{i} Review and cohese user studies
Closed this issue · 4 comments
┆Issue is synchronized with this Asana task
Danielle:
"Iwould use IE-Viz to define fever from clinical text. Specifically, thetask is to classify mentions as fever_present or fever_absent at the mention(quantifications and qualifications) and then document-level.
The target classes are TEMPERATURE, FEVER. The modifierclasses are QUANTIFICATION, INCREASED, HIGH_LEVEL, LOW_LEVEL.
The classification rules:
· TEMPERATURE + QUANTIFICATION -->fever_present
· FEVER + QUANTIFICATION --> fever_present
· FEVER + DEFINITE EXISTENCE--> fever_present
· TEMPERATURE + HIGH-LEVEL--> fever_present
· TEMPERATURE + LOW-LEVEL--> fever_present
· TEMPERATURE + INCREASED --> fever_present
· FEVER + HIGH-LEVEL --> fever_present
· FEVER + LOW-LEVEL --> fever_present
· FEVER + INCREASED --> fever_present
Else:
· FEVER + negation| historicity | future |indication --> fever_absent
Atthe document-level if a patient ever had a fever_present, I would flag forinclusion in my cohort."
[--- Commented from Asana.com
#commenter Bill Scuba
---[aa]
Garrett:
“
I am a Neurologist at the VA whowill be doing retrospective operational research concerning the most commonsigns and symptoms among veterans found during an initial neurological exam.The overall intent is to find which areas need the most attention when trainingdoctors and nurses with this veteran-specific population as well as to increasequality of care and readiness at the locations of initial exams. I don't have asolid list of terms and concepts yet because of the large number ofpossibilities. I would like to be able to input my directory of exam notes andbe given a list of common concepts, possibly sorted by number of occurrencesand severity of concept, of which I could then have a chance to go through andremove the less relevant ones myself. The NLP needs to be able to accuratelyreturn assertion numbers in a large number of semi-structured SOAP notes andoutput professional looking graphs or other visualizations appropriate forpapers and grant proposals, or at least output the data in a format wherevisualizations could easily be made with another program. Because I will beworking with VA data, it must be able to run on the VA's VISTA system.
“
[--- Commented from Asana.com
#commenter Bill Scuba
---[aa]
Wendy:
"I am using pyConText and need to expand the lexicon for targets. Iwould like to use Knowledge Author to 1) quickly and easily look at the typesof sentences identified by the regular expressions in my lexicon, 2) suggestsynonyms based on mining real data – it would be great if I could also point tomy own data if I have some, 3)show me word trees from the data that help mefigure out good terms.
I would like it to be able to read in my lexicon, support me ineasily adding new terms, deleting bad ones, and modifying existing terms, andgenerate regular expressions for the terms I select using Knowledge Author sothat they can be read into PyConText."
[--- Commented from Asana.com
#commenter Bill Scuba
---[aa]
Jianlin "I think one important case is that when a user is using arule-based system and trying to develop the rules to make the system runagainst the gold standard. I think it would be useful to ask users if they wantto show the changes that the new rules make. It will be useful to list the newfalse positive and false negative separately.
Probably the user would like to check the processeddifference after make a few rules. Then it will be useful to locate the falsecases with the corresponding rules."
[--- Commented from Asana.com
#commenter Bill Scuba
---[aa]