Hi Stuart, give ExtractText a try using
ExtractText -nonSeq -html and inspect the result. It does a fairly good job for the sample PDF. The reason why I'm suggesting the -html option is that Paragraphs of text are written out within a <p> tag. You can build on org.apache.pdfbox.util.PDFText2HTML and use that as a starter for enhancements if needed. As there is no structure information within the PDF these can not be taken to help enhancing the text extraction. The fact that you see boxes around text are graphics but not represented e.g. as articles. Of course you could try to use drawing commands as a hint but that's a lot of effort. Maybe the functionality available is already sufficient for you. BR Maruan Sahyoun Am 12.06.2013 um 22:04 schrieb Stuart Coleman <[email protected]>: > Hi, > > Thanks for the quick response. I have uploaded one of the pages at > > https://www.dropbox.com/s/7cqlul61pk53gd1/testpage.pdf > > Any pointers how I could extend things would be great. > > Thanks, > Stuart > > On 12 Jun 2013, at 20:52, Maruan Sahyoun wrote: > >> Hi Stuart, >> >> from the screenshot it's not clear how the PDF is layer out. In general >> there are some structures like article threads which PDFBox supports for >> text extraction. Also PDFBox is able to handle bookmarks, annotations …. >> although some of these informations are not taken into account when using >> the standard ExtractText functionality. But it's possible to extend existing >> functions. With the PDF as a sample it would be easier to understand which >> PDF features is used for the box and give you some additional hints. As the >> mailing list doesn't allow for PDF attachments please upload a sample at a >> public location if possible. >> >> BR >> Maruan Sahyoun >> >> Am 12.06.2013 um 21:35 schrieb Stuart Coleman <[email protected]>: >> >>> Hi, >>> >>> I have a PDF file which I am trying to extract text from. Unfortunately the >>> document is non sequential and has various boxes with supplementary >>> content. When I open the file in Acrobat Reader, Reader seems to be able to >>> distinguish these features and can surround them with a blue bounding box. >>> I would like to be able to extract text by area from within these bounding >>> boxes? Is PDFBox capable of detecting these features also? >>> >>> I have attached a screenshot showing the style of box I am referring to >>> (top right hand corner) >>> >>> Thanks >>> Stuart >>> >>> <Screen Shot 2013-06-12 at 20.17.31.png> >> >
