Margaret Masterman

• 1. Introduction

  ›

  Margaret Masterman was born in London on the 4th of May 1910, and by the time she died in 1986, the field she had helped create had largely forgotten her name. She spent decades arguing that computers could learn to understand language, not by mastering grammar rules, but by grasping the meaning beneath the words. At a time when those ideas were treated as implausible, she kept working anyway, running a small research unit out of a former museum with Buddhist sculptures still carved into its walls. How did a philosopher trained by Ludwig Wittgenstein end up founding one of the most quietly influential centres in computational linguistics? And why, when her ideas finally entered the mainstream, did almost no credit flow back to her?

• 2. Wittgenstein's Notebook

  ›

  Ludwig Wittgenstein's lectures at Cambridge in the early 1930s drew a small and carefully chosen audience. Masterman was among the handful of students selected to take notes during his 1933-34 course, and those notes became something lasting. Her records, combined with those of other students, were later compiled as The Blue Book, one of the key documents for understanding Wittgenstein's thought during that period. There is also a companion volume, the Yellow Book, which contains notes taken by Alice Ambrose and Masterman during pauses in the Blue Book dictation. Sitting in those sessions shaped Masterman's deepest instincts about language: she emerged as a committed Wittgensteinian sceptic, deeply suspicious of any system that claimed a limited sublanguage or logic could stand in for the whole of natural language. That scepticism would run through every research program she designed for the rest of her life.

• 3. The Unit on Adie's Lane

  ›

  In 1955, Masterman founded the Cambridge Language Research Unit, known as the CLRU, and directed it herself. It had started two years earlier, in 1953, as an informal discussion group with a heterogeneous mix of people interested in language from both philosophical and computational perspectives. The building it occupied had a striking past: Adie's Museum, a small but handsome structure that had once housed far eastern art. Small Buddhist sculptures were still built into its walls, and carved doors framed the entrance to what had become a research centre. For roughly twenty years, the CLRU worked on machine translation, computational linguistics, and quantum physics, though it sat entirely outside the official university structure. Its computing resources were minimal. The main machine was an ancient ICL 1202, and most serious computation happened on the Cambridge university machine at the Mathematical Laboratory, or at sites in the United States during visits by CLRU staff. The unit's staff never exceeded ten people, yet its funding came from American agencies including AFOSR, ONR, and NSF, as well as UK government bodies and eventually European Union funds.

• 4. Meaning Before Grammar

  ›

  During the years when Noam Chomsky's syntactic theory dominated linguistics, Masterman held a very different position. From 1951 through 1966, she argued that semantics had to come first in any serious attempt to process language by machine. Her core conviction was that language works through redundancy: a writer says the same thing repeatedly in different ways, and it is precisely that repetition that allows a reader to resolve ambiguities. She drew this idea partly from information-theoretic thinking about language as a signal. R. H. Richens' work on semantic primitives gave her a concrete method. She and colleagues at CLRU developed actual programs using a thesaurus as the primary vehicle for semantic operations, using it for tasks ranging from indexing to translation. She did not believe, however, that semantic primitives were fixed or foundational in a simple sense. She insisted they would only make sense if there were empirical procedures for discovering them and a theory that could account for the fact that they, too, would develop the same polysemy as any natural language. For some years, she and colleagues explored lattice theory as the formal structure underlying thesauri, seeking to let primitives emerge naturally from the classification of semantic relations.

• 5. Breath Groups and Rhythm

  ›

  In the later years of her work, Masterman became preoccupied with rhythm, stress, and what she called breathgroupings. Her claim was pointed and unusual: languages take the form they do, at least partly, because they are produced by creatures that breathe at fairly regular intervals. She believed that each breath group carried a distinct meaning unit, and that translation could work by breaking a sentence into breath group segments, translating each segment independently, and reconstructing the target sentence from the translated pieces. This approach stood in direct contrast to dominant methods of the time, such as Systran, which worked through a dictionary and rule-based system. In 1980, she attempted to restart the CLRU with William Williams, hoping that the newer generation of microcomputers could run her breath group algorithms. She walked seven miles from Millington Road in Cambridge to Orwell and bought two North Star Horizon computers from a company called Intelligent Artefacts. These machines were installed with the Forth programming language, written by David Sands, and students from the University of Cambridge programmed her algorithms into them. The project never reached the scale she had envisioned.

• 6. Kuhn's Missing Word

  ›

  In 1965, Masterman presented a paper called "The Nature of a Paradigm" at the Fourth International Colloquium in the Philosophy of Science, held in London. Her subject was Thomas S. Kuhn's book The Structure of Scientific Revolutions, published in 1962. She described it as "at once scientifically perspicuous and philosophically obscure" and praised Kuhn himself as "one of the outstanding philosophers of science of our time." Her central observation was that Kuhn used the concept of paradigm in at least twenty-one different senses, which she organised into three groups: metaparadigms, sociological paradigms, and artefact or construct paradigms. She pointed out that Kuhn's critics had focused only on metaparadigms, missing the other two. Kuhn accepted her criticism, and it was a key factor in his subsequent move away from the word paradigm toward the concept of incommensurability. A philosopher who was overlooked in one field had managed to redirect the central terminology of another.

• 7. Lost and Recovered

  ›

  When Masterman died on the 1st of April 1986, William Williams shut down the CLRU. Its library of early machine translation documents, spanning decades of rare primary material, was thrown into a skip. Two university bodies had offered to take it in, but the offer was not accepted. The ideas she had championed survived in a different form. Woody Bledsoe, in his 1985 presidential address to the American Association for Artificial Intelligence, spoke of those who had attempted machine translation in the 1950s and 1960s: he said they may have failed, but they were right to try, and the field had learned so much from their efforts. Masterman's specific ideas found no such acknowledgment. Her thesaurus-based approach to semantic operations, her use of redundancy as a processing principle, and her scepticism about syntax-first methods all anticipated directions that later became standard. When those ideas were written up by students or independently reached by others, the source was rarely traced back to her. Three CLRU alumni have since received the Annual Lifetime Achievement Award from the Association for Computational Linguistics: Martin Kay, Karen Sparck Jones, and Yorick Wilks.