1) The University of Hawaii system faced budget cuts, requiring data-driven budgeting for colleges based on key performance metrics for faculty and staff research, teaching, and service. 2) The College of Education captured faculty publications, grants, and service data in an OpenLDAP directory to support the college's case for funding. 3) A reporting API allows querying the LDAP data and returning results in various formats like XML, JSON, and PDF to analyze and report on faculty research metrics.

1. Making research « social »
with LDAP
Stephan Fabel
Logo

2. Aloha :-)
2
Logo

3. Situation

University of Hawaii System:
– 10 campuses

UH Manoa campus 20,000 students

Budget cuts across the entire UH System

UHM alone >$24M in 2013


Budget 2014 for the first time 51% based on
student tuition
2014 Budget for Colleges « data informed »
3
Logo

4. Challenge for our College
How do we capture key performance
metrics for faculty and staff to
support our case?
4
Logo

5. COE and Symas OpenLDAP (1)


Symas OpenLDAP crucial to College
infrastructure services
Pre-2011:
– user accounts (College-specific)
– groups (College-specific)

2011-2012:
– pass-through authentication to central IT
– groups (POSIX and groupOfNames) local
– authentication and groups in every application
rolled out at the College
5
Logo

6. COE and Symas OpenLDAP (2)

2013:
– Publications (100%)
– Grants (specification done)
– Service (currently being specified)

No teaching activity stored in our directory
– data available through Banner (Oracle)
– but it's on the list

We run our own Student Information System
which helps
6
Logo

7. Capturing Research



« pet project »
originally aimed at marketing efforts through
public website
idea was to present college research to
interested third parties
– Legislature
– General public
– Prospective students
– Other researchers
7
Logo

8. Public Website (1)
8
Logo

9. Public Website (2)
9
Logo

10. Aloha :-)
10
Logo

11. Schema
11
Logo

12. publications.schema ?

Dublin Core Schema:
http://tools.ietf.org/html/draft-hamilton-dcxl-02

We implemented it

We didn't like it:
– Distinction between authors, contributors,
editors not clear enough
– Everything a DirectoryString
– Goal was to be able to generate APA-style
citations: not possible using Dublin Core
12
Logo

13. publications.schema (1)

26 attributes capturing:
– Title Information, Author(s), Abstract, Type,
Publisher, Volume, Pages, Owner, Venue,
Location, Organization, Editor, Series, Edition,
Chapter Information, Thumbnail, PDF, Month
and Year
– Keywords, Flag for outstanding research

8 object classes (pubObject)
– Conference Proceedings, Journal Article,
Book, Book Chapter, Presentation, Research
Report and Multimedia Contribution
13
Logo

14. publications.schema (2)


classes are auxiliary
meant to be used in conjunction with the
document structural object class
– documentIdentifier
– documentAuthor
– documentLocation

For the most part, tried to keep logical
attributes away from pubObject
– with few exceptions
14
Logo

15. Determining Author- and Ownership
pubObject
documentAuthor : uid=firstAuthor
documentAuthor : uid=secondAuthor
cn : [uidNumberFirstAuthor]XXX
pubOwner : uid=firstAuthor
Goal :
- determine authorization to edit
- only first author gets rw, all others only get r
- thankfully first author never changes
XXX is
incrementing
number
Yes
It's redunant
:-/
Show all work from uid=sfabel :
(pubOwner=uid=sfabel)
Show all work where uid=sfabel was involved : (documentAuthor=uid=sfabel)
Logo

16. Document Identifier


cn locally unique
documentIdentifier supposed to be
globally unique
– DOI - http://dx.doi.org/
– ISBN - http://books.google.com/


We don't want to save the publications
themselves (copyright issues)
We link them based on DOI through our library
→ paywall if not part of our system, otherwise
direct access
Logo

17. Lessons learned / Still todo


Capture organizations as DN
How to organize this in hierarchical fashion
across multiple, distributed servers
– Change
– Federated access

Other things we're not aware of
17
Logo

18. Aloha :-)
18
Logo

19. Reporting API
19
Logo

20. Reporting API

Written in PHP

ReST based queries (HTTP)

Binds to LDAP server and executes search

Returns data in
– XML, JSON, PDF, CSV
– Net file

Currently no authentication layer
– Looking at possibly using OAuth 2.0
20
Logo

21. Publications by Person (1)
21
Logo

22. Publications by Person (2)
22
Logo

23. Publications by Person (3)
23
Logo

24. By Person → By Department

Using groupOfNames

Using slapo-memberof(5)


First Author is member of department →
publications can be aggregated
Relationship is dynamic (author moves to
different department, so do his/her
publications)
24
Logo

25. Publications by Department (1)
25
Logo

26. Publications by Department (2)
26
Logo

27. Publications by Department (3)
27
Logo

28. Publications by Department (4)
Bonus !
28
Logo

29. Expert Search

Goal is to find the person with the highest
caliber in publications around a given topic

Based on pubKeyword attribute values

Output is people (not publications!)

Ranking is performed by
– Publication count, type, # of collaborators
– Whether person was first author or not
29
Logo

30. Keyword Search (1)
30
Logo

31. Keyword Search (2)
Person claims « autism » as area of interest, which
guarantees being listed, but we have no publications in
our system to indicate any value of his contribution.
31
Logo

32. Aloha :-)
32
Logo

33. So, how is it « social »?
33
Logo

34. What is « social » ?

Social media:
– share information with networks of people
– interaction based on that shared information
– goal: create « virtual community »
34
Logo

35. What makes research « social » ?

Social research:
– topically bounded interaction based on shared
information
– networks emerge through work
– communities « pre-defined »:
•
•
•
•
fellow researchers
prospective students
public/legislature
administration
35
Logo

36. Collaboration ↔ Interaction

Collaboration Report:
– Find author pairs, calculate their “weight”
– Create score based on these weights


Total relevance score average of all
co-authors importance
Scoped by keyword or global
36
Logo

37. Collaboration Report (1)
37
Logo

38. Collaboration Report (2)
38
Logo

39. Collaboration Networks


Combination of expert and collaboration
search
Undirected graph:
– Nodes people, size indicating weight
– Vertices collaborative relationship, size
indicating strength of collaborative efforts
(number of publications, kinds of publications,
number of other collaborators, etc.)
39
Logo

40. Collaborative Network (example)
40
Logo

41. Outlook / Future Work




Comprehensive Dashboard in planning
Additional institutional research / business
intelligence based on further analysis of
collaboration networks
Web-enabled search interfaces available to
public Q1 2014
Internal reporting to be made available to all
colleges, aggregation of LDAP servers to
provide campus-level reporting
41
Logo

42. Outlook / Future Work

Organizations:
– Common thread between publications, grants,
awards and service data
– Will be central in future reporting

Portal for researchers:
– Finding other people that you haven't
collaborated with
– Leveraging success of grant applications
through collaboration
– Providing orientation for new hires
42
Logo

43. Aloha :-)
43
Logo

44. Thanks!
Logo