4

Lojban was created to be syntactically unambiguous.

I wonder though if there are natural languages that have the same property, or at least come much closer to having it than English does?

Has anyone attempted to quantify the relative ambiguity of languages and to rank them?

10

Has anyone attempted to quantify the relative ambiguity of languages and to rank them?

Your timing is excellent! The most comprehensive study I've seen on this topic was published less than a week ago (at the time of writing).

The first question is, what is ambiguity? Pretty much every sentence has some sort of ambiguity in it: if I say "I like cats", does "like" mean similar to, or feel affection toward (a lexical ambiguity)? The syntax here disambiguates, makes it clear that the answer is feel affection toward.

So why do English-speakers consider this ambiguity acceptable? Why isn't it a problem to have one word that means both similar to and feel affection toward? Because the two words are generally used in such different syntactic contexts that it isn't an issue. There are only so many strings of phonemes to use, and reusing some makes the average word length shorter, so it takes fewer phonemes to convey a meaning.

For a cross-linguistic example, Swahili has no separate words for "he", "she", "him", or "her": yeye is used regardless of case and gender. Ancient Greek has thirty (it just uses the demonstratives), allowing it to distinguish "the first man I mentioned" from "the second man I mentioned" from "a man newly introduced to the dialogue", along with five different grammatical cases (subject, direct object, indirect object, possessor, addressee).

So it seems that Swahili is significantly more ambiguous, right? It would need more words than Ancient Greek to convey the same amount of information: one Swahili word could correspond to any of thirty different Ancient Greek words, and only context can make it clear which is intended. Textbook ambiguity right there. That seems like something that can be objectively measured and ranked, using information theory.

Well, Coupé, Oh, Dediu, and Pellegrino did exactly that. (Less formal summary of their work here.) They calculated how many bits of information are contained in each word (and in each syllable) in seventeen different languages. Vietnamese, for example, conveys about eight bits per syllable, significantly more than English.

But when multiplied by the rate of speech—how many words/syllables are spoken per second—all the results were almost identical. Every single language they studied conveyed about 39.15 bits per second. Languages with a lower information density per word are spoken faster to compensate, getting in more words per second, while languages with a higher information density per word are spoken slower (since losing any individual word in transmission would be a bigger problem).

This implies that the answer to your question…

Are some human languages significantly less ambiguous than others?

…is no. Some languages may have more lexical ambiguities or more syntactic ambiguities, conveying less information per word. But they compensate by just using more words, putting in more context until the ambiguity is gone, in the same amount of time.

| improve this answer | |
  • A very satisfying answer. This reminds me of the similar question for information density per character. Yes, chinese script uses less characters, but I'd guess that No, it doesn't use less strokes and the characters aren't the same size. Some more involved character consisting of 4 individual characters each potentially consisting of ... basically look like an ink blot at my usual font size. – vectory Sep 8 '19 at 22:25
  • 1
    I don't see how rate of speech is related to ambiguity. Replay, say, Lojban and English on fast-forward, so that their rates match. Sorry, but I don't see how this is an answer. – MaxB Sep 8 '19 at 22:43
  • @MaxB Because the solution to ambiguity is providing more context. I'm assuming you wouldn't consider "I like cats" ambiguous, even though there's a clear lexical ambiguity there—that's because of the context provided. This study shows how languages with higher degrees of (mainly lexical) ambiguity handle it by providing more context for each word, resolving that ambiguity. – Draconis Sep 8 '19 at 23:39
  • @Draconis There's another non-disambiguated lexical ambiguity there. Namely do they taste agreeable, or do you wish them well? Being a veggie, I often respond to friends who ask "Do you like fish?" (etc) with "Yes, that's why I don't eat them!" – Araucaria - Not here any more. Oct 20 '19 at 18:58
  • 1
    @Araucaria Also true! That one requires pragmatics to disambiguate rather than syntax, but the solution also seems to be more context. – Draconis Oct 20 '19 at 19:02
5

Languages per se are not ambiguous or not ambiguous.

Rather, instances of language - sentences, phrases, words... - are.

All natural languages evolved to allow varying degrees of ambiguity/specificity, which speakers then choose for the situation.

Consistently forced disambiguation would be a highly undesirable property.

Ambiguity can be syntactic but that is hardly the only type, there are semantic ambiguities because of homophones and co-references.

If an experiment like lojban has any success in the real world it will also end up like that, if it's not already.

| improve this answer | |
  • You can define the ambiguity of a language as the ambiguity of its typical usage (mind-boggling, isn't it). Sorry, but I really don't see how your answer is an answer. It should be a comment, if you really believe this is helpful. – MaxB Sep 8 '19 at 22:40
  • Would that be a property of the language, or of the corpus, cultures etc? How do you do the accounting, eg is the ambiguity of pro-drop less than or greater than the ambiguity of using present tense for the future? The answer is a function (of all these params...) and you're insisting on a single number. – Adam Bittlingmayer Sep 9 '19 at 4:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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