7

I've been working on natural language processing/understanding for a great many years and sometimes get hung up on trying to figure what superordinate a word belongs to, such as a hammer being a type of tool.

A more difficult example is shade, such as the shade of a tree. I never have figured out a good superordinate for it. I'm not asking for an answer to this one example. I need a good list as a reference source for many such problems.

Googling for lists of superordinate, hyponyms, and word categories has turned up nothing. Looking up words in dictionaries and Wikipedia can sometimes help, but often does not, and is a lot more time-consuming than using a list.

3
  • 'Shade' would be a hyponym of 'protection', I would say.
    – Nigel J
    Commented Mar 15, 2018 at 14:47
  • I've seen the word 'classmates' used to mean 'words semantically connected'; if this sense is accepted, 'classmate' is a hypernym of 'hypernym'. You could look through a good thesaurus giving classmates rather than just synonyms (cutlery / knife / fork / spork ...).
    – Edwin Ashworth
    Commented Mar 15, 2018 at 17:42
  • Not an exact duplicate, but the answers to the Meta question List of nouns by category may help you. For something really specific like the shade of a tree there might not be a good answer, though your best bet might be to ask a question.
    – 1006a
    Commented Mar 16, 2018 at 2:15

1 Answer 1

Reset to default
2

WordNet is the state of the art ontology (to be described) for basic vocabulary of English. (there are many ontologies out there for technical vocabularies).

An ontology is a vocabulary along with relations among the terms that support the definitions. One of the most important relations is 'subset' or 'is-a' relation which gives both hypernyms and hyponyms (since the set of dogs is a subset of the set of mammals, the word 'mammal' is a hypernym of the word 'dog', and 'dog' is a hyponym of 'mammal'.

Because this is a tree-like (or really partial order) relation, and there are many relations with other properties, it does not form a simple list. But the WordNet vocab and relations are downloadable as a set of data that you can manipulate by programming.

Of course, as with much of electronic data, it is black and white, and may not encode things exactly as you see them. It may well have a good hypernym for shade, or it may have one that doesn't really suit your intuition.

Note that there is a difference between the concepts in a relation, and the words or terms used to express them. I.e. there may be a good concept that captures ideas of the dogs, the cats, and the bears, without including the rest of the carnivores, but there is no good existing single word for that single coherent concept. In other words, for a possibly very good and very common concept, there just may not be a single word for it.

See also Is WordNet trustworthy

10
  • Thanks. I've downloaded and looked at WordNik in the past, but had two problems: (1) the data didn't seem to suit my needs, and (2) I was reluctant to invest a lot of time in trying to figure out or to learn the format the data is in and to write code to access it when the data may not be helpful anyway.
    – user1828
    Commented Mar 16, 2018 at 19:19
  • Just in case, not Word Nik, but Word Net. If this isn't good for you, then, what would you like? a two column list, a word in field 1 and and its (least) hypernym in field 2?
    – Mitch
    Commented Mar 16, 2018 at 19:29
  • (I got cut off.) I just visited the WordNik web site and used their online interface to ask for "shade" and I didn't see anything resembling a usable superordinate. But like I said, I'm not concerned about one word ("shade"). What I was hoping for was a simple list that would help me with the hundreds or possibly thousands of words which are difficult to categorize. But, I've often wondered what you closed with -- maybe there is not a superordinate for every hyponym.
    – user1828
    Commented Mar 16, 2018 at 19:33
  • Yeah, Word NET is where I went and what I have downloaded, not WordNik. Sorry. But yes, a two-column list with a hyponym in the first column and its superordinate in the 2nd. In my OP, I gave the example "hammer" and "tool". In NLP, this allows "hammer" to inherit all the characteristics assigned to "tool", thus saving a lot of duplication.
    – user1828
    Commented Mar 16, 2018 at 19:39
  • You're not going to find that 2 col file anywhere, but you should be able to create it from WordNet.
    – Mitch
    Commented Mar 16, 2018 at 19:44
  • "maybe there is not a superordinate for every hyponym": philosophically, there must be one for every noun and that is 'concept' or 'thing' or 'entity' (just like in Java there is the one base class 'Object'). But there's no guarantee it is going to be ... salient (intuitively what you expect). Keep clicking on 'direct hypernym
    – Mitch
    Commented Mar 16, 2018 at 19:52
  • At this point you're really looking for programming advice. You might get a more thorough answer (where someone has done it before) on another SE site, like datascience.stackexchange.com or regular old stackoverflow.stackexchange.com. Ask for something like 'Using WordNet to extract simple hyponym-hypernym list of pairs'
    – Mitch
    Commented Mar 16, 2018 at 19:58
  • hammer is a tool. tool is an implement. implement is an article. article is an object. object is a thing. thing is ...? Do you run out of superordinates eventually?
    – user1828
    Commented Mar 16, 2018 at 20:00
  • "Do you run out of superordinates eventually?" That's another side to the phil. question. Maybe you can continue in both directions, keep hypernymming up by arbitrary unions, getting more and more specific hyponyms downwards by arbitrary intersections of properties. I sorta think that 'concept' is just the extreme maximum 'top' of the whole thing. But that's philosophical. For a given finite set of vocabulary, there will always be an end to the climb upwards because you only have a finite number of words that could be a successor.
    – Mitch
    Commented Mar 16, 2018 at 20:05
  • Googled "extracting data from wordnet". The good news is that wordnetport.sourceforge.net has already extracted data into spreadsheet tables for different groupings which are at least readable. The bad news is that the column which I take to be hyponyms & superordinates are not going to be very helpful for me, but I'm posting the site here in case someone else is looking for this type of data. Thanks for your help.
    – user1828
    Commented Mar 16, 2018 at 23:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .