TextMeta

TextMeta allows for processing of PDF files, it extracts the text & metadata using rules (Regular Expressions) which is then returned in the format { "text": "extracted_text", meta: {}}.

Try It

You can try it in your browser here: http://textmeta.azurewebsites.net/ using the sample rules.json & sample pdf file.

Installation

npm install textmeta

Usage

var textmeta = require('textmeta');

var rules = [
  {
    key: "Author",
    type: "FirstSingle",
    expression: "Author:\\s+([^\\n]+)",
    default: "Unknown"
  }
];

textmeta.extractFromPDFFile("samples/sample_doc.pdf", rules).then((result) => {
  console.log("Text => " + result.text);
  console.log("Metadata => " + JSON.stringify(result.meta, null, 4));
});

For an example of the rules see the rules.json file from a sample Azure Function that uses this module.

API

extractFromPDFFile(file, rules, options)

• Extracts text from the source PDF file and metadata using the supplied rules.
• The options parameter is optional, for more details see below.
• Returns { text: "...", meta: {} }.

extractFromPDFBuffer(buffer, rules, options)

• Extracts text from the supplied buffer and metadata using the supplied rules.
• The options parameter is optional, for more details see below.
• Returns { text: "...", meta: {} }.

processText(file, rules, options)

• Returns metadata using the supplied text.
• The options parameter is optional, for more details see below.

Rules

The format for a rule is

{
  "key": "<Metadata Name>",
  "type": "<Match Type>",
  "expression": "<Regular Expression>",
  "default": "<Default Value if no matches>"
  "startKeyword": "Optional: <Keyword for substring match start>",
  "endKeyword": "Optional: <Keyword for substring match end>",
  "options": {
    "flags": "<Optional RegularExpression Flags>"
  }
}

Rule Types:

• "FirstSingle" : Will capture the first match.
• "All" : Will capture all matches.
• "AllUnique" : Will capture all matches and return the list of unique strings.

Start/End Keywords

If you want to run your expression on a subset of the text, then specify the start/end keywords and only the text in between will be used.

Options

The default options are

{
    "numericPrecision": 0,
    "nonWhitespaceRegexp": "/\\S/",
    "replaceMultipleSpaces": true,
    "orderByYPos": true,
    "distanceToAddSpace": 1
}

• numericPrecision : specifies the decimal precision when doing position (X, Y) comparisons.
• nonWhitespaceRegexp: regular expression used (represented as a string) to determine if the text contains non whitespace characters.
• replaceMultipleSpaces: boolean indicating if multiple spaces should be replaced with just a single space.
• orderByYPos: boolean indicating if the Y position of the text should be used to determine order. Some PDF files can be generated with text out of sequence (since the actual position of the text is determined by it's the transformation matrix which contains the X & Y positions). When set to true, TextMeta will first extract all lines on a page keeping track of their Y position then sort them.
• distanceToAddSpace: Sometimes text in the PDF can be chucked (that is to say instead of full sentances sometimes words are broken up), this value indicates the minimum distance required between two text elements before a space ' ' is applied between them.

Sample PDF

The result of processing the sample file sample_doc.pdf using the sample rules.json is the following result:

Text

  Title: PDF to Text Function
  Author: Marc Gagne 
  
  Description: 
  An azure function that extracts text from PDFs, runs the regular expression captures found in rules.json 
  against the text and stores the results in DocumentDB. 
  
  Technologies used: 
   Azure Functions 
   pdf.js 
   JavaScript 
   Node.js 
  
  GitHub: https://github.com/m-gagne/PDF2AzSearch 

Meta Data

    {
      "id": "dafe8948ef379e6aef78cc1b059122cebcae436d7dd878375f16094a99a9243b",
      "name": "sample_doc.pdf",
      "text": "Title: PDF to Text Function \nAuthor: Marc Gagne \n \nDescription:  \nAn azure function that extracts text from PDFs, runs the regular expression captures found in rules.json \nagainst the text and stores the results in DocumentDB. \n \nTechnologies used: \n• Azure Functions \n• pdf.js \n• JavaScript \n• Node.js \n \nGitHub: https://github.com/m-gagne/PDF2AzSearch \n ",
      "last_updated": "2017-05-23T20:10:31.653Z",
      "meta": {
        "Title": "PDF to Text Function",
        "Author": "Marc Gagne",
        "Description": "An azure function that extracts text from PDFs, runs the regular expression captures found in rules.json",
        "Technologies": [
          "Azure Functions",
          "pdf.js",
          "JavaScript",
          "Node.js"
        ]
      }
    }

Why

Originally I created an Azure Function called PDF2Search that extracted text from a PDF file when stored in an Azure Storage blob container and realized it should exist outside of that project to be used elsewhere.

Primarily it's helpful for people/organizations that have PDF documents of a specific format that makes running regular expressions against them to extract metadata possible. The idea being that the extracted text & metadata can be indexed using Azure Search to more easily find the relevant document.

It quickly become apparent that this should exist as a standalone package to be used in various environments, so I refactored this logic out of the Azure Function and into it's own reusable Node.js package.