search / provide results by semantic classification

[dwhieb]

Allow the user to take advantage of the existing semantic classifications to find words that are semantically related to the one they're searching for.

Possible implementations:

• Amazon.com-style suggestions: "You might also be interested in..."
• Advanced search features for power users, ex: meaning:furniture

[aarppe]

An alternative, or parallel functionality would be to provide a browsing feature, where one peruse the contents of the dictionary either alphabetically or according to semantic classes, perhaps making use of the hierarchical organization of both WN and RW.

Beside search, we would want to make the semantic classes visible for the results of any regular dictionary search.

[eddieantonio]

I like the idea of related words! Very useful. Perhaps if we have some sort of word vectors, this would be more-or-less straightforward to implement.

[aarppe]

The current word-vector search already stealthily gives as a list of Cree words ranked in terms of their semantic closeness to the search words, as well as other items.

But what we'd want to implement here is an explicit way to extract entries classified under various WordNet and RW semantic classes, perhaps using tags like: rw:6.2 or rw:agriculture. For RW, we could simply show any entries within the indicated class and further down in the hierarchy. For WN, we'd need to incorporate its semantic hypernym/hyponym hierarchy, and effectively traverse down the entire path from the search word term. I'm wondering whether one would want to sort those in level by level, or with some other algorithm. What we'd also need to figure out is how to possibly disambiguate WN search terms.

Anyhow, probably starting with RW is the easiest. @dwhieb - we'd need you to add the WN and RW classifications to our dictionary DB for crk.

[dwhieb]

That can be done!