Hi Wendy,
Sharing with GPC-DEV that may have interested participants.

Russ

From: Wendy W Chapman [mailto:[email protected]]
Sent: Monday, July 11, 2016 11:02 PM
To: David Carrell; [email protected]; Mei Liu; Matthew Hoag; Russ Waitman; 
[email protected]; [email protected]; [email protected]; 
[email protected]; [email protected]; [email protected]; Wendy W 
Chapman; [email protected]; [email protected]; [email protected]; 
[email protected]; [email protected]; 
[email protected]; William Karl Thompson
Cc: Rachel Hess
Subject: PCORI CRG (collaborative research group)

CDRN NLP colleagues,

As you know, we are postponing the NLP workshop until September. In the 
meantime, we are seeking CDRN collaborators for CRG in Natural Language 
Processing (NLP) led by Dr. Wendy Chapman and PaTH. If you are interested, 
please fill out the following brief survey: https://goo.gl/CosEgj

We have also sent this out to the PIs of the CDRNs and PPRNs (we have only 
received two responses so far).

Here is the text we plan to submit - any suggestions appreciated.

Thank you very much!

Best,

Rachel, Kathleen, and Wendy
----------------------------
Many of the variables required for phenotypes of interest in PCORI are locked 
in clinical text resulting in cohorts that are less accurate than they could be 
or abandonment of important studies due to the inability to identify patients. 
Natural language processing (NLP) has been successfully applied for cohort 
identification but requires specific expertise to develop and apply. Not every 
CDRN or PPRN has access to that expertise. Moreover, even if possible, it would 
be inefficient and redundant to develop NLP algorithms within each CDRN. We 
propose a CRG for NLP as a strategic opportunity to provide guidance and 
services to CDRNs and PPRNs seeking to extract information from text, to 
support NLP research within the networks, and to begin working towards a 
unified vision of how NLP might be supported and applied within PCORI. We 
envision an NLP CRG to bring together two types of members: members with NLP 
expertise wishing to provide educational and other services across PCORI and 
members wishing to learn more and apply NLP within their CDRN or PPRN. We 
envision an NLP CRG that provides the following to participating CDRNs and 
PPRNs:


1)      Provide educational opportunities for members to learn about NLP 
methods and steps required for application and evaluation of NLP on clinical 
text. Educational opportunities could include webinars for large groups of 
participants and hands-on activities and workshops for small groups of 
participants (e.g., tutorials and participation in summer workshops, such as 
the Data Science summer school at the U of Utah).

2)      Support NLP research projects within the networks involving NLP. We 
will provide individual consulting to members on design, application, and 
evaluation of NLP activities. Consulting could include suggestions on the most 
suitable NLP tools to apply, guidance on tool customization, and direction on 
evaluation of tool performance. In addition, consulting could include 
activities such as development of guidelines and annotation schema, training of 
on-site annotators, and guidance on study design suitable for publication.
NLP research within CDRNs will likely need to be local activities at single 
institutions, due to lack of mechanisms for sharing data  within a network. The 
NLP CRG will support the development of local NLP research activities at 
individual activities while also working towards mechanisms to support NLP 
across institutions. A sub-group of the CRG interested in strategic visioning 
will meet regularly via Webex to discuss current techniques for applying NLP 
across research networks, new technologies and resources available for 
collaborative NLP development, privacy preservation in collaborative NLP 
activities, and perceived feasibility of different approaches within PCORI. 
This CRG will be a driver for new methodology and research development for the 
challenging problem of processing text across institutions while preserving 
privacy. Possible inter-institutional activities include deploying the same 
software at different institutions or de-identifying datasets to be centrally 
processed.




_______________________________________________
Gpc-dev mailing list
[email protected]
http://listserv.kumc.edu/mailman/listinfo/gpc-dev

Reply via email to