COMPUTING The use of an electronic device that accepts data, performs mathematical and logical operations at speed on those data, and displays the results. Computers, although initially developed as calculating devices and open to a range of uses, have become central to communicative technology, and relate to language in at least three ways: (1) They require their own artificial languages in order to function. (2) Their use has adapted natural language to new ends, such as the processing of texts by computer. (3) Their users have developed their own styles and registers for working with them and talking about them. Since the 1950s, these factors have developed explosively and are major influences on late 20c English, the language most closely involved in computing.

Nature

The present-day computer derives from British work during the Second World War on cryptographic machines and is the most recent in a line of calculating devices that includes the abacus, the Jacquard loom, Babbage's Analytical Engine, and Hollerith's tab-sorter. Its primary purpose has been to compute, not to compile or converse. There are two kinds of computer: analog and digital. Analog computers, which are related to the slide rule and tables of logarithms (and virtually obsolete), use the strengths of voltages to represent the size of numbers, whereas digital computers use electrical signals only in the on/off form. Currently, digital computers consist of four major parts: (1) A processor or central processing unit (CPU), which executes commands, performing arithmetical, logical, and manipulative operations on the data stored in the second part. (2) A memory, the information store. Most computers have at least two kinds of memory: primary and secondary. Primary memory is usually silicon chips, typically DRAM (dynamic random access memory) chips. ‘Random access’ means that any part may be obtained immediately, as with a book that can be opened to any page. The process is fast, usually less than one microsecond to obtain an item of information. Secondary memory is usually magnetic disk, made of one or more platters rotating under a reading head. It is not random access: a particular part of the disk cannot be read until it rotates under the reading head, which usually takes several milliseconds. Storage is measured in bytes, one byte containing eight bits, and representing storage for one character in European alphabets. See ASCII. (3) Input/output equipment, which enables the user to get information into and out of the machine. The information is entered most commonly through a keyboard but also through removable disks, tapes, and other devices. Output goes to display screens, to printers (which produce text etc., usually known as hard copy), and also to disks and tapes. (4) Communications equipment, which permits a computer to ‘talk’ to other machines and to people located at a distance from it. The equipment includes a modem (an acronym for ‘modulator demodulators’), which connect computers by telephone line, and networks to let machines talk at high speed to each other, as for example in using the INTERNET and the WORLD-WIDE WEB.

Computer programs

Since computers work very fast, they cannot be directed step by step. Instead, a script must first be written for the computer to follow. The script typically contains sequences to be repeated, so that the script is much shorter than the operation as executed. The computer responds to machine language, which is binary code (strings of 0s and 1s), in which the operations are very simple (such as elementary arithmetic or moving one piece of data from one place to another). Such scripts are written in higher-level languages called computer programs (BrE following AmE in this spelling, but AmE follows BrE in doubling the m in programming). A distinction is now universally made between the equipment as hardware and software, the latter now generally made available as commercial software packages.

Computer languages

Also programming languages, high-level languages. Digital computers can follow directions written in a great variety of artificial languages that provide precise specifications of operations to be done and the order in which they must be done. Although strings of letters are used to name commands in these languages, they are quite different from natural language. Among other things, they must be logical and unambiguous: unlike people, computers do not know that the and in I like bread and jam means ‘both together’, while the and in I like cats and dogs does not imply that both must be present at once (= ‘I like cats and I like dogs’). Compared with natural language, high-level computer languages normally have: (1) Very short words: most programmers save effort by giving variables names such as x, one or two letters long, and by using many abbreviations, such as del for delete. (2) Very short utterances: written English sentences might average 20 words in length, but statements in programming language are typically only six items long. (3) Little syntactic variety: the typical computer language at present has a grammar of about 100 rules, compared with thousands in a formal grammatical description of English.

Specific languages

The many programming languages are divided into business languages (verbose, emphasizing simple operations on complex data) and scientific languages (terse, emphasizing complex operations on simple data). They often have distinctive histories and functions, and names of etymological interest. ALGOL, a language suitable for expressing algorithms, is the computational equivalent of Esperanto, created in 1960 by an international committee. Its name, a reduction of Algorithm Language, is a homonym of the star Algol (Arabic, ‘the ghoul’). BASIC is short for Beginner's All-Purpose Symbolic Instruction Code, designed at Dartmouth College in New Hampshire in 1965 by J. Kemeny and T. Kurtz. It is often the first programming language learned and is similar to the Basic of BASIC ENGLISH, also an acronym. ADA was designed in a competition run by the US Department of Defense from 1974 to 1980, going through successive refinements with such names as Strawman, Woodenman, Tinman, Ironman. The French computer scientist Jean Ichbiah led the winning team. It was named after Lady Ada Lovelace, daughter of the poet Byron and a supporter of Charles Babbage, the inventor of the Analytical Engine, an early mechanical digital computer. She is often called the first programmer. For some years, the goal of ‘programming in English’ (that is, using a more or less unrestricted subset of the natural language) attracted attention, but it has so far proved unattainable.

Processing text

Computers, among other things, are extensions of writing and print systems, and have therefore been used with greater or less success to do such things as evaluate, index, parse, translate, correct, and ‘understand’ text. When a suitably programmed computer is fed English, it can process it at several levels, but with decreasing competence as the task becomes more complex. The following sequence is typical:

1. The character level.

Text can be entered into a computer by three means: keying it, typically into a word processor which will format the text (arranging the line lengths and character positions); scanning it, using a machine which transfers a paper version into an image followed by a program that seeks to recognize the characters in it; transferring it electronically, typically by diskette or telephone, from another compatible computer. Transfer is the fastest and most accurate method, but currently the least used. When a cleanly typed or printed original is available, without too many fonts or typographic complexities, scanning is faster and easier than rekeying. Once the text is entered, computers can print it in a wide variety of typefaces, sizes, and page formats, using either a printer or a desktop publishing system.

2. The word level.

A spelling checker can find some kinds of typing mistakes, usually by comparing words with a dictionary list and noting those that are not in that list. Programs can make word lists and concordances (lists of each word with some context before and after it). By noting the most frequent words in a document, and comparing the word frequencies in a particular text with the average word frequencies in English, a program can suggest words that might be used for indexing the document. The counting of relative word frequencies and comparison with word frequencies from a standard sample can also help in guessing the authorship of anonymous works or measuring the readability level of a text.

3. The sentence level.

On the level of syntax, PARSING programs can try to define the structure of sentences and relationships among words. This is typically done by applying grammar rules of the form ‘a verb phrase may be a verb followed by an adverb’. Unfortunately many sentences are ambiguous. In the preceding sentence, a computer would not know whether Unfortunately modified the verb (implying that it is sad that ambiguous sentences occur) or the adjective many (suggesting disappointment that ambiguous sentences are so frequent). Adding a comma after Unfortunately could, however, serve as a means of disambiguation. However, some kinds of grammatical and stylistic errors can be diagnosed, and grammar checkers and style checkers have become available to help in the writing of business letters and the propagation of PLAIN ENGLISH.

4. The message level.

At the level of word-and-sentence meaning, semantic analysis can map a sentence into a knowledge-presentation language. Some research projects have been able to take such sentences as Which ships are in port? and answer them by looking at a table of ship locations, but such systems currently operate in strictly limited subject areas. Other applications of semantics include machine translation and direct generation of language by computers (that is, the computer produces text without human input).

The above levels of activity depend on computational linguists writing rules of analysis, accumulating a GRAMMAR of syntactic and/or semantic rules for such a language as English. An alternative strategy for processing written language, however, uses reference books: the use of a MACHINE-READABLE dictionary or thesaurus may help a computer make reasonable guesses about which sense of an ambiguous word is intended in a particular context. Another strategy relies on the statistical properties of large corpora to determine word relationships. Such methods have allowed parsing without writing a grammar in advance, a higher quality of error correction in spelling, and the automatic recognition of phrases. However, they handle uncommon constructions less well than the grammar-based procedures handle them, and depend for their success on the fact that such constructions are uncommon. See COMPUTERESE, COMPUTER USAGE, CONCORDANCE, CORPUS, EMOTICON, ICON.