private static void ParseTextBlocks(XElement pageElement, TextGeometryModel model)
{
foreach (var textBlockElement in pageElement.TextBlockElements())
{
var textBlock = textBlockElement.ParseTextBlock();
if (!textBlock.IsEmpty())
{
model.AddTextBlock(textBlock);
}
}
}
//return HashSet coincidered words
private static IEnumerable <GMWord> EnumerateWordsFromFineReader(TextGeometryModel fineReaderModel, TextGeometryModel sampleModel,
LinesHashInGeometryModel tesseractCache)
{
foreach (var block in fineReaderModel.TextBlocks())
{
var blockBox = block.BoundingBox;
var thisBlock =
sampleModel.AddTextBlock(new BoundingBox(blockBox.XMin, blockBox.YMin, blockBox.XMax, blockBox.YMax));
foreach (var paragraph in block.Paragraphs())
{
var paragraphBox = paragraph.BoundingBox;
var thisParagraph =
thisBlock.AddParagraph(new BoundingBox(paragraphBox.XMin, paragraphBox.YMin, paragraphBox.XMax,
paragraphBox.YMax));
foreach (var line in paragraph.Lines())
{
var lineBox = line.BoundingBox;
var thisLine =
thisParagraph.AddLine(new BoundingBox(lineBox.XMin, lineBox.YMin, lineBox.XMax, lineBox.YMax));
foreach (var word in line.Words())
{
var wordBox = word.BoundingBox;
var accuracy = 50;
if (tesseractCache.Contains(word, String.Compare))
{
yield return(word);
accuracy = 100;
}
var thisWord = new GMWord(wordBox, word.Text, accuracy);
thisLine.AddWord(thisWord);
}
}
}
foreach (var word in block.StandaloneWords())
{
var wordBox = word.BoundingBox;
var accuracy = 50;
if (tesseractCache.Contains(word, String.Compare))
{
yield return(word);
accuracy = 100;
}
var thisWord = new GMWord(wordBox, word.Text, accuracy);
thisBlock.AddStandaloneWord(thisWord);
}
}
}
private static void AddWordsFromTesseract(TextGeometryModel tesseractModel, TextGeometryModel sampleModel, List <GMWord> set)
{
var box = new BoundingBox(tesseractModel.PageBox.XMin, tesseractModel.PageBox.YMin,
tesseractModel.PageBox.XMax, tesseractModel.PageBox.YMax);
var block = sampleModel.AddTextBlock(box);
var paragraph = block.AddParagraph(box);
var line = paragraph.AddLine(box);
foreach (var word in tesseractModel.Words())
{
//compare with special comparator, which compare boundingboxes with inaccuracy
if (!set.Contains(word, GMWordComparerWithInaccuracy.Instance))
{
var wordBox = new BoundingBox(word.BoundingBox.XMin, word.BoundingBox.YMin, word.BoundingBox.XMax, word.BoundingBox.YMax);
line.AddWord(new GMWord(wordBox, word.Text, 50));
}
}
}