Is there an entity extractor (or set of extractors) that can extract portions of a word as an entity?
For example, if a user inputs a string such as word30, I would like this to be parsed into two entities: word and 30. Duckling handles the number fine, though I get training warnings for mis-tagged entities if I just tag word in word30.
Thank you!
Tagging parts of words won’t work for the CRFEntityExtractor, no. You can however tag the whole word and map it to the word “word” using synonyms.
Thank you! Unfortunately I probably cannot map all of the combinations of “wordX” to “word” as synonyms because X is variable…well it actually is possible as it is just a large subset of integers. Though in the case where X is close to infinite, I am guessing I would need to make a custom extractor like duckling to parse this wordX combination.
@akelad can a word be two entities? like
“wordx” is “entity1”
“wordx” is also “entity2”
i am trying to find [aziz] (person)(human)
aziz is human and as well as person
No you should only label it as one, if you want to fill two different slots with it then you can do this in a custom action.