PDF- Can text chunk contains 2 or more words? - pdf

Question

Im using LocationTextExtractionStrategy to render text from PDF.
Text is rendered in function called RenderText.
So my question is: Can one chunk contains more than 2 words ?
For example we have text:
'MKL is a helpfull person'
Can it be written in chunks like (the most important chunk is bolded):
MK
L
is a h
elpfull
per son
?
Below is the code i use for word separation.
Im doing the word separation during adding text(chunk from renderText function) to current line.
public class TextLineLocation
{
public float X { get; set; }
public float Y { get; set; }
public float Height { get; set; }
public float Width { get; set; }
private string Text;
private List<char> bannedSings = new List<char>() {' ',',', '.', '/', '|', Convert.ToChar(#"\"), ';', '(', ')', '*', '&', '^', '!','?' };
public void AddText(TextInfo text)
{
Text += text;
foreach (char sign in bannedSings)
{
//creating new word
if (text.textChunk.Text.Contains(sign))
{
string[] splittedText = text.textChunk.Text.Split(sign);
foreach (string val in splittedText)
{
//if its first element, add it to current word
if (splittedText[0] == val)
{
// if its space, just ignore...
if (splittedText[0] == " ")
{
continue;
}
wordList[wordList.Count - 1].Text += val;
wordList[wordList.Count - 1].Width += text.getFontWidth();
wordList[wordList.Count - 1].Height += text.getFontHeight();
}
else
{
//if it isnt a first element, create another word
wordList.Add(new WordLocation(text.textChunk.StartLocation[1], text.textChunk.StartLocation[0], text.getFontWidth(), text.getFontHeight(), val));
//TODO: what if chunk has more than 2 words separated ?
}
}
}
}
else
{
//update last word
wordList[wordList.Count-1].Text += text.textChunk.Text;
wordList[wordList.Count - 1].Width += text.getFontWidth();
wordList[wordList.Count - 1].Height += text.getFontHeight();
}
}
public List<WordLocation> wordList = new List<WordLocation>();
}

Answer 1

Not sure from what library LocationTextExtractionStrategy comes, or what it does exactly, but in the PDF representation itself you can group characters together in a "chunk".
How this is used totally depends on the program that produces the PDF: Some programs keep words together, some programs only group word fragments (for example for kerning), some program do other, random things.
So, if LocationTextExtractionStrategy returns these as chunks, you can't rely on anything. If LocationTextExtractionStrategy doesn't return these, but instead relies on spacing heuristics to group characters into chunks, then this will be as good as the heuristics are.
Bottom line: A PDF doesn't contain text, and contains glyphs and their position on the page. Trying to reconstruct text from it is and remains guesswork. You may get it to work in the majority of cases, but there'll always be PDFs where whatever you are doing fails.