Clement Jonquet et al.
It’s hard for people to find data – annotating data with ontologies is a solution, but which ontology to use? There are many different formats, platforms, versions. Which is relevant to you? What happens if you get it wrong? Here’s where the recommender comes in. The NCBO Annotator workflow extracts annotations from text by concept recognition, expands annotations using knowledge in the ontologies, and then score annotations according to their context and return them to the user. They use a dictionary approact, which is a list of strings that identifies ontology concepts.
For the semantic expansion, use the original ontologies’ is_a hierarchy, and use mappings in UMLS Metathesaurus and NCBO Bioportal, and use semantic similarity algorigthms based on the is_a graph (ongoing work). It has a nice-looking web interface that has a resulting ontology tag cloud – a good way of displaying the results. In the results, you give high scores for big ontologies, and you identify key ontologies. Some ontologies appear only with a specifric type of data, and it makes it important to have an appropriate recommendation. The score does not follow linearly the number of annotations. In future, they want to enhance backend annotation workflow, and have different types of scoring methods. They’d also like to parameterize the scoring methods.
FriendFeed discussion: http://ff.im/4x1mo
Please note that this post is merely my notes on the presentation. They are not guaranteed to be correct, and unless explicitly stated are not my opinions. They do not reflect the opinions of my employers. Any errors you can happily assume to be mine and no-one else’s. I’m happy to correct any errors you may spot – just let me know!