Information extraction / Wikifier approach to enrich / categorise some material / technique property values

[tschleider]

We have a lot of datasets with mixed material / technique fields. Some of them (like MAD, e.g. this query) have mostly sentences that are not possible to parse with regex so we should use an approach like Wikifer to process their semantics properly.

[rtroncy]

The Wikifier SILKNOW UI available at http://wikifier.org/silknow.html can be used to do quick testing of what could be extracted from long sentences.

[rtroncy]

@tschleider Read actually the whole thread at https://lists.eurecom.fr/sympa/arc/silknow/2020-05/msg00015.html with the experiments made by Janez and the results stored at http://wikifier.org/silknow/JSI-all_text-wikified.zip (warning, this is a big 600MB JSON file that you best edit using Sublime or equivalent)