In “The Grammarian’s Five Daughters,” a fable by science fiction writer Eleanor Arnason, a mother bestows grammatical gifts to five daughters seeking their fortune in the world. The eldest daughter gets a bag full of nouns, the next gets verbs, the next adjectives, and the next adverbs. The youngest daughter is stuck with the leftovers, those “dull little words” overlooked by everyone else: the prepositions. But the prepositions ultimately bring order to a chaotic land, serving as the foundation for a strong and thriving nation organized under the motto “WITH.”
Arnason’s story of prepositional prosperity is pleasing to lexicographers who spend a lot of their time fretting about these “dull little words” and how best to define and categorize them. Compared to nouns and verbs and adjectives, prepositions are resolutely “unsexy,” quite the opposite of the ostentatious stunt words I looked at last week. But you can think of them as the connecting tissue of language, quietly forging relations between more easily graspable parts of speech. And now the lowly preposition is finally getting the attention it deserves with The Preposition Project (TPP), a monumental undertaking that systematically describes 673 distinct senses for 334 prepositions in English.
Oxford University Press played a contributing role in TPP, making available the definitions of all the prepositions found in the Oxford Dictionary of English (ODE), since the project creators found ODE’s sense inventory for prepositions to be the best available. The project took all of these preposition senses and mapped them onto a database of sentences from FrameNet, a resource that tags every word in its corpus with an appropriate semantic role (or “frame”). The relationships described by English prepositions can often seem vague and hard to get a hold of, even for native speakers, but TPP’s linking of dictionary senses to semantic frames turns the elusive workings of prepositions into something much more lucid and concrete.
TPP’s online lookup lets you navigate through the complex prepositional landscape. Take a seemingly simple word like through. You’ll find that ODE provides 5 main senses for through, and 13 different subsenses. For each sense and subsense, the TPP team has matched it up with FrameNet sentences in order to specify its exact semantic and syntactic properties. So, for instance, the basic sense of through meaning “moving in one side and out of the other side of” has a subsense used specifically to describe something permeating an obstacle. The thing doing the permeating can be as ethereal as a beam of light “streaming through a window,” or it can even be a perceived sound, as noises heard “through the walls.” Each little nuance of the preposition’s usage is analyzed to determine exactly how it can function with relation to other words in a sentence.
All of this research should be tremendously useful in the burgeoning field of Natural Language Processing (NLP), which allows computers to evaluate the sentences that humans produce. It’s useful for dictionary makers too, since it helps us see where the existing definitions of prepositions might not accurately mirror the range of usage found in a big corpus of texts. This kind of analysis can be grueling and downright obscure, bringing to mind Samuel Johnson’s famous definition of “lexicographer” as “a harmless drudge.” Or as Grant Barrett put it, “That’s exactly the kind of hard work lexicographers and computational linguists do that you don’t think of when you say to yourself, ‘Gee whiz! It’d be swell to make dictionaries, wouldn’t it?’” But that’s what life is like down in the dictionary trenches.
Ben Zimmer is an editor at Oxford University Press and a true word junkie. Once a week he surfaces from his dictionaries to write this column. Check out his “words of the week” on our main page (center column) or by clicking here.