Powered by Zoomin Software. For more details please contactZoomin

The Semaphore Fact Extraction Framework (FACTS)

Table of Contents

NO TITLE FOUND

Save PDF
Save selected topicSave selected topic and subtopicsSave all topics
Share
Share to emailCopy topic URL
Print
  • Last Updated: October 30, 2026
  • 1 minute read
    • Semaphore
    • Documentation
| | [Contexts](fact_extraction_framework_document_contexts.md) | [Fact Context Elements](fact_extraction_framework_document_contexts.md) | not fact in |

Example model and tests

The example extractor DF4 in the “FactExtraction-Example” model uses a “not fact”. Our example is a “Near Ordered Across Phrases in Document” extractor that is looking for a Taxonomy Fact, a negative Taxonomy Fact, and then another Taxonomy Fact.

Important: for this extractor to work, you need to ensure the broader relationship of the concept “DF4 - Near ordered phrases in the same sentence (negated fact)” is “document-identifying document fact in”:

Positive test content

Apple beer.

“Apple” is the first fact in the extractor, a Taxonomy Fact looking for concepts in the “Everyday food” taxonomy.

“beer” is the third fact in the extractor, a Taxonomy Fact looking for concepts in the “Everyday drinks” taxonomy.

The second fact in the extractor is the “negative fact”, a Taxonomy Fact looking for concepts in the “Poison” taxonomy. If found, it will cause the extractor to fail. It is not found here, so the extractor will succeed.

Negative test content

Apple bleach beer.

This will not fire since the negative fact, “bleach” from the “Poison” taxonomy, is present.

Alert