Hi Vicky There are two Time related annotation types and neither of them are incredibly useful on their own and there's no documentation that I've found, so far. What I've deduced is by trial and error with short phrases, trying to reverse engineer the behavior
TimeMention attempts to link a time/date assertion with one or more events. They use a pointer system where the Mention, whose range is a time or date expression, has a unique ID that is pointed to from a "TemporalTextRelation" node that links the time with an event through another ID. At the same time, there will be an event object (or sub object of one of the major semantic types like SignSymptomMention, which is annotated with "before" or "after" another Event. Something like "A occurred <before/after/during> B through time mention T DateAnnotation is a more simplistic Annotator that just identifies the range in which a date or time assertion occurs. Unfortunately neither are completely reliable, and you might get false negatives or false positives with either. Sometimes clinical statements can also be ambiguous and might be a pair of lab values interpreted as a date If you have both of these annotators in place, and you parse a phrase such as "On 12/24/1995 he had an embolism", you will get both annotation types. I suggest drawing out the pointer relationships from the output. I take the serialized CAS and pass it through a rather extensive post-processing phase (written in ruby) where I use both annotation types, a ruby temporal gem "Chronic" and some magic with the ConllDependency nodes to try to get a useful date in most cases, where it may not be obvious. Imagine the following "Patient admitted 4/6/2003 for observation. He will be sent to the Cath lab tomorrow" While the TimeMention correctly identifies "tomorrow", it doesn't interpret what absolute date it refers too. This is where one can try to infer the date as 4/7/2003 with post-processing tricks. It can get very tricky. Peter On 9/12/16, 2:48 AM, "Vicky Chawda" <vicky.cha...@praxify.com> wrote: >Hi, > >I am using "Ctakes-Temporal" >Link : >https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_healthnlp_ >examples_tree_master_ctakes-2Dtemporal-2Ddemo&d=DQIBaQ&c=B73tqXN8Ec0ocRmZH >MCntw&r=5LM1YwNyMUq7CWiSepCCsjTjwuVF4uswNF8BK5Orm10&m=6DaS2ZEk2UDo6JWBiHqY >5n-R2kteT52ZrkFYAQUCCJ8&s=tEj7xdiF4QISneiA7Qh1VNu1FYLoHEfYhTTMk4AMGuY&e= >https://urldefense.proofpoint.com/v2/url?u=http-3A__52.27.22.206-3A8080_in >dex.jsp&d=DQIBaQ&c=B73tqXN8Ec0ocRmZHMCntw&r=5LM1YwNyMUq7CWiSepCCsjTjwuVF4u >swNF8BK5Orm10&m=6DaS2ZEk2UDo6JWBiHqY5n-R2kteT52ZrkFYAQUCCJ8&s=wEMSC-0bNN6h >xP6VeyTddfUmTA8ZEcbYzesAfIXtL7Y&e= > >Can anyone please suggest how to interpret the XML response? is there any >documentation available for it? > >ViCky