private void processDocument()
{
// Document document = app.Documents.Open(@"C:\User_Pradeep\Transcript2.doc", ReadOnly: true);
Document document = app.Documents.Open(@"C:\User_Pradeep\Transcript3C.doc", ReadOnly: true); //test version
document.Activate();
applicationTime.Start(); //Start stopwatch
HashSet <string> processedWordList = new HashSet <string>(); // 10/25
string textOftheWord = "";
bool validWord = false;
//Console.SetWindowSize();
for (int i = 1; i < document.Words.Count; i++)
{
textOftheWord = document.Words[i].Text.ToString().Trim().ToLower(); // 10/26
//if ((textOftheWord.Length > 3) && (textOftheWord.Contains("_________") == false) && (textOftheWord != "-")) // 10/26 - 03/28/2017
//Check length of the word is > 3 and word contains letters, numbers or $,#,* only
if ((textOftheWord.Length > 2) && (Regex.IsMatch(textOftheWord, @"^[a-zA-Z0-9\$#*]+$"))) // 03/28/2017
{
validWord = true;
string processedWord = processedWordList.FirstOrDefault(w => w == textOftheWord); // 10/25
if (processedWord == null)
{
//If word is all numbers and equal to number at the begining of the sentence then it is a question number, ignore!
if (Regex.IsMatch(textOftheWord, @"^[0-9]"))
{
Range sentenceRange = document.Range(Start: document.Words[i].Start, End: document.Content.End); // 03/27/2017
string textOfTheCurrentSentence = sentenceRange.Sentences.First.Text; // 03/27/2017
string firstWordOfTheCurrentSentence = textOfTheCurrentSentence.Substring(0, textOftheWord.Length);
int valueInTheWord = 0;
bool isNumber = Int32.TryParse(firstWordOfTheCurrentSentence, out valueInTheWord); // 10/22
if ((isNumber) && (textOftheWord == firstWordOfTheCurrentSentence))
{
validWord = false;
}
}
if (validWord) //((!parsed) ) 03/28/2017
{
Range currentRange = document.Range(Start: document.Words[i].Start, End: document.Words[i].End); // 10/22
int pageNumberOfTheCurrentRange = currentRange.Information[WdInformation.wdActiveEndPageNumber]; // 10/22
if ((pageNumberOfTheCurrentRange > 1)) // 10/26
{
Console.WriteLine("Scanning transcript- processing word # " + i + " - In page # " + pageNumberOfTheCurrentRange);
var CustomWord = new TranscriptWord();
// Not in the array, it is a new word
processedWordList.Add(textOftheWord); // 10/24
CustomWord.Name = textOftheWord;
int wordFoundFrequency = 0;
Range searchRange = document.Range(Start: document.Words[i].Start, End: document.Content.End); // 10/23
searchRange.Find.Forward = true; // 10/26 all range words
searchRange.Find.Text = textOftheWord;
currentWord = textOftheWord;
searchRange.Find.Execute(MatchWholeWord: true);
int currentLineNumber = 0;
int currentPageNumber = 0;
int pageNumberOfTheWord = 0;
int lineNumberOfTheWord = 0;
while (searchRange.Find.Found)
{
Console.WriteLine("Looking for word : " + currentWord);
// Get currently searched sentence and extract the first word
string textOfTheSearchedRangeSentence = searchRange.Sentences.First.Text;
string firstWordOfTheSearchedSentence = textOfTheSearchedRangeSentence.Substring(0, textOftheWord.Length);
//If begining of the sentence is a number, it usually is a question number, now if it is a number and matches the searched text, it definitely
// cannot be a regular word, it got to be a question number, so ignore!
if ((firstWordOfTheSearchedSentence == textOftheWord) && (Regex.IsMatch(firstWordOfTheSearchedSentence, @"^[0-9]")))
{
}
else
{
wordFoundFrequency++;
currentPageNumber = searchRange.Information[WdInformation.wdActiveEndPageNumber];
currentLineNumber = searchRange.Information[WdInformation.wdFirstCharacterLineNumber];
if (currentPageNumber != 1) // Page # 1 is cover page
{
//Check whether current word is repeating in the same page and line number, if not, create "Occurrence" object.
if (wordFoundFrequency > 1)
{
if (pageNumberOfTheWord != currentPageNumber || lineNumberOfTheWord != currentLineNumber)
{
pageNumberOfTheWord = currentPageNumber;
lineNumberOfTheWord = currentLineNumber;
var CustomOccurrence = new Occurrence {
CustomPageNumber = pageNumberOfTheWord, CustomLineNumber = lineNumberOfTheWord
};
CustomWord.PageAndLine.Add(CustomOccurrence);
}
}
else
{
pageNumberOfTheWord = currentPageNumber;
lineNumberOfTheWord = currentLineNumber;
var CustomOccurrence = new Occurrence {
CustomPageNumber = pageNumberOfTheWord, CustomLineNumber = lineNumberOfTheWord
};
CustomWord.PageAndLine.Add(CustomOccurrence);
}
}
}
searchRange.Find.Execute(MatchWholeWord: true);
}
CustomWord.Frequency = wordFoundFrequency;
CustomWordDirectory.Add(CustomWord);
}
}
}
}
}
document.Close();
}
private void processDocument2(ref string[] finalDeDupedWordList)
{
Document document = app.Documents.Open(@"C:\User_Pradeep\Transcript3C.doc", ReadOnly: true); //test version
document.Activate();
HashSet <string> processedWordList = new HashSet <string>();
string finalSearchWord = "";
//Console.SetWindowSize();
for (int i = 0; i < finalDeDupedWordList.Length; i++)
{
finalSearchWord = finalDeDupedWordList[i].Trim();
//Words and sentences within double quotes(" ") should be identified, quotes should be removed in order to preserve correct print order (i.e. #'s $'s digits and actual words)
if (Regex.IsMatch(finalSearchWord, @"^[a-zA-Z0-9\$#]"))
{
}
else
{
//If the word is not all spaces and does not starts with one of the allowed charactors, then remove first position, could be a starting double quote or single quote
if (finalSearchWord != "")
{
finalSearchWord = finalSearchWord.Remove(0, 1);
}
}
//If the last position of the word in not one of the allowed charactors, then remove it, could be a closing double quote or single quote
if (finalSearchWord != "")
{
if (Regex.IsMatch(finalSearchWord.Substring(finalSearchWord.Length - 1, 1), @"[a-zA-Z0-9\$#]"))
{
}
else
{
finalSearchWord = finalSearchWord.Remove(finalSearchWord.Length - 1, 1);
}
}
//Cleanup any spaces created by above process if any
finalSearchWord.Trim();
if (finalSearchWord.Length > 2)
{
string processedWord = processedWordList.FirstOrDefault(w => w == finalSearchWord); // 10/25
if (processedWord == null)
{
Console.WriteLine("Scanning transcript- processing word # " + i);
var CustomWord = new TranscriptWord();
// Not in the array, it is a new word so add to processed word list and start processing....
processedWordList.Add(finalSearchWord);
CustomWord.Name = finalSearchWord;
int wordFoundFrequency = 0;
Range searchRange = document.Range(Start: document.Content.Start, End: document.Content.End); //Look for the word from start of the transcript to end
searchRange.Find.Forward = true; // 10/26 all range words
searchRange.Find.MatchCase = true; // 04/14/2016
searchRange.Find.Text = finalSearchWord;
currentWord = finalSearchWord;
searchRange.Find.Execute(MatchWholeWord: true);
int currentLineNumber = 0;
int currentPageNumber = 0;
int pageNumberOfTheWord = 0;
int lineNumberOfTheWord = 0;
while (searchRange.Find.Found)
{
Console.WriteLine("Looking for word : " + currentWord);
// Get current sentence being searched and extract the first word
string textOfTheSearchedRangeSentence = searchRange.Sentences.First.Text;
string firstWordOfTheSearchedSentence = textOfTheSearchedRangeSentence.Substring(0, finalSearchWord.Length);
//If sentence starts with a number, it usually is a question number, now if it is a number and matches the searched text, it definitely
// cannot be a regular word, it got to be a question number, so ignore!
if ((firstWordOfTheSearchedSentence == finalSearchWord) && (Regex.IsMatch(firstWordOfTheSearchedSentence, @"^[0-9]")))
{
}
else
{
// if (currentPageNumber != 1) // Page # 1 is cover page
// {
wordFoundFrequency++;
currentPageNumber = searchRange.Information[WdInformation.wdActiveEndPageNumber];
currentLineNumber = searchRange.Information[WdInformation.wdFirstCharacterLineNumber];
//Check whether current word is repeating in the same page and line number, if not, create the "Occurrence" object.
if (wordFoundFrequency > 1)
{
if (pageNumberOfTheWord != currentPageNumber || lineNumberOfTheWord != currentLineNumber)
{
pageNumberOfTheWord = currentPageNumber;
lineNumberOfTheWord = currentLineNumber;
var CustomOccurrence = new Occurrence {
CustomPageNumber = pageNumberOfTheWord, CustomLineNumber = lineNumberOfTheWord
};
CustomWord.PageAndLine.Add(CustomOccurrence);
}
}
else
{
pageNumberOfTheWord = currentPageNumber;
lineNumberOfTheWord = currentLineNumber;
var CustomOccurrence = new Occurrence {
CustomPageNumber = pageNumberOfTheWord, CustomLineNumber = lineNumberOfTheWord
};
CustomWord.PageAndLine.Add(CustomOccurrence);
}
// }
}
searchRange.Find.Execute(MatchWholeWord: true);
}
CustomWord.Frequency = wordFoundFrequency;
CustomWordDirectory.Add(CustomWord);
}
}
}
document.Close();
}
