Skip to content

Latest commit

 

History

History
450 lines (329 loc) · 12.7 KB

README.md

File metadata and controls

450 lines (329 loc) · 12.7 KB

Citation

Build Status

A fast, stand-alone legal citation extractor.

Currently supports:

  • usc: US Code
  • law: US Slip Laws (public and private laws)
  • stat: US Statutes at Large
  • cfr: US Code of Federal Regulations
  • dc_code: DC Code
  • dc_register: DC Register
  • dc_law: DC Slip Law

With limited, opt-in support for:

As you can see, Citation is currently US-only, but we'd love for that to change. There are lots more citation types out there, and it's easy to contribute, so please help us grow!

Compatible in-browser with modern browsers, including IE 9+.

Use

Citation can be used:

  1. In JavaScript, in browser or in Node. This method supports the most options, including passing in JavaScript functions as callbacks.
  2. Over HTTP, via GET or POST. Supports JSON and JSONP. Options require function callbacks are not supported (it won't eval JavaScript).
  3. On the command line or Unix pipes, over STDOUT. Options require function callbacks are not supported (it won't eval JavaScript).

But one way or another, you pass in text:

Citation.find("pursuant to 5 U.S.C. 552(a)(1)(E) and");

And you get back data about matched citations:

[{
  "match": "5 U.S.C. 552(a)(1)(E)",
  "citation": "5 U.S.C. 552(a)(1)(E)",
  "type": "usc",
  "index": "0",
  "usc": {
    "title": "5",
    "section": "552",
    "subsections": ["a", "1", "E"],
    "id": "usc/5/552/a/1/E",
    "section_id": "usc/5/552"
  }
}]

In the browser

Check out /browser for browser-ready compressed and uncompressed versions of the library.

Loading any of them with a <script> tag will result in a global Citation object being available for immediate use.

In Node

Install Node.js and NPM, then install Citation globally (may require sudo):

npm install -g citation

Or install it locally to a node_modules directory with npm install citation.

JavaScript API

Citation.find(text, options)

Check a block of text for citations of a given type, returning an array of matches with citations broken out into fields.

options can include:

  • types: (string | string array) Limit citation types to those given. e.g. ["usc", "law"]
  • excerpt: (integer) Return an excerpt of the surrounding text for each detected cite, with the given number of characters on either side.
  • parents: (boolean) For any cite, return any "parent" cites alongside it. For example, matching "5 USC 552(b)(3)" would return 3 results - one for the parent section, one for (b), and one for (b)(3).
  • filter: (string) Enable Filtering.
  • replace: (function | object) Enable Replacement.
  • links: (boolean) Include Links.
  • Also: see Cite-specific options to pass in options for a particular citation type.

Some examples:

Citation.find("pursuant to 5 U.S.C. 552(a)(1)(E) and");

// Yields:

[{
  "match": "5 U.S.C. 552(a)(1)(E)",
  "citation": "5 U.S.C. 552(a)(1)(E)",
  "type": "usc",
  "index": "0",
  "usc": {
    "title": "5",
    "section": "552",
    "subsections": ["a", "1", "E"],
    "id": "usc/5/552/a/1/E",
    "section_id": "usc/5/552"
  }
}]

Citation.find("that term in section 5362(5) of title 31, United States Code.", {
  excerpt: 10
})

// Yields:

[{
  "match": "section 5362(5) of title 31",
  "citation": "31 U.S.C. 5362(5)",
  "excerpt": "t term in section 5362(5) of title 31, United S",
  // ... more details ...
}]

HTTP API

Start the API on a given port (defaults to 3000):

cite-server [port]

GET or POST to /citation/find with a text parameter:

curl http://localhost:3000/citation/find?text=5+U.S.C.+552%28a%29%281%29%28E%29

curl -XPOST "http://localhost:3000/citation/find" -d "text=5 U.S.C. 552(a)(1)(E)"

Will return the results of running Citation.find() on the block of text, under a results key:

{
  "results": [
    {
      "match": "5 U.S.C. 552(a)(1)(E)",
      "citation": "5 U.S.C. 552(a)(1)(E)",
      "type": "usc",
      "index": "0",
      "usc": {
        "title": "5",
        "section": "552",
        "subsections": ["a", "1", "E"],
        "id": "usc/5/552/a/1/E",
        "section_id": "usc/5/552"
      }
    }
  ]
}

Supported options

Some HTTP-specific parameters:

  • callback: a function name to use as a JSONP callback.
  • pretty: prettify (indent) output.

And some of the options that the JavaScript API supports:

  • text: required, text to extract citations from.
  • options[excerpt]: include excerpts with up to this many characters around it.
  • options[types]: limit citation types to a comma-separated list (e.g. "usc,law")

Server deployment

See etc/ for an example upstart script to keep cite-server running in production.

Command line API

The shell command can accept a string to parse as an argument or through STDIN, and outputs results to STDOUT as indented JSON.

cite "section 5362(5) of title 31"

echo "section 5362(5) of title 31" | cite

cite "pursuant to 5 U.S.C. 552(a)(1)(E) and > results.json"

Options

Pass any options the library takes, using dot operators to pass nested options.

For example, searching among types:

cite --types=usc,law "section 5362(5) of title 31"

Passing nested options:

cite --dc_code.source=dc_code "and then § 3-101.01 happened"

Opt-in to using walverine to search judicial cites with --judicial:

cite --judicial "Smith v. Hardibble, 111 Cal.2d 222, 555, 558, 333 Cal.3d 444 (1988)"

Add --links to include links in the output.

Filters

Instead of treating the input text as just a blob of text that matches citations at a string index, you can apply a "filter" that will parse the input text and provide more precise context.

Lines

For each citation, return the line number and the relative character index of the match inside that line.

Example:

cite --pretty --filter=lines "I once met a cite named nancy
whose 5 usc 552 was awfully fancy
and then the poem ended"
{
  "citations": [
    {
      "type": "usc",
      "match": "5 usc 552",
      "index": 6,
      "citation": "5 U.S.C. 552",
      "usc": {
        "title": "5",
        "section": "552",
        "subsections": [],
        "id": "usc/5/552"
      },
      "line": 2
    }
  ]
}
XPath

For each citation, return an XPath statement identifying the match's specific node in the input document, and the relative character index of the match inside that node.

Example:

cite --pretty --filter=xpath_xml "
<?xml>
<document>
  <title>Best Bill of 2012</title>
  <bill>
    <introduction>Bill to enforce happiness amongst all the children</introduction>
    <closing>All information releasable through 5 U.S.C. 552 is now banned</closing>
    <footer>(c) Congress</footer>
  </bill>
</document>
"
{
  "citations": [
    {
      "type": "usc",
      "match": "5 U.S.C. 552",
      "index": 35,
      "citation": "5 U.S.C. 552",
      "usc": {
        "title": "5",
        "section": "552",
        "subsections": [],
        "id": "usc/5/552"
      },
      "xpath": "/document[1]/bill[1]/closing[1]/text()[1]"
    }
  ]
}

Replacement

You can perform a "find-and-replace" with detected citations, by providing a replace callback to be executed on each citation, that returns the string to replace that citation.

By passing a replace callback, a text field will be included at the top of the returned object, with the processed text.

Citation.find("click on 5 USC 552 to read more", {
  replace: function(cite) {
    var url = "http://www.law.cornell.edu/uscode/text/" + cite.usc.title + "/" + cite.usc.section;
    return "<a href=\"" + url + "\">" + cite.match + "</a>";
  };
});

The response will have a text field containing:

click on <a href="http://www.law.cornell.edu/uscode/text/5/552">5 USC 552</a> to read more

This feature is only available in the JavaScript API.

Include Links

With the links option, each matched citation will include URLs to access the content of the citation on the web. For:

Citation.find("pursuant to 5 U.S.C. 552(a)(1)(E) and", { links: true });

you will get back an extended object with permalinks:

[{
  "match": "5 U.S.C. 552(a)(1)(E)",
  "type": "usc",
  ...
  "usc": {
    "id": "usc/5/552/a/1/E",
    ...
    "links": {
      "usgpo": {
        "source": {
          "name": "U.S. Government Publishing Office",
          "abbreviation": "US GPO",
          "link": "http://www.gpo.gov",
          "authoritative": true,
          "note": "2014 edition. Sub-section citation is not reflected in the link."
        },
        "pdf": "http://api.fdsys.gov/link?collection=uscode&year=2014&title=5&section=552&type=usc",
        "html": "http://api.fdsys.gov/link?collection=uscode&year=2014&title=5&section=552&type=usc&link-type=html",
        "landing": "http://api.fdsys.gov/link?collection=uscode&year=2014&title=5&section=552&type=usc&link-type=contentdetail"
      },
      "cornell_lii": {
        "source": {
          "name": "Cornell Legal Information Institute",
          "abbreviation": "Cornell LII",
          "link": "https://www.law.cornell.edu/uscode/text",
          "authoritative": false,
          "note": "Link is to most current version of the US Code, as available at law.cornell.edu."
        },
        "landing": "https://www.law.cornell.edu/uscode/text/5/552#a_1_E"
      }
    }
  }
}]

The links object maps sources to one or more renditions. The rendition types are pdf, html (for raw HTML content), landing for a landing page (i.e. a website) about the document refered to by the citation, and mods (US GPO MODS XML files).

Cite-specific options

You can pass arbitrary options to individual citators, if that citator supports them.

By using a key is the key of a citator, e.g. usc or dc_code, that citator's processors will get the value of that key passed in as an argument.

Example: DC Code relative cites

For example, the dc_code citator accepts a source option, to indicate what the text source is. If the value of source is itself "dc_code", then the citator will apply a looser pattern to detect internal cites.

That looks like this:

Citation.find("required under § 3-101.01(13)(e), the Commission shall perform the", {
  dc_code: {source: "dc_code"}
})

That will match § 3-101.01(13)(e), because the dc_code citator assumes it's processing the text of the DC Code itself, and internal references are unambiguous.

Court opinions

Citation can integrate with walverine to detect and return results for US court opinions.

To use walverine, you may need to "opt-in" to including judicial-type citations.

In JavaScript:

Citation.types.judicial = require("./citations/judicial");

In CLI:

cite --judicial "Text to scan"

The HTTP server, cite-server actually loads judicial cites by default, since the performance penalty is absorbed on start-up.

Unsupported features

walverine's support for extra features is limited. When detecting judicial-type cites, there is no support for:

  • Returning parent citations
  • Replacing detected text
  • A character index of detected citations

Tests

This project is tested with nodeunit.

To run tests, you'll need to install this project from source and install its node dependencies:

git clone [email protected]:unitedstates/citation.git
cd citation
npm install
npm test

Test cases are stored in the test directory. Each test case covers a subsection of the code and ensures that citations are correctly detected: for instance, see test/stat.js.

To run all tests:

nodeunit test

To run a specific test:

nodeunit test/usc.js

Public domain

This project is dedicated to the public domain. As spelled out in CONTRIBUTING:

The project is in the public domain within the United States, and copyright and related rights in the work worldwide are waived through the CC0 1.0 Universal public domain dedication.

All contributions to this project will be released under the CC0 dedication. By submitting a pull request, you are agreeing to comply with this waiver of copyright interest.