#56: Research Paper: ‘Technology as an Alchemy of Language and Communication’ by Mansa Ram Maity

Technology as an Alchemy of Language and Communication: A Critical Estimation of Machine Translation

by

Mansa Ram Maity, Lecturer in English, Rajiv Gandhi University – IIIT, Basar

The century that has gone by, witnessed far-reaching changes in linguistics, Anthropology, literature and philosophy leading to the change in the earlier status of translation considered as  ‘a secondary activity’. With the explosion of information technology, with the ever-increasing interaction between languages and cultures bridging the yawning hiatus between them, and with comparative literature gaining importance because of inter-dependence of literature in different regions, the crucial role played by translation is striking new grounds for readers and researchers.

The unintelligibility brought about by the plethora of linguistic diversities vitiated the biblical tower of Babel. Translation is a metaphor of the spirit of the source language and is a metamorphosis of language. Translations have a marvelous way of liberating a text from the confines of its regional context and have helped to break the linguistic barriers and to promote research at the national and international levels. Learning and research become easier if they are done through one’s mother-tongue. In advanced countries like America, England, Germany, France, and Russia research findings are presented only in their respective mother-tongue. If the developing countries are to reap the benefits by these advanced research findings, they need translation of these records. In science and technology, translation helps to share knowledge and to avoid duplication of research among scientists. Translations help learners and researchers to develop self-confidence, self-reliance and freedom and to build up their own independent research findings.

Translations help to eschew parochialism and to promote national integration and international understanding. At the global level, translations of literary works help people to understand common problems and aspirations and help to grasp the genius of the language and the rich resources in it.

English literary studies in India are in state of great transition and formidable flux today. Soon, all of us would agree, we have to make fresh choices and evolve a new curriculum for ourselves. We realize that we have to relocate ourselves in a new socio-literary and academic context. Having taken note of this in time, many of us have already made our choices; others are in the process of doing so. The study of linguistics being boosted by the computational thinking has opened up new dimensions and created new arena for the gennext. The nations which have done this are geared to face the future. Others will slowly, but surely, have to follow suit. Our new literary on academic baggage is a curious mix of great variety. Yet, our curricular cartography is under pressure from newer areas awaiting inclusion.

The overall map of our literary reading is much more complex now than it was ever before. If has regional, national and cosmopolitan, references; it has sounds and echoes from all over the land. Literatures written originally in English, and those in translation, have started engaging greater attention today. A translated text is no longer the ‘other ‘text; it is a major component of our new text logy. This holds true much more appropriately in India than anywhere else. We need to identify and know our text(s), co-text(s) and context(s), as we need to evolve a way of reading. This rich and varied context gives us a reason, more than others, to read texts in translation, and to read them with a certain discipline. It mutually calls for theoretical interventions.

Several efforts have been made, more especially, in the recent past, to help read translations as discipline. Even though the Indian text and context have been under sharp focus, translation in India is yet to achieve a canonical status. The movement towards achieving certain acceptability is quick, and the academic proceeds, too, seem to be encouraging.

Now comes the question, what’s translation?

Translation is an anglicized form of a Latin word. In it ‘trans’ means ‘across’ and ‘latum’ means ‘to carry’. In literal terms, to translate means to make another language like one’s own, to preserve meaning and significances across vocabularies, grammars and syntaxes. In other words, it is an art of carrying across f the matters of one source language (SL) into a Target Language (TL)

Translation, involves the rendering of a Source language (SL) text into the Target Language (TL) so as to ensure that (1) the surface meaning of the two will be approximately similar and (2) the structures of the SL will be preserved as closely as possible but not so closely that the TL structures will be seriously distorted.

The language we translate is called Source Language, called for short, SL. The language into which we translate is called Target Language, called for short, TL, also known as Receptor Language. The original text is called Source Language Text (SLT). The translated text is called Target Language Text (TLT) on Receptor Language Text (RLT).  Translation is not, in the modern context, secondary to original literature. It has an independent existence of its own. It is not reproduction but recreation. Translation is neither a creative art nor an initiative art but stands somewhere between the two.

Translation should reflect accurately the meaning of the original text –Source Text. Nothing should be arbitrarily added on removed, though occasionally part of the meaning can be “transposed”. Translation, so far has been performed by human beings is done by the machines (computer) now-a-days. Though it does not have a brain of its own, both linguist and the programmer feed everything that are required for translation. Then it acts according to the algorithm designed by them.

Literacy through computers is now an important part of everyone’s education. The computer is now a common household appliance with uses ranging from recreation to household accounts to word processing. It has played a possible part of its own by translating books, film from one language on culture to another language or culture, In a computer for translation process, a single accumulator machine is arranged and its memory is sequentially organized into words. The growth in computing has been primarily in applications where numeric calculations play a very secondary role, Non-numeric applications such as word processing and electronic mail, have paved the way to the acceptance of the computer as a standard business on home commodity.

Means of Translation: There are four means of Translation:

1. Human Translation: Translation a performed by human beings some linguist, translation theoreticians.
2. Human Aided Machine Translation (HAMT): The rules of Humans in this system are editing the input text preparing the dictionary and writing the program according to the pre-edited input text. HAMT parses the sentences, refers dictionary, processes the noun, verbs and other categories of the sentences and records the structure, so as to suit the target language structure.
3. Machine Aided Human Translation(MAHT): Translation as performed by both the machine and human beings in conjunction. Computer is used as an auxiliary device to translate a text.
4. Pure Machine Translation(MT): In the system the source language(SL) text has to be fed into the computer as such. In this work the edition in the input text is mostly avoided except in linking compounded units with hyphen. In this type of translation, asset of rules called transfer rules are required. These rules convert the surface structure of one language. Also these rules operate on the surface structure of one language and derive the same of another language.

A general impression is prevailing over that soon machines will be doing all types of translating quickly, producing better performance than human beings. Of course, the machines (computers) must be properly given a straight forward algorithm by the specialists so that they can keep every instruction in their memory. In the beginning the machine translation (MT) was applied only to very brief passages – few sentences in a brief article, comprising subjects like nuclear physics, mathematical procedures or chemical procedures.

However, there are some limitations to Machine Translation at this point of time. Literary translation cannot be done meaningfully by the machines because of its limitations arid moreover, it is not possible to feed a machine with sufficient background data to resolve the numerous formal and semantic problems preserving the original meaning and retaining the syntactic structure of the original, reflecting the thought process, considering form and content, and fusing the manner and matter of the original.

The Analysis of Machine Translation (MT):

Major parts in a computer:

There are three major parts in a computer. They are

1. The input
2. The storage and processor
3. The output

i. The input: The computer needs input in order to get the desired result. The input is nothing but source information required for the computational process. It is the data given by the user-the stored data, for example, the dictionaries of two languages.

ii. The storage and the processor:

The storage : The  computer needs a set of pre-defined rules to perform the operation which is called processing. The instructions are stored in the storage.

The processor: The processor analyses the data to be given as output. The processing of the text includes procedures as identification of words and meaningful combination of word and so on.

iii     The output: The translated text is the output.

Function of a computer Machine in Translation:

The machine knows only ‘0’ and ‘1’. They are called binary codes.

1. The data given by the user is first of all converted into binary code numerals. The changed alphabets into binary codes are fed into the computer. For example, the alphabets are symbolized as follows:

A00001

B00010

C00011

Doo1oo etc

ii                           The binary codes from one language are converted into the binary codes in another language.

Binary code ————————————-Binary Code

(Source Language)                                           (Receptor Language)

Iii            The binary codes of Receptor Language are converted into Receptor Language.

Binary code (Receptor Language) —————-Receptor Language

The major parts in a computing machine, such as the input, the main storage and processor, and the output, do the processing work. For example, the electrical impulses on a tape must be reproducible as holes in a card, on positive and negative changes on a disc, or photo sensitive contacts on a photo scope plate.  These impulses can be translatable back into letters and later into words allowing the material to be read. This process of intelligible translation, with the help of the linguist and the programmer, is done by making a series of simple comparisons and calculations and sending the results to an output component which makes out the checks.

Procedure of Machine Translation:

A bilingual dictionary is fed into the machine. It is the stored data of source Language and Receptor Language. An MT dictionary transforms the input of words in on language (SL) into an output of words into another language(RL).

Processing takes place in three stages:

1. Analysis
2. Comparison
3. Synthesis

In the analysis stage, the semantic units are identified; the structural patterns are analyzed stored. In the comparison stage, both the Source Language text and the Receptor Language text re compared. Comparison involves the storing of the data and the instructions for matching specifies.

In the synthesis stage, MT looks for the semantic and structural specifiers in the Receptor Language and generates appropriate sentence with closest approximation.

Language Structure in MT:

(Machine Translation involves translating) a structures, not a string of words. It must identify the language structure, the borders of constructions and the key words around which, the parts of the sentence may be constructed. There are two methods of treating language structure:

1. Phrase structure Approach and
2. Predictive Method
3. i. Phrase Structure Approach

The                 Man                went            home           The             Man           went        home

Article àNounàVerbàNoun

A              N           V         N

The man went home (making the subject and predicate depend upon the verb)

The machine, according to the instructions fed, acts and identifies such units –A,N,V,N and related them. With the data stored in the memory the machine can pass, from left to right or from smaller to bigger units on the reverse order to reach successively a more detailed comprehensive structure. This is known as “pass procedure”.

II. Predictive Method:

This method closely represents the mathematician’s view of the language structure expectations of what is to follow an initial word. A permanent memory of this type of sequences predicts what follows the initial word, analyses the meaningful combinations of words and sets up the necessary specifies so that the corresponding features of the Receptor language might be produced.

Whatever method one uses to identify the grammatical combinations of words, there are two types of information which may be taken into account.

1. The degree of cohesiveness between words.
2. The types of , dependencies involved in terms of agreement (e.g. gender and number), government (e.g. case endings required by certain grammatical positions), and cross reference.

Functions in Machine Translation:

The machine must be provided a bilingual dictionary through which the input of words in one language can be transformed into an output of words in another language. It is a fact that natural language is made understandable to the computer; certainly one can have a sigh of relief with the fifth generation computers in his hands. So the facts of the natural language must be collected and have to be processed for inclusion in a general system. In this system the SL resource language text has to be fed into the computer as such. The sentences in a language can be classified according to the nature of the verbs they take. The verbs denoting “psychological state”(know, understand etc.) and motion( go, come etc.) take the constituents experience and patient, and the arguments agent, source, goal, path and means respectively.  A dictionary called preposition identifier is prepared from which the computer predicts the successive nouns as goal, source, path, means respectively and takes the equivalents from the general dictionary where marked constituents are entered. As a result, the computer can perform particular set of operations from among the different types already given. Then in the post-edition, the nouns and verbs re processed.

The sentence in the Source Language is identified and maintained in the following order:

SL Text (English): Agent +verb +Goal+Adverb+Time+Source+Path+Means.

SLT: He went to Chennai yesterday from trichy through Tanjore by train.

SLT Analysis:

He (Agent), went (verb), to Chennai (Goal), yesterday (Adverbial of Time), from Trichy (source), through Tanjore (path), by train (means).

TLT:

वह कल त्रिची से चेन्नई, तानज़ोर होते हुए ट्रेन के  ज़रिए गया.

TLT Analysis:

वह (Agent),  कल (Adverbial of time), त्रिची से (source), चेन्नई (Goal),

तानज़ोर होते हुए (Path), ट्रेन के ज़रिए(means), गया (verb).

Various phases of MT:

Original textàpre-edited text à word by word substitution of the target languageàrearranged structure àoutputàpost editing àinputànormal word orderànatural target language sentences.

The input fed into the computer is modified to facilitate the process of MT. Success in MT therefore, depends on both the analysis of language and the data available with the machine. A bilingual dictionary was created with necessary grammatical information. Each lexical item entered in the dictionary has not only equivalent item in the target language but also the grammatical rules of both the source language and target languages are fed into computer to translate and generate grammatical sentences in the target language.

Both pre-editing and post-editing are done to give minimum rules as instructions to the memory component of the machine. Since the machine we are working with a limited memory power we are forced to involve in the process of pre-editing in our translations.

Basic Limitations and Word Order in MT:

To translate the structure of language into a form that the machine can handle, there is a need for an intermediate technical language which can be automatically adopted to the special limitations of the machine. It is convenient to use a kind of special built in language. For example , a detail computer order bust be given for a simple linguistic commands as such.

“Identify the noun in the following sentence.”

Some of the special languages for computers are comity Fortran and Mimic which are making the correspondences between linguistic procedures and machine processes.

Word order and choosing of the correct alternative are the main problems in MT.

For example, the Spanish word ‘radio’ has its English alternative as follows:

Radius (in geometry or anatomy)

Radium (in chemistry)

The only way by which the machine can determine the correct alternative is to identify certain diagnostic words in the context which point to the topic of discourse. It is not at all an easy task to set up an adequate dictionary for MT , since the dictionary needs careful classification of a form and listing of various types of words, so that the proper equivalent may be selected.

Likewise,  one- to- one correspondence and many –to-one correspondence, may not prose serious problems in the process of Machine Translation. But while doing one-to-many, depending upon the usage of subject works he, she, it, and they poses certain problems. And in word order, there may be differences.

It is possible to do effective translation through computers only after a systematic comparison of the two languages involved. Even though most of the translations believed that the translation of the creative literature is not immediately possible, a systematic an time bound language analysis using the advanced computer technologies would give fruitful results in the field of machine translation.

Change is inevitable in all cases, and so with the story study of language and linguistics in the dimension of translatology. There is always a room for improvement with research on artificial intelligence, the development of language and linguistics in the dimension of translatology  will be  a boost and will refine the output.