iTextSharp：执行GetTextFromPage时指定编码

Question

I'm using iTextSharp to extract some informations from a PDF file. Everything is almost perfect(quite impressed in fact), I just have some issue with some words.

By example, in the PDF, I've the following sentence:

Dès la fin de soirée, [...]

When I look the PDF, I see exactly that, but when I receive the text from the following code:

ITextExtractionStrategy strategy = new SimpleTextExtractionStrategy();
for (int page = 1; page <= pdfReader.NumberOfPages; page++)
{
    string currentPageText = PdfTextExtractor.GetTextFromPage(pdfReader, page, strategy);
    if (currentPageText.Contains(PAGE_MARKER))
    {
        return currentPageText;
    }
}

The text in question is the following:

Dès la fi n de soirée, [...]

It's strange, but the "fi" are in fact only one character, and there is a space.

When I open the same PDF in Foxit reader or Adobe acrobat, it's looking fine, but if I copy paste the text, I've the following text:

Dès la fi n de soirée, [...] (So the correct characters but with one space)

It's one example, but I've some another ones.

Any idea how to fix this?

Answer 1

For this to make sense, you need some background in pdf syntax.

In the most rudimentary form, a pdf document contains only the instructions needed to render a document in a viewer. In other words, there is not concept of "text" being rendered. Just something like "draw character 'A' at location 150, 877" and so on.

In fact, this is a snippet from a .pdf document (opened with a simple text editor)

[a, -28.7356, p, 27.2652, p, 27.2652, e, -27.2652, a, -28.7356, r, 64.6889, a, -28.7356, n, 27.2652, c, -38.7594, e, 444] TJ

TJ is the "draw text" instruction. The array contains pairs for characters and their kerning info.

Now, for any kind of text extraction to work (both in iText and in the copy/paste functionality of Foxit, Adobe, etc) you need a bit of guesswork. (A heuristic as it commonly called).

You need to decide when some characters get stuck together and form a word, and when two characters are far enough apart that there should be a space between them.

In your usecase, it seems that the distance between "n" and "i" is greater than the expected distance for that font.

Sadly, iText will not be able to (easily) help you there. Since the input document simply seems to be incorrect. Or rather most readers/viewers seem to be getting it wrong, so it is likely to simply be an issue in the pdf.

Of course, you can implement the TextExtractionStrategy. This class gives you access to the TextRenderInfo objects that contain the characters and graphic state in the pdf. Most TextExtractionStrategies will then check what the size of a space is in the font being used, and use that as a reference to decide when characters should be concatenated, and when they should be separated.

Lastly, if you want to delve deeper into this issue, you could always attach the input document.

Kind regards, Joris