hypothesis / vision

Envisioning the future of the Hypothesis.
https://github.com/hypothesis/vision/issues/
40 stars 8 forks source link

[feature] highlight/annotate non-continuous sequences of text #220

Open jibe-b opened 6 years ago

jibe-b commented 6 years ago

It happens that valuable pieces of text contain parts that we don't want to highlight.

A behaviour like

would then create one annotation with the selected text as target and an indication such as [...] to indicate the brank.

robertknight commented 6 years ago

The annotation model in Hypothesis supports this, in that a single annotation can refer to multiple pieces of a document. Firefox actually supports creating discontiguous selections natively IIRC, but no other browser does.

In order to devote attention to supporting this in the client, we'd need to have some compelling use cases though.

jibe-b commented 6 years ago

Here is a use-case:

the manual annotation of named-entities in order to constitute a supervised machine learning training set requires non-continuous annotation of sequences of text.

Named-entity recognition by text-mining requires to annotate strings with the class to which the string is an instance.