1

What word is valid across the largest number of different languages, and as different part-of-speech? (The precise term is interlingual homographs/heteronyms/polysemes)

Examples:

  1. 'rate' is both verb, noun and adjective in English, so count that as 3. 'rate' is a loanword but also a (distinct) noun in German ('instalment'), in French: (noun) 'spleen' and also as verb (present tense of 'rater'), and again a loanword in Norwegian. Giving us a grand total of 3+2+2+1 = 8.

  2. 'baba' has meanings across English, French, Spanish, Portuguese, Turkish, as a noun meaning variously: 'rum cake','saliva','father'...

  3. 'like' can be 8 or 9 valid POS in English alone ("What word can fulfill the most parts of speech?"), plus 3 in Norwegian, for a total of 12.

Clarifications:

  • I am not asking about homonyms within the same language (e.g. 'run' or 'set' have >400 meanings in English). I don't care about straight homonyms, for these purposes that all only boils down to 3 distinct meanings (noun, verb, adjective) for set.
  • Ignore capitalization (e.g. for German nouns)
  • Exclude universal loanwords like pizza (because it's the same word across languages, not a homograph.)
  • Exclude proper nouns (unless they also have a separate meaning, e.g. 'scotch', 'go dutch'..).
  • Include noun-, adjective- and verb-stemming
  • Include accents as being significant, hence plies, pliés are both valid in English and French.
  • Include colloquial usage, as long as it's common usage.
  • Words must be 2+ letters (arbitrarily, to avoid the 'letter 'A' is the name for the letter A, in languages X,Y and Z' definition that Muke quoted. But allowing unusual definitions like wye or cym, as he cited.

A good multi-lingual dictionary for verifying how many languages a word-spelling is legal in is thefreedictionary.com or these others.

(Motivation: this question arose (legitimately) from a SO question "Data structure for multi-language dictionary?"

On-topicness: If anyone quibbles whether this is a practical, answerable question based on actual problems that you face, see preceding Motivation remark; also I wanted to ask this on NLP/CL, but that's sadly closed, and non-English questions are offtopic for EL&U, and this is probably neither on-topic nor well-suited to the general StackOverflow audience. So yes, Computational Linguistics&NLP is on-topic for Linguistics.SE)

18
  • 3
    I doubt you'll be able to get a good answer to this, since linguists aren't particularly interested in homographs (other than as a practical concern). But I think it's a legitimate question, since Computational Linguistics questions are on-topic, so +1. – arjan May 14 '12 at 16:23
  • @arjan: easy to generate all combinations of 2,3,4-letter words (containing at least one vowel from {a,e,i,o,u,y}), then run them through a dictionary(/ies) like Farlex, and sum the count of distinct homographs for distinct POS across multiple languages. – smci May 14 '12 at 20:47
  • I think this should be closed as it's not about linguistics. It's about the same sequence of written symbols representing words in the writing systems of many languages. It's not actually about "words" as they're understood in linguistics. – Gaston Ümlaut Jun 1 '12 at 4:00
  • @Gaston: you're flat wrong, here's citation: Computational linguistics has subfields about real-world considerations like 'words' as they're found on the internet, aka reality: outside of labeled academic corpora, we can only statistically infer what the person who wrote the word meant, e.g. rate l'examen vs rate my apearance. Specifially the subfields are Design of parsers or chunkers for natural languages and Design of taggers like POS-taggers (part-of-speech taggers) – smci Jun 2 '12 at 17:04
  • 1
    No, let's exclude guesses. Guesses mean lots of answers, little content. So make it clear in your question. – Alenanno Jun 3 '12 at 15:58
3

I choose /ki/, which is a noun class prefix in a good hundred Bantu languages. It also means "key" in English, "key" in Langalanga, "who" in French, "give" in Ngwe.

1
  • /ki/ means "who" in italian as well, and "tree" in Japanese. :) – Alenanno Jun 4 '12 at 8:44
1

I think it's interesting enough to try to lay out a method that one might use to try to answer this question, and I think the question is more interesting as one about words written in a basic / lossy IPA (such as how we tended to use Kirshenbaum on Usenet) rather than the languages' respective orthographies.

If I had the relevant tools available, I'd try this:

  • Two-sound CV combinations should probably win out.

  • Look for two-sound Austronesian words. There are over 800 Austronesian languages, and many Malayo-Polynesian of them can be quite close.

  • There are 250-500 Bantu languages, and they can be quite close as well, so if bare prefixes are allowed (and also even if they aren't), jlovegren's suggestion of /ki/ should be taken quite seriously, especially since both the consonant and the vowel are extremely prevalent worldwide.

  • It's important to note that along with importation of the Roman alphabet we get at least 20-30 loanwords for names of letters into each language. This applies to the hundreds, possibly thousands of languages that "use" the Roman alphabet to some capacity.

  • Languages using scripts in the Brahmic tradition tend to name consonants with a trailing schwa. /kə/, for example. Whether this vowel is involved in many words in other languages seems to rely heavily on how we're transcribing.

My guess (heavily influenced by how Malayo-Polynesian languages tend to pronounce the alphabet) is we would end up getting /ka/, /ke/, /ki/, /nu/, /pe/, /pi/, /te/, or /ti/.

Another question: For each length n, what is the most popular n-letter (in IPA) word? For n at least 4, I have no idea what one could do short of crawling a massive database.

Edit: Just noticed that we're supposed to exclude universal loanwords, so I'll rework my answer soon or take it down.

Not the answer you're looking for? Browse other questions tagged or ask your own question.