Archive for the ‘Uncategorized’ Category
Facetted navigation: enabling powerful search strategies with a few, simple mechanisms
I am currently developing a prototypical facetted search application for aTags. See
http://hcls.deri.org/atag/search/
(I repeat: very prototypical)
This contains all entities from DBpedia, as well as hundreds of thousands of aTags aggregated from the web of linked data.
Type in “ginkgo” in the search box. You should see some aTags that have been fetched from
http://hcls.deri.org/atag/data/tcm_atags.html
Results are ranked based on relevance measures (using Lucene/Solr).
The ‘tags’ cloud lists entities that are contained in the current result set. Click on ‘Therapy’ (alternatively, you can type it in the search field and use autocompletion). The result is narrowed down (’zoomed in’) to a few aTags that propose Ginkgo as a therapy for some kind of ailment.
You can then remove the filter for the “Therapy” tag by clicking on the (x) on the upper left (’zoom out’). Or you can decide to navigate to related tags, for example by clicking on the ‘memory’ tag beneath one of the aTags (’panning’ to assertions that are related in some way).
Since we are using entities from ontologies / the linked data cloud, we can also make use of this large amount of background knowledge. Start with a fresh search on ‘ginkgo’, then look at the ‘Broader tags’ tag cloud on the left. This tag cloud is populated with entities that are ‘broader concepts’ of the entities used in the aTags in the result set, based on the hierarchy in DBpedia. Clicking on ‘disease’, for example, narrows the result set to aTags that mention some kind of disease (such as ‘Attention-deficit hyperactivity disorder’).
The implementation here is still very rudimentary, but I think that this kind of facetted navigation based on ontological background knowledge could be very powerful when fully realized. You could, for example, run a reasoner over a set of chemical compounds, let them be classified based on background knowledge (e.g., classifying them under ‘drug that is currently researched in a clinical trial’ based on data from http://linkedct.org), and then search for statements based on this inferred information.
The interesting thing here is that it is sufficient to offer the user a small set of mechanisms (typing in text, selecting tags, removing tags, clicking tags) that are easy to learn and predictable for most kinds of datasets. Nonetheless, by enriching the underlying taxonomy with hierarchies that have been asserted in ontologies / linked data resources, or that have been inferred from these ontologies / data, we can achieve very impressive results that would not have been possible without these technologies.
Some aTags about neuropharmacology etc.
Below I have collected some interesting statements from research papers I recently stumbled upon. They are encoded as aTags.
| “Huperzine A acts as a non-competitive antagonist of the NMDA receptors” (Source) |
| “some effects of CDP-choline could be mediated by changes in brain platelet-activating factor (PAF) levels” (Source) |
| “Changes in brain striatum dopamine and acetylcholine receptors induced by chronic CDP-choline treatment of aging mice” (Source) |
| “changes in ERK phosphorylation in hippocampus and PFC were regulated by GABAA receptor in a learning and memory paradigm under acute restraint stress conditions” (Source)|
| “our data suggest actions of memantine beyond NMDA receptor antagonism, including stimulating effects on cholinergic signalling via muscarinic receptors” (Source)|