5

In lexical semantics, a lot of meaning in individual words, the concept behind the utterance, is captured in ontological relations: is-like for synonymy, is-a for a hypernym hierarchy. But this doesn't capture a lot of the incidental nuances. Simply relating one word to another doesn't include all the non-word connotations; not every concept has a single word to label it.

In analogy with phonology, the area of generative semantics sometimes uses semantic features, often binary, to label lexical items. For example, one might label 'bucket' and 'pail' as follows: - 'bucket' - [+ receptacle] [+ wide open top] [+ big] [+ (opt) metal] - 'pail' - [+ receptacle] [+ wide open top] [- big] [+ (opt) milk] [+ (opt) toy]

(and one can see how a hypernym relation can be extracted from analyzing the containment of features)

Most dictionaries attempt some connections, usually synonyms and antonyms. But I haven't seen anything that is binary, except maybe if the word is very distinctive, a better dictionary might label the subculture it is specific too ('chemistry', 'vulgar', 'archaic').

Are there any published dictionaries, print or online, that attempt to do a binary feature analysis beyond toy examples? Assuming none, is this method simply an old academic trend that dies out long ago out of fashion or out of impracticality?

  • I've never seen a real-life Katzian dictionary either. I think such a dictionary would not fit current trends, since essentialist linguistics as a whole is fading away, and ideas like prototypes and fuzzy categories now take centre stage. – WavesWashSands Dec 14 '16 at 8:11
  • @WavesWashSands Does 'essentialist' mean rule based? But prototypes and fuzzy categories still need dimensions to be fuzzy on, and binary ones are the simplest dimensions. And then aren't those two ideas also a bit old-fashioned now (1980's)? – Mitch Dec 14 '16 at 14:20
  • 1
    I was just using 'essentialist' in its usual sense, i.e. characterisable by necessary and sufficient conditions. Yeah the ideas aren't new (Labov's cups was from what - the 70s?), but I think they've only gained traction since then. Though I'm not very familiar with the semantics literature - I might be wrong about the trends. I only get the impression that the field as a whole now embraces them (e.g. Croft's work on word classes) - what do you think? – WavesWashSands Dec 14 '16 at 17:11
  • 1
    And dictionaries are of course essentialist, but I don't think lexicographers are philosophically inclined to be essentialists. The reason we have dictionaries is usually pragmatic (those with a documentary purpose aside) - to explain words to language learners, translators, etc. This goal is best served using un-fuzzy definitions. An actual Katzian dictionary, however, would primarily be an intellectual venture, with little pragmatic purpose. So its creators will probably have some essentialist position in mind. – WavesWashSands Dec 14 '16 at 17:15
  • 1
    The success of generative lexicon theory, I think, is a good example of how non-essentialist linguistics now is, and that might partially explain the absence of a binary features dictionary. Compared to the generative lexicon (which has applications in computational linguistics), its empirical coverage is lacking. – WavesWashSands Dec 14 '16 at 17:23
1
+50

I think what you need is WordNet. This is a huge lexical database and it contains a lot words which are related in many different ways: hyperonymy, hyponymy, meronymy, antonymy, etc.

For more on WordNet, you can visit their website, read the papers, or just use their online browser to get a better feeling of what it does and what can you get from it.

There are also plenty of free open-source APIs on different programming languages on the net about it, so you can easily use it for a lot of stuff.

| improve this answer | |
  • WordNet is great. Their primary manner of structuring word knowledge is through subset or member semantic relations (all those -onymies you mention). Does WordNet also explicitly use binary semantic features? – Mitch Dec 16 '16 at 14:44
  • 1
    Maybe not directly, but you can definitely assign them with the other information of the words provided by WordNet. For example, Man is [+HUMAN], [-WOMAN], [+ADULT]. The ones with "+" property can be every word in the synset (you have "human" in the WordNet's synset of "man") or their hypernyms ("adult" is a direct hypernym of "man" in WordNet). As for the "-" property, you can take direct antonyms of each item in the synset ("woman" is a direct antonym of "man" in WordNet). – Belphegor Dec 16 '16 at 14:54
  • playing around with the WordNet interface for 'Man', I don't see any way to discover [-woman] (similarly with 'woman' to find [-man]) – Mitch Dec 16 '16 at 15:04
  • 1
    You are using it wrong then. When you have "man" for example, go on the first noun, and click on "S", then click "antonym" and there you can find the word "woman". – Belphegor Dec 16 '16 at 15:23
  • OK, nice, didn't realize what 'S' is for. So there is a full ontology underneath WordNet, and one can extract binary features from membership (it is philosophically and practically a question whether non-membership is the same as antonym: sometimes it is, sometimes it is not. But I am still concerned that the ontology doesn't allow classes without labels from the lexicon (eg, there is a conceptual superclass of both dogs and cats. there is no lexical name for it (maybe the biologists have one) but the concept may still be viable). – Mitch Dec 16 '16 at 15:40

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.