In 1954, in a famous public demonstration of a research project known as the “Georgetown-IBM experiment,” Russian was automatically translated into English by an “electronic brain” in the form of an IBM 701 computer, shown here.

1954

NATURAL LANGUAGE PROCESSING

In 1954, an IBM press release proclaimed: “Russian was translated into English by an electronic ‘brain’ today for the first time. . . . The famous 701 computer . . . within a few seconds, turned the sentences into easily readable English. A girl who didn’t understand a word of the language of the Soviets punched out the Russian messages on IBM cards.” The press release continued by explaining that “the ‘brain’ dashed off its English translations on an automatic printer at the breakneck speed of two and a half lines per second.”

In 1971, computer scientist Terry Winograd (b. 1946) wrote SHRDLU, a program that translated human commands such as “Move the red block next to the blue pyramid” into physical actions. Today, natural language processing (NLP) often involves many AI subfields, including speech recognition, natural language understanding (e.g., machine reading comprehension), and speech synthesis. One of the goals is to facilitate natural interactions between humans and computers.

The early days of NLP usually involved the use of complicated sets of manually created rules; but in the 1980s, NLP made increasing use of machine learning algorithms that learned rules through analysis of large sets of example language input. Typical NLP tasks may involve machine translations (e.g., translating Russian into English), question answering (e.g., “What is the capital of France?”), sentiment analysis (emotions and attitudes on a topic), etc. Analyzing input from text, audio, and video, NLP is employed in diverse realms, including spam-filtering for email, summarizing information in long articles, and question-answering by smartphone apps.

NLP is challenging for many reasons. For example, in speech recognition, the sounds of adjacent words blend into one another, and the computing system must also take into account syntax (i.e., grammar), semantics (i.e., meaning), and pragmatics (i.e., purpose or goal)—along with the many classes of ambiguity in language where words take on different meanings in different contexts. Today, significant use of artificial neural network methods helps to improve accuracy.