public IEnumerable <AnalyzedText> Analyze(OcrResult ocrResult)
{
var continuousText = ocrResult.AsContinuousText().ReplaceFaultyCharacters(Constants.NumericAnalysisOcrFixDictionary);
var amounts = Constants.TextAnalysisConfiguration.AmountRegexes
.Select(x => Regex.Matches(continuousText, x))
.SelectMany(y => y.Cast <Match>()
.Select(x => x.Value));
amounts = amounts.Select(x => Regex.Replace(x, Constants.TextAnalysisConfiguration.AmountIgnoreRegex, string.Empty));
amounts = amounts.Select(x =>
{
if ((x.Count(c => c == '.') > 0 && x.Count(c => c == ',') > 0) || x.Count(c => c == '.') > 1 || x.Count(c => c == ',') > 1)
{
Regex regex = new Regex("(,|\\.)");
return(regex.Replace(x, string.Empty, 1));
}
return(x);
});
amounts = amounts.Replace(",", ".");
return(amounts.Select(x => new AnalyzedText()
{
Text = x,
TextType = TextType.Amount.ToString(),
BoundingBox = ocrResult.BoundingBox
})
.Distinct());
}
private string ParseTextDefinition(TextDefinition textDefinition, OcrResult ocrResult)
{
string text = string.Empty;
switch (textDefinition.GetAs)
{
case GetTextAs.Text:
text = ocrResult.AsString();
break;
case GetTextAs.Continuous:
text = ocrResult.AsContinuousText();
break;
}
if (textDefinition.Replace != null && textDefinition.Replace.Length > 0)
{
foreach (var replaceDefinition in textDefinition.Replace)
{
text = ParseReplaceDefinition(replaceDefinition, text);
}
}
return(text);
}
public IEnumerable <AnalyzedText> Analyze(OcrResult ocrResult)
{
var possibleNames = Constants.TextAnalysisConfiguration.PercentageRegexes
.Select(x => Regex.Matches(ocrResult.AsContinuousText().Replace("-", ""), x, RegexOptions.IgnoreCase))
.SelectMany(x => x.Cast <Match>());
return(possibleNames.Select(x => new AnalyzedText()
{
Text = x.Value,
TextType = TextType.Percentage.ToString(),
BoundingBox = ocrResult.BoundingBox
}));
}
public IEnumerable <AnalyzedText> Analyze(OcrResult ocrResult)
{
var continuousText = ocrResult.AsContinuousText().ReplaceFaultyCharacters(Constants.NumericAnalysisOcrFixDictionary);
var continuousNumericSequences = Constants.TextAnalysisConfiguration.NumberRegexes
.Select(x => Regex.Matches(continuousText.Replace("-", "").Replace(".00", "").Replace(",00", ""), x))
.SelectMany(y => y.Cast <Match>()
.Select(x => (x.Value)));
return(continuousNumericSequences
.Select(x => new AnalyzedText()
{
Text = x,
TextType = TextType.Number.ToString(),
BoundingBox = ocrResult.BoundingBox
}));
}
public IEnumerable <AnalyzedText> Analyze(OcrResult ocrResult)
{
var possibleDates = Constants.TextAnalysisConfiguration.DateRegexes
.Select(x => Regex.Matches(ocrResult.AsContinuousText(), x))
.SelectMany(y => y.Cast <Match>()
.Select(x => x.Value));
return(possibleDates
.Select(x => new AnalyzedText()
{
Text = TryParseAndFixDate(x),
TextType = TextType.Date.ToString(),
BoundingBox = ocrResult.BoundingBox
})
.Where(x => !string.IsNullOrWhiteSpace(x.Text))
.Distinct());
}
public IEnumerable <AnalyzedText> Analyze(OcrResult ocrResult)
{
var firstPassPageNumbers = Constants.TextAnalysisConfiguration.FirstPassPageNumberRexeges
.Select(x => Regex.Matches(ocrResult.AsContinuousText(false), x, RegexOptions.IgnoreCase))
.SelectMany(y => y.Cast <Match>()
.Select(x => (x.Groups["page"])))
.FirstOrDefault()
?.Value;
if (string.IsNullOrWhiteSpace(firstPassPageNumbers))
{
return new AnalyzedText[] { }
}
;
var secondPassPageNumber = Constants.TextAnalysisConfiguration.SecondPassPageNumberRegexes
.Select(x => Regex.Matches(firstPassPageNumbers, x, RegexOptions.IgnoreCase))
.SelectMany(y => y.Cast <Match>()
.Select(x => (x.Groups["page"])))
.FirstOrDefault()
?.Value;
if (string.IsNullOrWhiteSpace(secondPassPageNumber))
{
return new AnalyzedText[] { }
}
;
return(new AnalyzedText[]
{
new AnalyzedText()
{
Text = secondPassPageNumber,
TextType = TextType.PageNumber.ToString(),
BoundingBox = ocrResult.BoundingBox
}
});
}
}
}
public IEnumerable <AnalyzedText> Analyze(OcrResult ocrResult)
{
var files = Directory.GetFiles((System.AppDomain.CurrentDomain.RelativeSearchPath ?? System.AppDomain.CurrentDomain.BaseDirectory) + "\\" + Constants.TextTypeDataFilePath);
return(files.SelectMany(filePath =>
{
var fileName = Path.GetFileNameWithoutExtension(filePath);
var matchedResult = ocrResult.BinarySearchMatchFromFile(fileName, filePath, false);
if (matchedResult.Count() == 0)
{
return matchedResult;
}
var fromFileConfiguration = Constants.FromFileConfigurations.FirstOrDefault(x => x.TextType.Equals(fileName, System.StringComparison.InvariantCultureIgnoreCase));
if (fromFileConfiguration != null)
{
if (!string.IsNullOrWhiteSpace(fromFileConfiguration.PostCaptureRegex))
{
return matchedResult
.Select(x =>
{
var matches = new Regex(fromFileConfiguration.PostCaptureRegex.Replace("{textpart}", x.Text)).Matches(ocrResult.AsContinuousText());
return matches;
})
.SelectMany(x => x.Cast <Match>()
.Select(y => new AnalyzedText()
{
BoundingBox = ocrResult.BoundingBox,
TextType = fileName,
Text = $"{y.Groups["textpart"].Value} {y.Groups["numberpart"]?.Value?.PadLeft(1)}".Trim()
}));
}
}
return matchedResult;
}));
}