@inproceedings{pusapati-singh-2026-gladiator,
title = "Gladiator at {MEDIQA}-{SYNUR} 2026: Contextual Clinical Extraction: Integrating Foundation Models with Domain-Specific Validation Rules",
author = "Pusapati, Siva Satyanarayana Raju and
Singh, Ankit",
editor = "Ben Abacha, Asma and
Bethard, Steven and
Bitterman, Danielle and
Naumann, Tristan and
Roberts, Kirk",
booktitle = "Proceedings of the 8th Workshop on Clinical Natural Language Processing (Clinical {NLP}) @ {LREC} 2026",
month = may,
year = "2026",
address = "Palma, Mallorca (Spain)",
publisher = "ELRA Language Resources Association (ELRA)",
url = "https://aclanthology.org/2026.clinicalnlp-1.20/",
doi = "10.63317/2tkvpwgzzcn2",
pages = "183--191",
abstract = "We present a hybrid extraction system that combines large language model capabilities with rule-based precision for extracting structured clinical observations from nursing dictation transcripts. Our approach leverages Claude Opus 4.5 as the primary extractor, enhanced with comprehensive prompt engineering that includes the complete 193-concept schema, few-shot examples, and detailed validation rules covering respiratory, cardiac, diagnosis, and mental status fields. The LLM output undergoes extensive post-processing with six specialized filters that remove speculative diagnoses, validate physiological ranges, ensure unit-field dependencies, and verify contextual appropriateness. Five correction mechanisms normalize breathing patterns, map dyspnea severity, standardize assistance levels, clean STRING fields, and handle multi-select conjunctions. A supplementary rule-based component employs 400+ regex patterns with contextual validation to capture high-confidence observations, particularly for vital signs and categorical fields. The system requires cardiac keywords for heart rate extraction and respiratory context for respiration rates, preventing false positives from unrelated numeric values. Results are merged through an intelligent strategy that prioritizes LLM comprehensiveness while supplementing with rule-based findings. A strict schema validation layer ensures all four value types (NUMERIC, STRING, SINGLE{\_}SELECT, MULTI{\_}SELECT) conform to enumerated options and physiological ranges. This multi-layered approach balances recall through LLM reasoning with precision through rule-based validation, effectively structuring natural nursing narratives into standardized EHR-ready observations."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="pusapati-singh-2026-gladiator">
<titleInfo>
<title>Gladiator at MEDIQA-SYNUR 2026: Contextual Clinical Extraction: Integrating Foundation Models with Domain-Specific Validation Rules</title>
</titleInfo>
<name type="personal">
<namePart type="given">Siva</namePart>
<namePart type="given">Satyanarayana</namePart>
<namePart type="given">Raju</namePart>
<namePart type="family">Pusapati</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ankit</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 8th Workshop on Clinical Natural Language Processing (Clinical NLP) @ LREC 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Asma</namePart>
<namePart type="family">Ben Abacha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Steven</namePart>
<namePart type="family">Bethard</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Danielle</namePart>
<namePart type="family">Bitterman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tristan</namePart>
<namePart type="family">Naumann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kirk</namePart>
<namePart type="family">Roberts</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resources Association (ELRA)</publisher>
<place>
<placeTerm type="text">Palma, Mallorca (Spain)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>We present a hybrid extraction system that combines large language model capabilities with rule-based precision for extracting structured clinical observations from nursing dictation transcripts. Our approach leverages Claude Opus 4.5 as the primary extractor, enhanced with comprehensive prompt engineering that includes the complete 193-concept schema, few-shot examples, and detailed validation rules covering respiratory, cardiac, diagnosis, and mental status fields. The LLM output undergoes extensive post-processing with six specialized filters that remove speculative diagnoses, validate physiological ranges, ensure unit-field dependencies, and verify contextual appropriateness. Five correction mechanisms normalize breathing patterns, map dyspnea severity, standardize assistance levels, clean STRING fields, and handle multi-select conjunctions. A supplementary rule-based component employs 400+ regex patterns with contextual validation to capture high-confidence observations, particularly for vital signs and categorical fields. The system requires cardiac keywords for heart rate extraction and respiratory context for respiration rates, preventing false positives from unrelated numeric values. Results are merged through an intelligent strategy that prioritizes LLM comprehensiveness while supplementing with rule-based findings. A strict schema validation layer ensures all four value types (NUMERIC, STRING, SINGLE_SELECT, MULTI_SELECT) conform to enumerated options and physiological ranges. This multi-layered approach balances recall through LLM reasoning with precision through rule-based validation, effectively structuring natural nursing narratives into standardized EHR-ready observations.</abstract>
<identifier type="citekey">pusapati-singh-2026-gladiator</identifier>
<identifier type="doi">10.63317/2tkvpwgzzcn2</identifier>
<location>
<url>https://aclanthology.org/2026.clinicalnlp-1.20/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>183</start>
<end>191</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Gladiator at MEDIQA-SYNUR 2026: Contextual Clinical Extraction: Integrating Foundation Models with Domain-Specific Validation Rules
%A Pusapati, Siva Satyanarayana Raju
%A Singh, Ankit
%Y Ben Abacha, Asma
%Y Bethard, Steven
%Y Bitterman, Danielle
%Y Naumann, Tristan
%Y Roberts, Kirk
%S Proceedings of the 8th Workshop on Clinical Natural Language Processing (Clinical NLP) @ LREC 2026
%D 2026
%8 May
%I ELRA Language Resources Association (ELRA)
%C Palma, Mallorca (Spain)
%F pusapati-singh-2026-gladiator
%X We present a hybrid extraction system that combines large language model capabilities with rule-based precision for extracting structured clinical observations from nursing dictation transcripts. Our approach leverages Claude Opus 4.5 as the primary extractor, enhanced with comprehensive prompt engineering that includes the complete 193-concept schema, few-shot examples, and detailed validation rules covering respiratory, cardiac, diagnosis, and mental status fields. The LLM output undergoes extensive post-processing with six specialized filters that remove speculative diagnoses, validate physiological ranges, ensure unit-field dependencies, and verify contextual appropriateness. Five correction mechanisms normalize breathing patterns, map dyspnea severity, standardize assistance levels, clean STRING fields, and handle multi-select conjunctions. A supplementary rule-based component employs 400+ regex patterns with contextual validation to capture high-confidence observations, particularly for vital signs and categorical fields. The system requires cardiac keywords for heart rate extraction and respiratory context for respiration rates, preventing false positives from unrelated numeric values. Results are merged through an intelligent strategy that prioritizes LLM comprehensiveness while supplementing with rule-based findings. A strict schema validation layer ensures all four value types (NUMERIC, STRING, SINGLE_SELECT, MULTI_SELECT) conform to enumerated options and physiological ranges. This multi-layered approach balances recall through LLM reasoning with precision through rule-based validation, effectively structuring natural nursing narratives into standardized EHR-ready observations.
%R 10.63317/2tkvpwgzzcn2
%U https://aclanthology.org/2026.clinicalnlp-1.20/
%U https://doi.org/10.63317/2tkvpwgzzcn2
%P 183-191
Markdown (Informal)
[Gladiator at MEDIQA-SYNUR 2026: Contextual Clinical Extraction: Integrating Foundation Models with Domain-Specific Validation Rules](https://aclanthology.org/2026.clinicalnlp-1.20/) (Pusapati & Singh, ClinicalNLP 2026)
ACL