Apache cTAKES- NLP in Healthcare
0 likes331 views
The document provides an overview of Apache cTAKES, a natural language processing (NLP) system designed for analyzing healthcare data. It discusses key features like named entity recognition related to medications and procedures, and outlines various algorithms and pipeline structures. Additionally, it encourages community involvement and contributions to the project, emphasizing opportunities for support and development in NLP within the healthcare industry.
![Natural Language Processing (NLP)
3
“A way for computers to analyze, understand and derive
meaning from human language” - algorithmia [1]
[1] - https://blog.algorithmia.com/introduction-natural-language-processing-nlp/
● Feasibility
Big Data / Machine Learning / Apache Projects
● Challenges
Ontology / Specialization / Anonymization](https://image.slidesharecdn.com/alexzbarcea-apachectakes-nlpinhealthcare-191010010709/85/Apache-cTAKES-NLP-in-Healthcare-3-320.jpg)
![● Approaches:
Extraction
Generation
● Algorithms:
Rule-based
Machine Learning
● Linguistic annotations:
Penn TreeBank [1]
GENIA [2]
How it works
4
corpus
[1] - https://www.clips.uantwerpen.be/pages/mbsp-tags
[2] - https://orbit.nlm.nih.gov/browse-repository/dataset/human-annotated/83-genia-corpus](https://image.slidesharecdn.com/alexzbarcea-apachectakes-nlpinhealthcare-191010010709/85/Apache-cTAKES-NLP-in-Healthcare-4-320.jpg)
![● Documentation (confluence [3])
● ctakes-examples
● main Classes
[alex ~/ctakes {trunk %} ]$ grep -nRI --include="*.java" "main(String[] args)" | wc -l
171
cTAKES: Exploring Examples
9
[1] - https://builds.apache.org/analysis
[2] - https://builds.apache.org/view/C/view/Apache%20cTAKES/
[3] - https://cwiki.apache.org//confluence/display/CTAKES
● smokingstatus
● coreference
● NexEx
● pipelines
● training
● temporal
● relationextractor
● etc
● Run on real data (i.a. LibreHealth / OpenEMR)](https://image.slidesharecdn.com/alexzbarcea-apachectakes-nlpinhealthcare-191010010709/85/Apache-cTAKES-NLP-in-Healthcare-9-320.jpg)
![Apache cTAKES Demo
10 [1] - https://github.com/azbarcea/ctakes-examples](https://image.slidesharecdn.com/alexzbarcea-apachectakes-nlpinhealthcare-191010010709/85/Apache-cTAKES-NLP-in-Healthcare-10-320.jpg)
Apache cTAKES- NLP in Healthcare
- 1. Apache cTAKES NLP in Healthcare Alex Zbarcea (FannieMae / cTAKES committer)
- 2. 2 Episode of Care medications imaging pathology inpatient services and procedures outpatient services and procedures medications imaging pathology inpatient services and procedures outpatient services and procedures research EMR notes
- 3. Natural Language Processing (NLP) 3 “A way for computers to analyze, understand and derive meaning from human language” - algorithmia [1] [1] - https://blog.algorithmia.com/introduction-natural-language-processing-nlp/ ● Feasibility Big Data / Machine Learning / Apache Projects ● Challenges Ontology / Specialization / Anonymization
- 4. ● Approaches: Extraction Generation ● Algorithms: Rule-based Machine Learning ● Linguistic annotations: Penn TreeBank [1] GENIA [2] How it works 4 corpus [1] - https://www.clips.uantwerpen.be/pages/mbsp-tags [2] - https://orbit.nlm.nih.gov/browse-repository/dataset/human-annotated/83-genia-corpus
- 5. 5 Apache cTAKES: Overview plain text CDA Named Entity * drug * disease/disorder * sign/symptom * anatomical site * procedures Pipeline based - combining techniques: ● Rule-based ● Machine Learning (ML) Java, Modular Measurable performance (standard)
- 6. Boundary detection Tokenization Normalization (Lemma) Part-of-speech Shallow parsing Entity recognition cTAKES System 6 NLM Apache OpenNLP SPECIALIST NLP Tools Apache Lucene UMLS, SNOMED-CT, RxNORM ICD10/9, Mayo Clinic, Custom
- 7. Tasks in NLP (cTAKES example) 7
- 8. cTAKES: Pipelines 8 (e.g. examples/pipeline/ProcessDir.piper ) // This file contains commands and parameters to run the ctakes-examples "Hello World" pipeline readFiles org/apache/ctakes/examples/notes // Load a simple token processing pipeline from another pipeline file load DefaultTokenizerPipeline.piper // Add non-core annotators add ContextDependentTokenizerAnnotator // Collect discovered Entity information for post-run access collectEntities https://cwiki.apache.org/confluence/display/CTAKES/Piper+Files
- 9. ● Documentation (confluence [3]) ● ctakes-examples ● main Classes [alex ~/ctakes {trunk %} ]$ grep -nRI --include="*.java" "main(String[] args)" | wc -l 171 cTAKES: Exploring Examples 9 [1] - https://builds.apache.org/analysis [2] - https://builds.apache.org/view/C/view/Apache%20cTAKES/ [3] - https://cwiki.apache.org//confluence/display/CTAKES ● smokingstatus ● coreference ● NexEx ● pipelines ● training ● temporal ● relationextractor ● etc ● Run on real data (i.a. LibreHealth / OpenEMR)
- 10. Apache cTAKES Demo 10 [1] - https://github.com/azbarcea/ctakes-examples
- 11. Apache Software Foundation ● Community ○ Linguist experts Users Developers ● Mature Software Lifecycle ○ Support Issues SCM - Collaboration Jenkins Sonar Distribution ● Popularize 11
- 12. Get involved (You don’t need to be a software developer) 12 ● Help new users and provide feedback ● Give feedback on required features ● Write or Update documentation ● Test the code and report bugs ● Fix bugs ● Write and update the software ● Create artwork ● Extend docs references ● Recommend the project to others ● Gamification ● Volunteer valuable skills ● Learn about communities - the Apache Way ● Requirements Engineering ● Learn about NLP and Healthcare ● What a strong product is about ● Test Automation and Software Engineering ● Develop code with high quality ● Build strong Software Development skills ● Explore your creativity ● Marketing ● Use your time wisely ● Help research community git: https://github.com/apache/ctakes wiki: https://cwiki.apache.org//confluence/display/CTAKES e-mail: https://ctakes.apache.org/mailing-lists.html
- 13. Thanks! Any questions? You can find me at: ● https://linkedin.com/in/azbarcea ● alexz@apache.org 13