Microsoft does Scholar.... Visualising Literature...

At the recent eResearch Australasia Conference, a visualisation contest brought me into contact with Microsoft Academic Search (MAS).

MAS provides an API to interrogate the database of 38M articles programmatically. Including access to title, author, journal, abstract, keywords, citations. See User Manual here.

Cyrus from Versi and I proposed a service where a tag cloud is generated from abstract text from articles with a certain keyword. Here is my text to a Microsoft reviewer after the competition. Kudos to David Flanders for passing this on: -->mail begins...

The basic pitch is to assist researchers to do literature reviews with targetted tag cloud summaries of relevant articles.


1. By topic (provided by MAS)
2. By abstract text. We had trouble with the API getting to the abstracts.

ie extract abstract text by API, and build tag cloud. Each word links to list of articles with an abstract containing the selected text fragment.

3. By keyword. We had trouble accessing publication keywords with the API.

ie extract keywords for target articles by API, and build tag cloud. Each keyword links to list of articles with selected keyword.

This is similar to the co-author path, but for keywords.
ie this code gave us blanks:

http://academic.research.microsoft.com/table.svc/search?AppId=&FullTextQuery=data+mining&ResultObjects=publication&PublicationContent=Keyword&StartIdx=1&EndIdx=10

My APP-ID: xxx [Microsoft issued me an API within 24 hours. Fast service.]
Benefits: The MAS API has a lot of potential to automate (do the heavy lifting) for researchers saving potentially many hours. My demo was aimed to replicate a three day literature review I did recently (for corrections on my PhD; valman.blogspot.com) in close to sixty seconds. Saving researchers time like this could likely result in strong word of mouth recommendations for Microsoft Academic Search.

Any questions, please let me know. I enjoyed the faceted browsing of MAS, and hope the API will be improved to improve the stability of the functions we tried to get working.


Likes: faceted search, tabular output, access (theoretically) to abstract, keywords, author, title. Good access to references. A lot of potential going forward. Many articles. Good journals included in my field (Business, economics).

Dislikes: no community content in forum to answer api problems, difficulty getting API to work as documented (ie difficulty accessing keywords and abstracts). Little community activity on forum suggests need more promotion of tool. Looking at one journal (eg http://academic.research.microsoft.com/Journal/4773/ind-market-manag-industrial-marketing-management) it was unclear which selection criteria were used on articles displayed and in what order they were presented.

Keep up the good work. Lot of potential here...