Hypernyms

Hypernyms are words that are supersets of other words. For example, pasta is a general description, with lasagna, fettuccine, and tortelli being subsets, also known as hyponyms, within the general category of pasta.

graph TD;
    Pasta --> Lasagna & Fettuccine & Tortelli

Hypernyms can be used to help users find products when they're searching for the hypernym, as products, which might fit the description, might only be found on the more specific terms.

For example, if a user is searching for pasta with no hypernyms registered, they will not be able to find products, which are only described as tortelli. After registering the hypernym above, users can now find all three kinds when searching for "pasta".

As opposed to synonyms, hypernyms are one-way. This means that the user will not find products, which are only described as pasta when searching for "tortelli", even after registering the hypernym.

As there is no boost reduction when matching products based on text analysis, hypernyms can also be used to improve the ordering of products in a search result. If a product has tortelli in the title and pasta in the LongDescription, the product will be seen as a better search result if the hypernym has been created, as matches in the title generally have higher search relevance matches in the LongDescription.

Nested hypernyms and shared hyponyms

A hyponym can itself be a hypernym for more specific terms, which forms a nested hierarchy. A hyponym can also belong to more than one hypernym at the same time — these are called shared hyponyms — and this is a valid, supported configuration.

For example, cat can be a hyponym of both animal and mammal, while cat is itself a hypernym of lion and tiger:

graph TD;
    animal --> cat & shark
    mammal --> cat & dog
    cat --> lion & tiger

Hypernyms are displayed as a fully expanded tree, with one tree per top-level hypernym (a hypernym that is not a hyponym of any other hypernym). A shared hyponym, together with everything nested beneath it, appears under each of its parents. In the example above, cat and the terms nested under it (lion and tiger) are shown under both animal and mammal. These repeated entries are intentional: they show the complete inheritance structure, and a term appearing in more than one place confirms that the relationships are configured correctly.

The only configuration that cannot be displayed as a tree is a circular reference, where a hypernym is — directly or through a chain of other hypernyms — its own hyponym. In that case the hypernyms are shown as a flat list instead of a nested tree.

Hypernyms are created as a part of a hypernym dictionary.