Skip to main content
v2026.11,986 entries · CC-BY 4.0

Apache cTAKES: Clinical NLP Architecture, UIMA Pipeline, UMLS Fast Dictionary Lookup & Negation Assertion

Technical guide to Apache cTAKES (clinical Text Analysis and Knowledge Extraction System): UIMA pipeline, UMLS CUI extraction, assertion attributes, and EHR use.

Written and maintained by CASRAI Editorial Board

Published

On this page

Biomedical informatics researchers, clinical data warehouse engineers, and computational phenotyping teams use this guide to configure Apache cTAKES (clinical Text Analysis and Knowledge Extraction System), understand its Apache UIMA annotator pipeline and Common Type System (CTS), map unstructured clinical text to UMLS SNOMED CT and RxNorm Concept Unique Identifiers (CUIs), and populate observational research databases.

What Is Apache cTAKES and How Is It Used in Clinical Research and Healthcare?

Apache cTAKES (clinical Text Analysis and Knowledge Extraction System) is an open-source, modular Java natural language processing (NLP) system designed specifically to extract structured clinical information from unstructured electronic health record (EHR) free text. Originally developed at Mayo Clinic by Guergana K. Savova, Christopher G. Chute, and colleagues (Journal of the American Medical Informatics Association, 2010) and promoted to an Apache Software Foundation (ASF) Top-Level Project in 2013 under the Apache License 2.0, cTAKES combines rule-based and machine-learning annotators built on the Apache UIMA (Unstructured Information Management Architecture) framework and the Apache OpenNLP toolkit.

In clinical research and healthcare informatics, Apache cTAKES is widely deployed across academic medical centers and consortia—including the Electronic Medical Records and Genomics (eMERGE) Network, Informatics for Integrating Biology and the Bedside (i2b2), and Observational Health Data Sciences and Informatics (OHDSI)—to automate cohort phenotyping, pharmacovigilance adverse event detection, tumor registry abstraction, and computer-assisted coding (CAC).

The Apache cTAKES Default Clinical Pipeline Architecture

In Apache cTAKES, clinical text processing is orchestrated through a sequential pipeline (defined via Piper files—.piper configuration scripts—such as DefaultFastPipeline.piper) where each UIMA Analysis Engine reads and enriches a shared in-memory Common Analysis Structure (CAS):

Pipeline Stage & Module Primary Annotator Class / Component Clinical NLP Function & Output Annotations
1. Section & Sentence Segmentation BsvRegexSectionizer / SentenceDetector Splits clinical notes into HL7 CDA / clinical header sections (HPI, Past Medical History, Family History, Medications) and detects sentence boundaries using OpenNLP maximum-entropy models trained on clinical text
2. Tokenization & Lexical Normalization TokenizerAnnotatorPTB & LvgAnnotator Segments sentences into Penn Treebank-style tokens (`WordToken`, `NumToken`, `PunctuationToken`) and applies NLM Specialist Lexicon Lexical Variant Generation (`LVG`) to normalize inflections to canonical base forms
3. POS Tagging & Shallow Parsing POSTagger & Chunker Assigns part-of-speech tags and groups tokens into syntactic noun phrases (`NP`), verb phrases (`VP`), and prepositional phrases (`PP`) to bound candidate clinical concepts
4. Fast Dictionary Lookup (NER) ctakes-dictionary-lookup-fast (DefaultJCasTermAnnotator) Uses rare-word indexing over an embedded HSQLDB Unified Medical Language System (UMLS) database (`sno_rx_16ab` or custom UMLS subset) to identify clinical entities and assign UMLS CUIs, SNOMED CT codes, and RxNorm codes
5. Context & Assertion Attribution ctakes-assertion / ContextAnnotator (NegEx / MITRE ML) Evaluates surrounding linguistic context for every `IdentifiedAnnotation` to set six critical clinical modifier attributes: polarity, uncertainty, conditional, generic, subject, and historyOf
6. Dependency Parsing & Temporal Extraction ClearNLPDependencyParserAE & ctakes-temporal Extracts anatomical locations (`locationOf`), severity (`degreeOf`), clinical events (`EventMention`), time expressions (`TimeMention`), and Document Time Relations (`DocTimeRel`: BEFORE, OVERLAP, BEFORE/OVERLAP, AFTER)

The cTAKES Common Type System (CTS) and Assertion Attributes

Unlike general-purpose NLP frameworks that emit untyped string spans, Apache cTAKES enforces a formal, strongly typed UIMA Common Type System (CTS) aligned with clinical semantic groups (`org.apache.ctakes.typesystem.type.textsem`). Every extracted clinical entity inherits from IdentifiedAnnotation and is categorized into one of five primary semantic classes:

  • DiseaseDisorderMention: Diagnoses and pathological conditions (mapped to SNOMED CT Disorder hierarchy).
  • SignSymptomMention: Clinical signs, symptoms, and qualitative findings (mapped to SNOMED CT Finding hierarchy).
  • ProcedureMention: Diagnostic, therapeutic, and surgical procedures.
  • MedicationMention: Active ingredients, brand names, dosages, routes, and frequencies (mapped to RxNorm and SNOMED CT).
  • AnatomicalSiteMention: Body structures and laterality modifiers (`BodyLateralityModifier`, `BodySideModifier`).

The 6 Assertion Modifiers on Every IdentifiedAnnotation

To prevent negated or family-history mentions from corrupting clinical research cohorts—a foundational requirement in clinical natural language processing (cNLP)—cTAKES populates six contextual attributes on every mention:

  1. polarity: 1 (affirmed/present, e.g., “Patient has pneumonia”) vs. -1 (negated/absent, e.g., “Denies chest pain or dyspnea”).
  2. uncertainty: 1 (hedged/probable/rule-out, e.g., “Cannot exclude pulmonary embolism”) vs. 0 (certain).
  3. conditional: true (hypothetical/instructional, e.g., “Return to ED if fever develops”) vs. false.
  4. generic: true (non-patient-specific mention, e.g., “Discussed stroke risk factors”) vs. false.
  5. subject: "patient" vs. "family_member" (e.g., “Mother had breast cancer at age 52”) or "donor".
  6. historyOf: 1 (historical condition resolved prior to the current encounter) vs. 0 (current/active).

What Standards or Regulatory Requirements Govern Apache cTAKES?

Deploying Apache cTAKES in healthcare and clinical research requires compliance with three technical and licensing standards:

  • NLM Unified Medical Language System (UMLS) Metathesaurus License: While Apache cTAKES software is open-source under the Apache License 2.0, its default clinical dictionary (`sno_rx`) contains SNOMED CT and RxNorm concepts governed by the National Library of Medicine (NLM) UMLS license. Users must authenticate with a free NLM UMLS Terminology Services (UTS) API key (`umlsKey`) at runtime or build custom dictionaries using the cTAKES Dictionary Creator GUI.
  • HIPAA De-Identification & On-Premises Execution (45 CFR § 164.514(b)): Because cTAKES runs entirely locally inside a Java Virtual Machine (JVM) without external cloud API calls, hospitals deploy it behind the institutional firewall on raw Protected Health Information (PHI) notes to extract de-identified CUI vectors before exporting research datasets.
  • OMOP Common Data Model (`NOTE_NLP` Table) and HL7 FHIR Mapping: The cTAKES output (UMLS CUI, character offsets `begin`/`end`, `section_concept_id`, `term_modifiers` for negation/temporality, and `nlp_system` provenance) maps directly to the OHDSI OMOP CDM v5.4 NOTE_NLP table as well as HL7 FHIR Condition and FHIR DocumentReference clinical indexing pipelines. When lighter-weight Python rule pipelines are preferred over Java UIMA, teams often pair or compare cTAKES with medspaCy.

How Does Apache cTAKES Integrate with EHR and Health Informatics Workflows?

In enterprise health systems, Apache cTAKES is typically deployed via three execution patterns: (1) batch execution using runPiperFile.sh / PiperFileRunner reading notes from an EHR Clarity/Caboodle database via JDBC and writing structured CUI rows back to a relational table; (2) distributed scale-out across millions of notes using Apache UIMA AS (Asynchronous Scaleout) or Apache Spark clusters; or (3) containerized REST microservice wrappers (`ctakes-web-rest`) that accept clinical note text and return JSON/XMI serialized CAS graphs in real time.

Frequently Asked Questions

What is apache ctakes and how is it used in clinical research and healthcare?

Apache cTAKES (clinical Text Analysis and Knowledge Extraction System) is an open-source Java clinical NLP platform originally developed at Mayo Clinic and stewarded by the Apache Software Foundation (Savova et al., 2010) that uses Apache UIMA and OpenNLP to extract UMLS SNOMED CT and RxNorm Concept Unique Identifiers (CUIs), negation, uncertainty, family history, and temporal relations from unstructured EHR clinical notes.

What standards or regulatory requirements govern apache ctakes?

Apache cTAKES is licensed under the Apache License 2.0, requires an NLM Unified Medical Language System (UMLS) Terminology Services API key to access SNOMED CT and RxNorm dictionaries, implements the OASIS/Apache UIMA standard architecture, and outputs structured concept annotations compatible with the OHDSI OMOP CDM NOTE_NLP table and HIPAA-compliant on-premises EHR data warehouses.

How does apache ctakes integrate with EHR and health informatics workflows?

Health systems integrate Apache cTAKES into EHR reporting databases using declarative .piper pipeline files (`DefaultFastPipeline.piper`), JDBC database readers/writers, or RESTful microservices (`ctakes-web-rest`) to transform unstructured pathology, radiology, and progress notes into queryable SNOMED CT and RxNorm tables for cohort discovery and phenotyping.

Sources

  • Savova GK, Masanz JJ, Ogren PV, Zheng J, Sohn S, Kipper-Schuler KC, Chute CG. Mayo clinical Text Analysis and Knowledge Extraction System (cTAKES): architecture, component evaluation and applications. Journal of the American Medical Informatics Association (JAMIA). 2010;17(5):507–513. https://pmc.ncbi.nlm.nih.gov/articles/PMC2995668/ (2026-11-09).
  • Apache Software Foundation. Apache cTAKES Official Documentation & Component Guide. https://ctakes.apache.org/ (2026-11-09).

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about Apache cTAKES: Clinical NLP Architecture, UIMA Pipeline, UMLS Fast Dictionary Lookup & Negation Assertion

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →