Tag Archives: documentcloud

The Iraq Inquiry documents on DocumentCloud

I’ve uploaded all 2.6million words of the Chilcot Report (also known as the Iraq Inquiry) to the document analysis service DocumentCloud. The documents can be found on the site here.

You’ll need a DocumentCloud account to analyse the documents by entity or timeline (more on that here), but if you don’t have an account you can also search all the documents using the embedded view below.



Savile extracted

On Friday the BBC released documents from The Pollard Report into the Savile inquiry.

These were published as scanned PDFs, making it impossible to search text or count mentions of particular terms.

We’ve used document extraction service DocumentCloud to convert the two key documents – appendices 10 (statements) and 12 (emails and documents) – into text. These are linked below. If you use them, let us know so we can continue to do this.

Savile Transcript appendix 10 (PDF)
Savile Transcript appendix 10 (Text)

 

Savile Appendix12 (PDF)
Savile Appendix12 (Text)

 

7 ways to get data out of PDFs

A frequent obstacle in data journalism is when the information you want to analyse is locked away in a PDF. Here are 6 ways to tackle that problem – with space for a 7th:

1) For simple PDFs: Google Docs’ conversion facility

Google Docs recently added a feature that allows you to convert a PDF to a ‘Google document’ when you upload it. It’s pretty powerful, and about the simplest way you can extract information.

It does not work, however, if the PDF was generated by scanning – in other words if it is an image, rather than a document that has been converted to PDF.

2) For scanned documents and pulling out key players: Document Cloud

Document Cloud is a tool for journalists to convert PDFs to text. It will also add ‘semantic’ information along the way, such as what organisations, people and ‘entities’ such as dates and locations are mentioned within it, and there are some useful features that allow you to present documents for others to comment on. 

The good news is that it works very well with scanned documents, using Optical Character Recognition (OCR). The bad news is that you need to ask permission to use it, so if you don’t work as a professional journalist you may not be able to use it. Still, there’s no harm in asking.

3) For scanned documents: The Data Science Toolkit

The Data Science Toolkit allows you to do lots of clever things, including converting PDFs using OCR with theFile2Text converter. Upload your document, and you’re away. Also works on other document formats, and PNGs, TIFFs and JPEGs.

4) For stripping out tables: PDF2XL

If you’re willing to shell out around £70 then PDF2XL is recommended as a useful piece of software for stripping out tables from Excel files. 

5) For automating the process: Scrape from PDF to XML using Scraperwiki

Scraperwiki is a collaborative website for scraping all sorts of hard-to-find information into some sort of useful format, so it’s no surprise that PDFs are a common problem there. They have a template scraper for converting PDF documents to XML (a more structured format) – if you can understand a little bit of programming then you can try to adapt it to your own purposes.

6) If it’s held by a public body and you have time: a well-written FOI request

Do you need all the data in the PDF or just some? Is that data available elsewhere? Try an advanced search using a phrase from the data in quotes and adding filetype:xls to see if you can find the spreadsheet it comes from. Or submit an FOI request for the data stipulating that it be provided in spreadsheet or CSV (comma separated values) format (if the PDF was supplied in response to an FOI request in the first place, go back and ask for the information to be provided in spreadsheet or CSV (comma separated values) format). 

It’s a good idea to also ask how the information is stored, including any software used, as you can check with the software vendor how easily the information can be extracted and bat away any excuses the body may come back at you with.

7) Add your own here

There must be others – tell me your own tips.

UPDATE: On Twitter: Simon Rogers uses Acrobat Pro; Kevin Anderson uses Omnipage. And Jack Schofield uses Zamzar.