta soshite
[soldiers][at the womanl[firedl[and]
wareware-wakare-ra-nos~nin~gataoreru-no-wo
[we][of them][several]If all]
", "type_str": "table", "text": "]. Lit. It may safely be said that. by this,", "num": null, "html": null } } } }