This report is divided in different parts to explain what the MT (Machine Translation) is and its main problems. The explanation about Machine Translation will be shown in the Introduction, later, I will focus the report on a especial theme, I mean, this report looks at some problems which face the builder of MT systems. We characterized them as problems of ambiguity (lexical and syntactic) and problems of lexical and structural mismatches. We can see too the problem of non-compositionality (as exemplified by idioms and collocations). All these different problems are structured in points which are headed up by questions,thus the reader will be better placed in the text.
INTODUCTION TO MT
Machine Translation and Computer-Assisted Translation
Machine Translation (MT) has been the great hope and the great disappointment of CL(Computational Linguistics). Taking into account the linguistic and technical challenges and difficulties of the task - translating is without any doubt one of the most complex linguistic processes conceivable, involving large amounts of data and highly complex mental operations that are by no means well-known and understood - it is surprising that MT should have been one of the earliest attempted applications, actually the first non-numerical application of electronic data processing. From the middle of the 1940s information theorists tried to tackle the problem of MT on a mathematical basis as a cryptographic or, more generally, statistical problem. They viewed the translation of text from one language to another as a computable transformation from one method of encoding corresponding information into a different one.
As this approach did not work out linguists entered the scene. They first tried literal word for word substitution, adding some supplementary rearrangement rules. Further development saw successive extensions and refinements of the rules, until syntactic surface description was complemented by the analysis of underlying logical and semantic structures. Currently available commercial systems running on personal computers still use this linguistic technology.
But, in fact, not all information that is necessary to correctly translate the content (let alone the stylistic) features of a text is explicitly encoded in lexical and syntactic structures. Human translators also refer to implicit (unexpressed) linguistic and extra-linguistic knowledge (cotextual knowledge and contextual, so-called "world knowledge"). The problem is how to represent and organise this implicit information in order to make all and only the relevant data available to an MT system.
Trying to solve this and other problems, MT developers have sought on the one hand to achieve further improvements on the linguistic level, and on the other hand to take advantage of methods and achievements of artificial intelligence and "knowledge processing". These efforts are referred to as "third generation MT" - after a first, information theory-based, and a second (computational) linguistics-based generation. In parallel, quantitative methods have emerged which, based on large parallel corpora of existing translations, rely on probabilities of interlingual correspondence or on (partial) literal and structural matches. Currently, efforts in CL concentrate on the (often comparative) evaluation of existing MT systems, emphasising a differentiating and pragmatic approach to translation quality and usability in function of parameters like, among others, text types, text categories or quality and cost requirements.
Quantitative methods have lead one of the most interesting and promising developments in the field of computer-assisted translation, viz. translation memories. These use archive databases containing parallel sentence for sentence versions of previously translated texts. Input (source) sentences which are identical or very similar to sentences in the archive then do not have to be reanalysed and retranslated. Instead, their translation can be used again, if necessary with some modifications. The strength of such programs lies in retrieving and intelligently handling less than 100% matches of new input sentences with archived previously translated ones. They are particularly useful and effective for translating new versions of, for instance, users' guides and operation manuals, where large oarts of the text may be not at all or only slightly altered in comparison to former versions. Translation memories do not translate automatically, i.e. without any human intervention during the translation process proper, but interactively, displaying potential matches of input sentences from their archives with their translation, including possible differences. It is up to the user in each case to decide whether to accept, modify or reject the solution offered. It should be noted that there are some interactive MT systems, too. And there are some commercially available systems with integrated translation memories now.
© Nico Weber 1998/2000 – Last Update: Nov. 2002
After having read the introduction that have expalined what Machine Translation is and sometheings about MT's background and-we will go on with the presentation of the main preoblems of MT, examples and some ideas to solve them:
1. Which are the main problems of MT?
We will consider some particular problems which the task of translation poses for the builder of MT systems --- some of the reasons why MT is hard. It is useful to think of these problems under two headings: (i) Problems of ambiguity , (ii) problems that arise from structural and lexical differences between languages and (iii) multiword units like idiom s and collocations .Of course, these sorts of problems (ambiguity)are not the only reasons why MT is hard. Other problems include the sheer size of the undertaking, as indicated by the number of rules and dictionary entries that a realistic system will need, and the fact that there are many constructions whose grammar is poorly understood, in the sense that it is not clear how they should be represented, or what rules should be used to describe them. This is the case even for English, which has been extensively studied, and for which there are detailed descriptions -- both traditional `descriptive' and theoretically sophisticated -- some of which are written with computational usability in mind. It is an even worse problem for other languages. Moreover, even where there is a reasonable description of a phenomenon or construction, producing a description which is sufficiently precise to be used by an automatic system raises non-trivial problems.
Thu Dec 21 10:52:49 GMT 1995
2. Select two examples of lexical ambiguity and select one example of structural ambiguity(due to the similarity of these themes I will explain them in the same question or point)
Ambiguity is divided in syntactically ambiguity problems and lexically ambiguity problems.When a word has more than one meaning, it is said to be lexically ambiguous. However, when a phrase or sentence can have more than one structure it is said to be structurally ambiguous. Ambiguity is a pervasive phenomenon in human languages. It is very hard to find words that are not at least two ways ambiguous, and sentences which are (out of context) several ways ambiguous are the rule, not the exception. This is not only problematic because some of the alternatives are unintended (i.e. represent wrong interpretations), but because ambiguities 'multiply'. If we want to see some examples of these problems of lexically ambiguous, which is less complex than the second case which is structurally ambiguous, to understand them better we could put some examples like these:
Lexical ambiguity is by far the more common. Everyday examples include nouns like 'chip', 'pen' and 'suit', verbs like 'call', 'draw' and 'run', and adjectives like 'deep', 'dry' and 'hard'. There are various tests for ambiguity. One test is having two unrelated antonyms, as with 'hard', which has both 'soft' and 'easy' as opposites. Another is the conjunction reduction test. Consider the sentence, 'The tailor pressed one suit in his shop and one in the municipal court'. Evidence that the word 'suit' (not to mention 'press') is ambiguous is provided by the anomaly of the 'crossed interpretation' of the sentence, on which 'suit' is used to refer to an article of clothing and 'one' to a legal action.
The above examples of ambiguity are each a case of one word with more than one meaning. However, it is not always clear when we have only one word. The verb 'desert' and the noun 'dessert', which sound the same but are spelled differently, count as distinct words (they are homonyms). So do the noun 'bear' and the verb 'bear', even though they not only sound the same but are spelled the same. These examples may be clear cases of homonymy, but what about the noun 'respect' and the verb 'respect' or the preposition 'over' and the adjective 'over'? Are the members of these pairs homonyms or different forms of the same word? There is no general consensus on how to draw the line between cases of one ambiguous word and cases of two homonyous words. Perhaps the difference is ultimately arbitrary.
Sometimes one meaning of a word is derived from another. For example, the cognitive sense of 'see' seems derived from its visual sense. The sense of 'weigh' in 'He weighed the package' is derived from its sense in 'The package weighed two pounds'. Similarly, the transitive senses of 'burn', 'fly' and 'walk' are derived from their intransitive senses. Now it could be argued that in each of these cases the derived sense does not really qualify as a second meaning of the word but is actually the result of a lexical operation on the underived sense. This argument is plausible to the extent that the phenomenon is systematic and general, rather than peculiar to particular words. Lexical semantics has the task of identifying and characterizing such systematic phemena. It is also concerned to explain the rich and subtle semantic behavior of common and highly flexible words like the verbs 'do' and 'put' and the prepositions 'at', 'in' and 'to'. Each of these words has uses which are so numerous yet so closely related that they are often described as 'polysemous' rather than ambiguous.
Structural ambiguity occurs when a phrase or sentence has more than one underlying structure, such as the phrases 'Tibetan history teacher', 'a student of high moral principles' and 'short men and women', and the sentences 'The girl hit the boy with a book' and 'Visiting relatives can be boring'. These ambiguities are said to be structural because each such phrase can be represented in two structurally different ways, e.g., '[Tibetan history] teacher' and 'Tibetan [history teacher]'. Indeed, the existence of such ambiguities provides strong evidence for a level of underlying syntactic structure. Consider the structurally ambiguous sentence, 'The chicken is ready to eat', which could be used to describe either a hungry chicken or a broiled chicken. It is arguable that the operative reading depends on whether or not the implicit subject of the infinitive clause 'to eat' is tied anaphorically to the subject ('the chicken') of the main clause.
It is not always clear when we have a case of structural ambiguity. Consider, for example, the elliptical sentence, 'Perot knows a richer man than Trump'. It has two meanings, that Perot knows a man who is richer than Trump and that Perot knows man who is richer than any man Trump knows, and is therefore ambiguous. But what about the sentence 'John loves his mother and so does Bill'? It can be used to say either that John loves John's mother and Bill loves Bill's mother or that John loves John's mother and Bill loves John's mother. But is it really ambiguous? One might argue that the clause 'so does Bill' is unambiguous and may be read unequivocally as saying in the context that Bill does the same thing that John does, and although there are two different possibilities for what counts as doing the same thing, these alternatives are not fixed semantically. Hence the ambiguity is merely apparent and better described as semantic underdetermination.
Although ambiguity is fundamentally a property of linguistic expressions, people are also said to be ambiguous on occasion in how they use language. This can occur if, even when their words are unambiguous, their words do not make what they mean uniquely determinable. Strictly speaking, however, ambiguity is a semantic phenomenon, involving linguistic meaning rather than speaker meaning; 'pragmatic ambiguity' is an oxymoron. Generally when one uses ambiguous words or sentences, one does not consciously entertain their unintended meanings, although there is psycholinguistic evidence that when one hears ambiguous words one momentarily accesses and then rules out their irrelevant senses. When people use ambiguous language, generally its ambiguity is not intended. Occasionally, however, ambiguity is deliberate, as with an utterance of 'I'd like to see more of you' when intended to be taken in more than one way in the very same context of utterance.
Kent Bach, Routledge Encyclopedia of Philosophy entry
3. Lexical and Structural Mismatches
At the start of the previous section we said that, in the best of all possible worlds for NLP, every word would have exactly one sense. While this is true for most NLP, it is an exaggeration as regards MT. It would be a better world, but not the best of all possible worlds, because we would still be faced with difficult translation problems. Some of these problems are to do with lexical differences between languages --- differences in the ways in which languages seem to classify the world, what concepts they choose to express by single words, and which they choose not to lexicalize. We will look at some of these directly. Other problems arise because different languages use different structures for the same purpose, and the same structure for different purposes. In either case, the result is that we have to complicate the translation process. In this section we will look at some representative examples.
Examples like the ones in (a) below are familiar to translators, but the examples of colours (c), and the Japanese examples in (d) are particularly striking. The latter because they show how languages need differ not only with respect to the fineness or `granularity' of the distinctions they make, but also with respect to the basis for the distinction: English chooses different verbs for the action/event of putting on, and the action/state of wearing. Japanese does not make this distinction, but differentiates according to the object that is worn. In the case of English to Japanese, a fairly simple test on the semantics of the NPs that accompany a verb may be sufficient to decide on the right translation. Some of the colour examples are similar, but more generally, investigation of colour vocabulary indicates that languages actually carve up the spectrum in rather different ways, and that deciding on the best translation may require knowledge that goes well beyond what is in the text, and may even be undecidable. In this sense, the translation of colour terminology begins to resemble the translation of terms for cultural artifacts (e.g. words like English cottage, Russian dacha, French château, etc. for which no adequate translation exists, and for which the human translator must decide between straight borrowing, neologism, and providing an explanation). In this area, translation is a genuinely creative actfootnode.html - 4582footnode.html - 4582, which is well beyond the capacity of current computers.
Calling cases such as those above lexical mismatches is not controversial. However, when one turns to cases of structural mismatch, classification is not so easy. This is because one may often think that the reason one language uses one construction, where another uses another is because of the stock of lexical items the two languages have. Thus, the distinction is to some extent a matter of taste and convenience.
A particularly obvious example of this involves problems arising from what are sometimes called lexical holes --- that is, cases where one language has to use a phrase to express what another language expresses in a single word. Examples of this include the `hole' that exists in English with respect to French ignorer (`to not know', `to be ignorant of'), and se suicider (`to suicide', i.e. `to commit suicide', `to kill oneself'). The problems raised by such lexical holes have a certain similarity to those raised by idiom s: in both cases, one has phrases translating as single words. We will therefore postpone discussion of these until Sectionnode55.html - Idiomsnode55.html - Idioms.
One kind of structural mismatch occurs where two languages use the same construction for different purposes, or use different constructions for what appears to be the same purpose.
Cases where the same structure is used for different purposes include the use of passive constructions in English, and Japanese . In the example below, the Japanese particle wa, which we have glossed as `TOP' here marks the `topic' of the sentence --- intuitively, what the sentence is about.
Satoo-san wa shyushoo ni erabaremashita.
Satoo-hon TOP Prime Minister in was-elected
Mr. Satoh was elected Prime Minister.
Example indicates that Japanese has a passive-like construction, i.e. a construction where the PATIENT, which is normally realized as an OBJECT, is realized as SUBJECT. It is different from the English passive in the sense that in Japanese this construction tends to have an extra adversive nuance which might make a) rather odd, since it suggests an interpretation where Mr Satoh did not want to be elected, or where election is somehow bad for him. This is not suggested by the English translation, of course. The translation problem from Japanese to English is one of those that looks unsolvable for MT, though one might try to convey the intended sense by adding an adverb such as unfortunately. The translation problem from English to Japanese is on the other hand within the scope of MT, since one must just choose another form. This is possible, since Japanese allows SUBJECTs to be omitted freely, so one can say the equivalent of elected Mr Satoh, and thus avoid having to mention an AGENT. However, in general, the result of this is that one cannot have simple rules for passives. In fact, unless one uses a very abstract structure indeed, the rules will be rather complicated.
We can see more examples with the difference between languages:
Figure:venir-de and have-just
The first example shows how English, German and French choose different methods for expressing `naming'. The other two examples show one language using an adverbial ADJUNCT ( just, or graag(Dutch) `likingly' or `with pleasure'), where another uses a verbal construction. This is actually one of the most discussed problems in current MT, and it is worth examining why it is problematic. These representations are relatively abstract (e.g. the information about tense and aspect conveyed by the auxiliary verb have has been expressed in a feature) , but they are still rather different. In particular, notice that while the main verb of is see, the main verb of b) is venir-de. Now notice what is involved in writing rules which relate these structures (we will look at the direction English and French)
Thu Dec 21 10:52:49 GMT 1995
In conclusion, Machine Translation can contain some errors and problems as we have read before, I have to say that in the last years Computers has improved and in future wi will go on finding less problems .We have to say that language is very complex in order to have it transalted:for instance,when you ask a program to translate you the word "manzanilla" it can may be answer you with "little apple" (obviously ,this does not always happens,it only happens when the transaltor is not good enough or when it does not have high quality).So,Computers is not able to register a whole language.
I definetely have to admit that the work has helped me a lot to improve my knowledge in relation with this topic. But not only it will help me knowing more,but it will let me do my university works and reports more completely and in a more eficient way.
Gala Díez Pedrosa