public static async Task HighlightHits(ProcessedDocument procesedDocument, Stream originalImage, CloudBlockBlob markedUpImage)
{
using (Image <Rgba32> image = Image.Load <Rgba32>(originalImage))
{
foreach (var line in procesedDocument.Lines)
{
int maxX = max(line.BoundingBox.x1, line.BoundingBox.x2, line.BoundingBox.x3, line.BoundingBox.x4);
int maxY = max(line.BoundingBox.y1, line.BoundingBox.y2, line.BoundingBox.y3, line.BoundingBox.y4);
int minX = min(line.BoundingBox.x1, line.BoundingBox.x2, line.BoundingBox.x3, line.BoundingBox.x4);
int minY = min(line.BoundingBox.y1, line.BoundingBox.y2, line.BoundingBox.y3, line.BoundingBox.y4);
using (Image <Rgba32> highlight = new Image <Rgba32>(new Configuration(), maxX - minX, maxY - minY, Rgba32.Yellow))
{
image.Mutate(x =>
{
x.DrawImage(highlight, .5f, new SixLabors.Primitives.Point(minX, minY));
});
}
}
markedUpImage.Properties.ContentType = "image/jpeg";
using (var outboundStream = await markedUpImage.OpenWriteAsync())
{
image.Save(outboundStream, ImageFormats.Jpeg);
}
}
}
public async Task <string> TestDocument(Document document)
{
ProcessedDocument processedDocument = _iPreProcessTextService.PreProcessDocument(document);
string topic = _iLearningService.Classify(processedDocument);
return(topic);
}
public async Task <Document> TrainDocument(Document document)
{
ProcessedDocument processedDocument = _iPreProcessTextService.PreProcessDocument(document);
_iLearningService.Train(processedDocument);
return(await _documentRepository.AddAsync(document));
}
private static DataField FindMoneyLine(ProcessedDocument document, string name, string title, MortgageApplicationDocument mortgageApplication, params string[] titlesToFind)
{
DataField dataField = FindDataLine(document, name, title, mortgageApplication, titlesToFind);
if (dataField != null)
{
double tmpVal;
if (double.TryParse(CLEANUPMONEY.Replace(dataField.Value, ""), out tmpVal))
{
dataField.Value = tmpVal.ToString("c");
}
}
return(dataField);
}
public string Classify(ProcessedDocument document)
{
Console.WriteLine("## Start classifying document ##");
_document = document;
//todo keep this value in cache
_nrDistinctWords = _iWordRepository.GetTotalCountOfDistinctWords();
CalculateProbabilityPerTopic("business");
CalculateProbabilityPerTopic("entertainment");
CalculateProbabilityPerTopic("politics");
CalculateProbabilityPerTopic("sport");
CalculateProbabilityPerTopic("tech");
Console.WriteLine("## Finished classifying document ##");
return(ChooseClass());
}
public ProcessedDocument PreProcessDocument(Document document)
{
Console.WriteLine("## Pre processing the document ##");
string text = document.Text;
LowerText(text);
ReplaceNonAlphaNumericCharacters(text);
List <string> bagOfWords = SplitInList(text);
RemoveSingleLetters(bagOfWords);
RemoveStopWords(bagOfWords);
ProcessedDocument processedDocument = new ProcessedDocument(document, BuildWordList(bagOfWords, document));
Console.WriteLine("## Finished pre processing the document ##");
return(processedDocument);
}
public static async Task Run([BlobTrigger("parseddocuments/w2/{name}", Connection = "Documents")] CloudBlockBlob parsedBlob,
[Blob("parseddocuments", Connection = "Documents")] CloudBlobContainer blobDirectory,
[Blob("rawdocuments/{name}", FileAccess.Read)] Stream originalImage,
[Blob("parseddocuments/markedup/{name}", FileAccess.ReadWrite)] CloudBlockBlob markedUpImage,
ILogger log, Microsoft.Azure.WebJobs.ExecutionContext context, CancellationToken cancellationToken)
{
try
{
MortgageApplicationDocument mortgageApplicationDocument = new MortgageApplicationDocument();
mortgageApplicationDocument.PopuplateFromBlobProperties(parsedBlob);
mortgageApplicationDocument.HasParsedResults = true;
mortgageApplicationDocument.DocumentType = "w2";
string jsonContents = await parsedBlob.DownloadTextAsync();
ProcessedDocument document = JsonConvert.DeserializeObject <ProcessedDocument>(jsonContents);
CloudBlockBlob finalBlob = blobDirectory.GetBlockBlobReference(parsedBlob.Name.Replace("w2/", "final/"));
mortgageApplicationDocument.Status = MortgageApplicationStatus.Processed;
mortgageApplicationDocument.DateProcessed = DateTimeOffset.UtcNow;
mortgageApplicationDocument.SetBlobProperties(finalBlob);
FindDataLine(document, "SSN", "Social Security Number", mortgageApplicationDocument, "a. Employee's social security number", "social security number", "Employee's soc. sec. number", "Employee's social security number");
FindMoneyLine(document, "Wages", "Total Wages", mortgageApplicationDocument, "Wages");
FindMoneyLine(document, "FedTax", "Federal Tax Witholding", mortgageApplicationDocument, "2. Federal income tax", "Federal income tax withheld");
FindDataLine(document, "CopyType", "Copy Type", mortgageApplicationDocument, "Copy");
FindMoneyLine(document, "SocialSecurityWage", "Social Security Wage", mortgageApplicationDocument, "Social security wages");
FindMoneyLine(document, "SocialSecurityTax", "Social Security Tax", mortgageApplicationDocument, "Social security tax withheld");
FindMoneyLine(document, "MedicareWage", "Medicare Wage", mortgageApplicationDocument, "Medicare wages and tips");
FindMoneyLine(document, "MedicareTax", "Medicare Tax", mortgageApplicationDocument, "Medicare tax withheld");
FindMoneyLine(document, "SocialSecurityTips", "Social Security Tips", mortgageApplicationDocument, "Social security tips");
FindDataLine(document, "EmployerDetails", "Employer Details", mortgageApplicationDocument, "Employer's name, address, and ZIP code");
FindDataLine(document, "EmployeeDetails", "Employee Details", mortgageApplicationDocument, "Employee's first name and initial", "Last name", "Employee's name, address, and ZIP code", "Employee's name (first, middle indial, last)");
await finalBlob.UploadTextAsync(JsonConvert.SerializeObject(mortgageApplicationDocument));
await MarkupService.HighlightHits(document, originalImage, markedUpImage);
}
catch (Exception ex)
{
log.LogError(ex, "W2 Processor Failed");
}
}
private static DataField FindDataLine(ProcessedDocument document, string name, string title, MortgageApplicationDocument mortgageApplication, params string[] titlesToFind)
{
ProcessedLine foundLine = null;
foreach (var titleToFind in titlesToFind)
{
string cleanTitle = CLEANUP.Replace(titleToFind, "");
foundLine = document.Lines.Where(l => CLEANUP.Replace(l.Text, "").IndexOf(cleanTitle, StringComparison.OrdinalIgnoreCase) > -1).FirstOrDefault();
if (foundLine != null)
{
break;
}
}
var dataLine = foundLine?.BoundingBox.FindClosestsBelow(document.Lines) ?? null;
if (dataLine == null)
{
return(null);
}
DataField dataField = new DataField()
{
FieldName = name,
FieldTitle = title,
LabelBox = foundLine.BoundingBox,
ValueBox = dataLine.BoundingBox,
Value = dataLine.Text
};
if (mortgageApplication.DataFields == null)
{
mortgageApplication.DataFields = new List <DataField>();
}
mortgageApplication.DataFields.Add(dataField);
return(dataField);
}
//todo: build a dict with distinct words before insert in the db to save time
public void Train(ProcessedDocument document)
{
_iWordRepository.AddMultipleWords(document.Words);
return;
}
