2 ms·
Indeed, but even the standard definitions are flawed. When I fist was interested in NLP, I thought "word" was easily defined as "sequence of letters bounded by
by ben_w 3y ago
Indeed, but even the standard definitions are flawed.
When I fist was interested in NLP, I thought "word" was easily defined as "sequence of letters bounded by non-letters". But we don't normally pronounce spaces, do we? AndyoucanwritewithoutspacestoindicatespeakinglikeYahtzeeCroshawonZeroPunctuation. And when we've got a contraction, is that "we've" two words or one? Compound words in general, for which German is famous ("Antibabypille") but that's also present in English ("antidisestablishmentarianism", or less severely "internet" and "smartphone")? What about single nouns with spaces in the middle like "remote control" or "mobile phone")? And as my original definition was letters, what about the apostrophe in "we've" or "O'Neil", or the hyphen in older posters saying "to-day"? If all names are words, what about numbers in product names such as "iPhone 15"?
And all those pedants complaining about "to boldly go" in Star Trek because of the split infinitive; I'm told that's because infinitives in Latin were single words and therefore un-splittable, but should admit that I don't know Latin and I don't care to learn it.
- mostlylurks 3y ago> antidisestablishmentarianism This is not actually a compound word, merely a word which stacks several layers of derivation on top of a single stem. > What about single nouns with spaces in the middle like "remote control" or "mobile phone")? These are compound words, identical in their nature to the stereotypical long German compound words, regardless of the fact that they're written with a space in the middle. The difference is merely orthographical. > When I fist was interested in NLP, I thought "word" was easily defined as "sequence of letters bounded by non-letters". There's an additional problem with this view which you didn't mention, which is that written language is merely one possible encoding of the actual underlying language, which is not merely a theoretical concern, but also a practical one. After all, even in English the encoding is far from 1-to-1, since you have multiple differing spellings for what would generally be considered to be the same word (e.g. "color" and "colour"), as well as identical sequences of letters which would generally be considered to be different words (e.g. "lead" and "lead").