A JavaScript/TypeScript utility that bridges the gap between TEI/XML documents and plaintext processing tools.
This library creates a reversible mapping between TEI/XML markup and character offsets in plaintext, allowing you to apply text analysis tools to TEI documents without losing markup context.
TEI/XML: <p>This is a <hi>sample</hi> text.</p>
Plaintext: This is a sample text.
^----^
Identified entity
Result: <p>This is a <hi><placeName>sample</placeName></hi> text.</p>
The core logic was ported to TypeScript from the excellent Python standoffconverter by @millawell.
npm install @recogito/standoff-converterText analysis tools (e.g. for Named Entity Recognition) typically work with plaintext only. However, TEI/XML documents contain rich structural markup that must be stripped away before processing. When analysis tools identify entities or features at specific character positions in the plaintext, it's hard to map those positions back to the original TEI markup structure.
This library:
- Creates a linearized representation that maintains the relationship between plaintext character positions and XML markup.
- Allows you to process the plaintext with any text analysis tool.
- Maps the identified features back to the exact location in the original XML.
- Allows you to modify the TEI/XML structure while preserving all existing markup.
Perfect for enriching TEI documents with automatically extracted entities, annotations, or other textual features!
- Extract plaintext from TEI/XML while preserving a bidirectional mapping between character offsets and markup.
- Convert between plaintext character offsets and TEI XPointer expressions.
- Insert new inline tags at specific character positions (e.g. add
<placeName>or<persName>tags based on NER results). - Preserve all original markup when serializing changes back to TEI/XML.
- Works in both Node.js and browser environments.
This library works in Node (using xmldom and xpath internally).
import { parseXML } from '@recogito/standoff-converter';
const xml = `
<TEI xmlns="http://www.tei-c.org/ns/1.0">
<teiHeader>
<fileDesc>
<titleStmt>
<title>Sample TEI Document</title>
</titleStmt>
</fileDesc>
</teiHeader>
<text>
<body>
<p>This is a <hi rend="italic">sample</hi> paragraph with <term>markup</term>.</p>
</body>
</text>
</TEI>
`;
const parsed = parseXML(xml);
// Get plaintext
const text = parsed.text();
// XPointer expression from character position
const xpointer = parsed.getXPointer(550);
// Character position from XPointer expression
const position = parsed.getCharacterOffset('//TEI/text[1]/body[1]/p[1]/::5');
// Add inline tag at character position
parsed.addInline(5, 7, 'rs', { resp: 'aboutgeo' });
// Modified markup as a DOM Element
const el = parsed.xml();
// Modified markup serialized to string
const xml = parsed.xmlString();You can use this library in the browser in combination with CETEIcean.
import { parseXML } from '@recogito/standoff-converter';
window.onload = async function () {
const CETEIcean = new CETEI();
CETEIcean.getHTML5('paradise-lost.xml', data => {
document.getElementById('orig').appendChild(data);
const el = document.getElementById('orig').firstChild;
// Parse CETEIcean content
const parsed = parseXML(el);
// Get XPointer expressions from plaintext character offsets
console.log(parsed.getXPointer(550));
// Get character offsets from an XPointer expression (format: path::offset)
const xpointer = '//text[@xml:id="text-1"]/body[1]/div[1]/p[4]/hi[1]::5';
console.log(parsed.getCharacterOffset(xpointer));
// Add inline tags at character positions
parsed.addInline(550, 560, 'tei-note', { type: 'comment', resp: 'aboutgeo' });
// Serialize back to TEI/XML
const teiElement = parsed.toXML();
document.getElementById('serialized').appendChild(teiElement);
});
};| Function | Description | Parameters | Return Value |
|---|---|---|---|
parseXML(input) |
Parse TEI/XML | input: XML string or Element |
parsed instance |
parsed.text() |
Get plaintext | None | string |
parsed.tokens |
Access linearized token array | - | Array of token objects |
parsed.getXPointer(offset) |
Convert plaintext character offset to XPointer | offset: number |
string XPointer expression |
parsed.getCharacterOffset(xpointer) |
Convert XPointer to character offset | xpointer: string |
number |
parsed.addInline(start, end, tagName, attrs) |
Insert inline tag at character positions | start: numberend: numbertagName: stringattrs: object |
void |
parsed.xml() |
Get TEI/XML (DOM Element) | None | Element |
parsed.xmlString() |
Get XML (serialized string) | None | string |
| Function/Method | Description | Parameters | Return Value |
|---|---|---|---|
parsed.annotations(standOffId?) |
Get standoff annotations from all or a specific TEI <standOff> element |
standOffId?: string |
Array of standoff annotation objects |
parsed.addStandOff(id) |
Add a new TEI <standOff> element |
id: string |
string annotation ID |
parsed.addAnnotation(standOffId, annotation) |
Add Recogito annotation to standOff element |
standOffId: stringannotation: standoff annotation |
void |
parsed.addStandOffTag(standOffId, start, end, tag) |
Add Recogito annotation to standOff element that represents a simple (NER) tag |
standOffId: stringstart: numberend: numbertag: string or { id, label } |
void |
- standOff Anchors and inline markup modifications. Using
.addInlinewill change the TEI markup. This has the potential to break XPath anchors for annotations in<standOff>blocks. In order to prevent this, we would need to do before/after checks for affected anchors, and update them accordingly, so that they remain in sync with the changed TEI document.