Hi Peter, In short (I have to run right now):
I am really glad that your employer will give you time to work with ctakes! #1 The jdk 11 build compatibility was a necessity. It turned out that the Apache Jenkins system no longer has a version of maven built with java 8. They seem to also be bumping up their lowest version of java. Unless I just couldn't find one. It is still a selectable choice in the Jenkins configuration wizards, it just doesn't actually exist. According to jira reports others found the same thing. So, the only solution was for me to make ctakes buildable with jdk 11 and maven built with jdk 11. However, it builds through java 11 as java 8 compatible byte code. That was done for our CI on Jenkins. I can still build ctakes with jdk 8. #2 I am not certain what issues you had. Unfortunately my employer's email system mangles anything that looks like a link, and they become completely unreadable. #3 Dennis Johns and I are trying to get the 5.0 wiki into shape, including differences between versions 4 and 5. You guessed right about pbj. It stands for "Python Bridge to Java" and uses artemis as a go-between for ctakes and python pipelines. Though it allows python to java, it can be used pretty much any way you would imagine. java to java, python to python, java-python-java, python-java-python, scaling out, joining, splitting tasks between different systems, etc. There are a few java-python-java examples in the ctakes resources, one of which uses some python from our friends at cnlpt (https://github.com/Machine-Learning-for-Medical-Language/cnlp_transformers). Unfortunately dictionary lookup doesn't have any recent major improvements. Sean ________________________________ From: Peter Abramowitsch <pabramowit...@gmail.com> Sent: Friday, May 26, 2023 1:46 PM To: dev@ctakes.apache.org <dev@ctakes.apache.org> Subject: What's new in 5.0 && testing JDK 11 [EXTERNAL] * External Email - Caution * Hi Sean, It looks like I may get some support from my employer to explore 5.0 this summer, and while doing so, also test the jdk11 build, but I have a couple of quick questions. 1. If the system would still require 1.8 to run due to certain dependencies, what would be the advantage of building it under 11? - or were you suggesting that an 11 runtime would be possible by upgrading those dependencies too. 2. In building the complete 5.0 from git, I've run into a problem with maven blocking certain artifacts due to http/https issues. There are global fixes and project by project specific fixes. Which do you recommend? Ideally should maven be run with -o? *[INFO] --- maven-remote-resources-plugin:1.4:process (default) @ ctakes-core ---Downloading from maven-default-http-blocker: https://urldefense.com/v3/__http://0.0.0.0/org/apache/ctakes/ctakes-models/5.0.0-SNAPSHOT/maven-metadata.xml__;!!NZvER7FxgEiBAiR_!rSuvDzYcI-VTRvyiFTaFxr9HI_L4MzFigaaWnvS0cqKIaLEF7BMIeepWBtVYMLmRMVIiIlcrud1QTdtlm_M70sb8DEL1-0WrqQ$ <https://urldefense.com/v3/__http://0.0.0.0/org/apache/ctakes/ctakes-models/5.0.0-SNAPSHOT/maven-metadata.xml__;!!NZvER7FxgEiBAiR_!rSuvDzYcI-VTRvyiFTaFxr9HI_L4MzFigaaWnvS0cqKIaLEF7BMIeepWBtVYMLmRMVIiIlcrud1QTdtlm_M70sb8DEL1-0WrqQ$ >* 3. Finally, I had asked a while back if someone could point me to a list of improvements or significant additions to cTakes that have occurred over the last year or so. Since no one responded, I decided to look at all the SVN and Git commit messages and diffing the sources. I did come across the PBJ project. The readme doesn't actually explain what it is for and there are various meanings of the term PBJ in the python community. This one looks like infrastructure to allow ctakes to be called from a python pipeline using Artemis to decouple the processes -- or am I wrong and it is the reverse (calling python from within a cTakes pipeline) If there are any areas where concept lookup has been improved through better semantic contextualization please let us know! Peter
