This is a read only archive of pad.okfn.org. See the
shutdown announcement
for details.
opendiscovery-mozfest
I didn’t know I could know so much! The open science way to do literature research
Mozfest, 30 Oct 2016, 11-12am, Room 902
Hashtags: #openscience #mozfest #okmaps
http://openknowledgemaps.org/
Shared Notes
For a field completely new to me, I'd like an overview and list of jargon first, so I start at Wikipedia before I go to Google Scholar (@steltenpower)
Questions
- How good is the clustering/similarity algorithm?
- It uses Wards Method of minimum variance, currently we have only human judgement for evaluation, since it is unsupervised there is no direct metric for evaluation - we're thankful for suggestions and also feedback is some clustering appears incorrect
- What ranking is used for the list on the right?
- The Pubmed Search currently provides sorting by citations, title, author or year
- Is this based on full text analysis, abstracts or titles?
- the grouping of papers is based on the subject keywords of papers. In cases where they are missing from the metadata, we approximate them from abstract and title
- Considering sources/coverage is it more useful in certain fields than in others?
- currently Pubmed and DOAJ are available, Pubmed is more useful in bioscience and medical fields, DOAJ more useful for humanities and social sciences - but some overlap always exists
- Is there a danger in 'just' using the top 100 for the analyses? Does that bring in limitations of the ranking of the source databases used?
- Does it also use some structured text capabilities, like linked data, etc. ?
- At this stage there are no structured capabilities - if you have some specific ideas or suggestions, please raise an issue on https://github.com/pkraker/Headstart and we can discuss more
- The keywords or (better) concepts found as something in common on many papers, could be explained ...
Suggestions:
- Please provide a Tor hidden service and full Tor Browser support!
- Checkboxes next to each paper so that at the end I can export all of the checked papers to my reference manager such as Zotero or BibTex.
- As above, a kind of 'favourites' option. Complies a list of the papers you were most interested in.
- Be able to make your own knowledge maps. Could then share these with collegues as a visual representation of research you've read
- Specifically the ability to manually drag and drop papers around and define the relationships between them.
- document/display applied search filters somewhere on the visualization as a quick reminder of "what did I search for again?"
Annoying feature requests ;-)
- add doi to article information displayed
Late participant suggestion! For Pubmed: could you: do a visualisation of the % of diseases/conditions that exist, for which there is an Open Access pubmed-published research paper? You could compare PubMed to: http://www.who.int/classifications/icd/en Thanks!