Six months ago we here at the Hesburgh Libraries finished writing up our final report regarding an IMLS grant on the theme of topic modeling, cross-disciplinary research, and machine learning, and below is a statement of our goal, our findings, and our recommendations:

  Goal - To understand the unique current practices of domain
  experts, librarians, and computer science specialists and to
  identify possibilities to use topic modeling and NLP to enhance
  or augment current library classification in order to meet
  current cross-disciplinary research needs

  Findings: 1) Interest in machine learning is high and appears to
  be on a precipice, 2) The biggest issues with cross-disciplinary
  research are not discovery related, 3) There is a high need for
  interdisciplinary collaboration, 4) Community effort for greater
  ROI, 5) "Garbage in, garbage out,"; machine learning requires
  good data, 6) Ethics are a really big concern for machine
  learning, especially regarding bias, and 7) there is a need for
  greater machine learning literacy

  Recommendations: 1) Increase the the community, 2) Develop
  machine learning education for scholars and library
  professionals, 3) Form learning communities and networks, 4)
  Create and curate a clearinghouse for machine learning models, 5)
  Support consortia around subject strength to develop machine
  learning tools, 6) Develop processes to enhance discovery tools,
  and 7) Support diversified machine learning innovations.

As a bonus, we also published a freely available edited volume of fourteen essays on machine learning, entitled "Machine Learning, Libraries, and Cross-Disciplinary Research: Possibilities and Provocations". From the preface:

  The resulting essays cover a wide ground. Some present a
  practical, "how-to" approach to the machine learning process for
  those who wish to explore it at their own institutions. Others
  present individual projects, examining not just technical
  components or research findings, but also the social, financial,
  and political factors involved in working across departments (and
  in some cases, across the town/gown divide). Others still take a
  larger panoramic view of the ethics and opportunities of
  integrating machine learning with cross-disciplinary higher
  education, veering between optimistic and wary viewpoints.

For more full access to the final report as well as edited volume, please see:

  1. project home page - https://innovation.library.nd.edu/crossdisciplinary-research/
  2. final report - https://curate.nd.edu/downloads/7p88cf98m4g
  3. edited volume - https://curate.nd.edu/downloads/f1881j95p8f

We here at the Libraries thought some of you here on the mailing list might be interested in these topics, thus FYI.

-- 
Eric Morgan for Team IMLS Machine Learning Grant
Hesburgh Libraries
University of Notre Dame

574/631-8604