Temporal Overview
Annotator for Clinical Data can detect temporal expressions and link them to related concepts as contextual features. This temporal capability is currently available through Clinical Insights.
The Annotator for Clinical Data demo application allows you to see how temporal linking works. In the screenshot below, hovering over hip surgery shows that is linked to the date 5/17.
Concepts are linked to dates by JSON structures described below.
temporal
| Feature | Description |
|---|---|
| begin | The start position of the temporal entity as a character offset into the text. The smallest possible start position is 0. |
| end | The end position of the temporal entity as a character offset into the text. The end position points at the first character after the annotation, such that end-begin equals the length of the coveredText. |
| temportalType | Provides information about the type of a temporal entity. Currently, dateScore is the only score provided in this section. Anything that scores sufficiently low as a date will not be surfaced. For example, in the image above 2/10 in the context of the patient’s current pain level scores low as a date and is not surfaced. |
| relationType | Provides information about how the temporal entity relates to an entity. Currently, overlapsScore is the only relation type defined and it indicates how strongly the temporal entity is linked to the concept. |
Temporal features
"begin": 585,"end": 596,"coveredText": "hip surgery","negated": false,"insightModelData": {"procedure": {"usage": {"explicitScore": 0.999,"pendingScore": 0,
It is possible for multiple temporal links to be added to a concept. For example, given the following text:
PMH: Patient had bariatric revision surgery in April 1999 and again in July 2003, knee replacement in 2011, and underwent arthroscopic surgery on her shoulder in 2016.
The temporal section for revision surgery would have two associated dates.
"temporal": [{"begin": 47,"end": 57,"coveredText": "April 1999","temporalType": {"dateScore": 1},"relationTypes": {"overlapsScore": 0.988