31:[] 2f:{"title":"Compromise-offsets","cells":[{"id":0,"value":"# Compromise-offsets\n[compromise](https://observablehq.com/@spencermountain/nlp-compromise) offers the ability to parse and match arbitrary points in a text. Often you want to apply this analysis back to the original source text, by highlighting or replacing parts of the text.\n\ncharacter-offsets are notoriously tricky to deal with. They get hung-up on unicode, and whitespace/punctuation.\nWhen the document changes, the offsets must be re-caculated...\n\n---\n\n### Getting offsets\nFrequently though, character-offsets are an unavoidable thing. They can be included as part of a **[.json()](https://observablehq.com/@spencermountain/compromise-json)** result, like this:","pinned":false,"hidden":false,"mode":"md"},{"id":5,"value":"{\n let doc = nlp('Who holds back the electric car? Who made Steve Guttenberg a star?')\n let nouns = doc.match('#Person+')\n return nouns.json({offset:true})[0].offset\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":7,"value":"**`index`** is the term #, from the start of the document. Note that this number includes [implicit terms](https://observablehq.com/@spencermountain/compromise-tokenization).\n\n**`start`** is the first character of the term, not including its whitespace or leading punctuation.\n\n**`length`** is the last character of the word, not including it's punctuation or whitespace.\n\n---\n\n### Naive highlighting\nYou can plunder your original text, using the **start** and **length** properties, like this: \n","pinned":false,"hidden":false,"mode":"md"},{"id":9,"value":"{\n let txt=' 🎶 Who made Steve Guttenberg a star? 🎵'\n let doc = nlp(txt)\n let json = doc.match('#Person+').json({offset:true})[0]\n let offset = json.offset\n //replace the string at offset\n let before = txt.substring(0, offset.start)\n let replacement = `**${json.text}**` //add highlight\n let after = txt.substring(offset.start + offset.length)\n return `${before}${replacement}${after}`\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":29,"value":"remember though, that making one change to the text will throw off all subsequent offsets. Working with offsets like this is not recommended. It will address the offset-slip issues, described above:\n\n---\n\n### .wrap()\ncompromise offers a function that lets you add custom markup before and after any arbitrary match.\n","pinned":false,"hidden":false,"mode":"md"},{"id":67,"value":"{\n let doc = nlp(\"soft and yielding like a nerf ball\")\n return doc.wrap({\n '#Adjective': (m) => `[${m.text()}]`\n })\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":73,"value":"regarding ***overlapping markup***, remember that in most markup languages, this sort of thing is invalid:\n\n```bold and italic```\n\nyou can see .wrap() will avoid creating any overlapping markup:","pinned":false,"hidden":false,"mode":"md"},{"id":75,"value":"{\n let doc = nlp(\"bold and italic\")\n return doc.wrap({\n 'bold and': (m) => `${m.text()}`,\n 'and italic': (m) => `${m.toUpperCase().text()}` //ignored\n })\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":82,"value":"{\n let doc = nlp(\"it was bold and uppercase\")\n return doc.wrap({\n 'bold': (m) => `${m.text()}`,\n 'bold and uppercase': (m) => `${m.toUpperCase().text()}` \n })\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":159,"value":"note that **.wrap()** does not change the document in any way.","pinned":false,"hidden":false,"mode":"md"},{"id":52,"value":"---\n### Generating HTML:\nIn addition to the issues with plaintext markup, creating HTML markup involves additional concerns about xml encoding. Compromise comes with a facility for handling this, and generating html spans from text matches:","pinned":false,"hidden":false,"mode":"md"},{"id":54,"value":"{\n let doc = nlp(`The aurora borealis? At this time of year?`).debug()\n let str= doc.html({\n '.red': 'aurora #Noun', //a class name : match\n })\n return html`${str}`\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":146,"value":"---\n\n### demo:\ncreate a link to wikipedia, for every person mentioned in a text:","pinned":false,"hidden":false,"mode":"md"},{"id":98,"value":"viewof sentence = html``","pinned":false,"hidden":false,"mode":"ojs"},{"id":94,"value":"{\n let doc = nlp(sentence)\n let out = doc.out({\n '#Person+': (m) => {\n let page = encodeURIComponent(m.text().replace(/ /g, '_')) //some wikipedia encoding\n return `${m.text()}`\n }\n })\n return html`${out}`\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":153,"value":"### Notes: \n* ***markup-to-markup***\ncompromise assumes all input is plaintext. If you are trying to augment or change text in a markup language, there is no easy way to do this.\n\n* ***highlighting editable text***\nfor creating WYSIWYG text interactions on the web, we recommend using [codemirror](https://codemirror.net/).","pinned":false,"hidden":false,"mode":"md"},{"id":11,"value":"

\n","pinned":false,"hidden":false,"mode":"html"},{"id":13,"value":"nlp=require('compromise')","pinned":false,"hidden":false,"mode":"ojs"},{"id":15,"value":"nlp.version","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W31","stdlib":"1","titling":"auto"}