public static bool CheckPersonName(String s)
{
if (s.Contains('.'))
{
return(true);
}
bool hasSpaces = s.Trim().Any(Char.IsWhiteSpace);
if (!hasSpaces)
{
return(false);
}
string[] words = Regex.Split(s, @"[\,\s\n]+");
if (TextHelpers.CanBePatronymic(words[words.Length - 1]))
{
return(true);
}
if (words.Count() != 3)
{
var predictedField = ColumnByDataPredictor.PredictByString(s);
if (!HeaderHelpers.IsNameDeclarationField(predictedField))
{
return(false);
}
}
return(true);
}
void DivideNameAndOccupation()
{
var nameCell = GetDeclarationField(DeclarationField.NameAndOccupationOrRelativeType);
NameDocPosition = adapter.GetDocumentPosition(GetRowIndex(), nameCell.Col);
string v = nameCell.GetText(true);
if (DataHelper.IsEmptyValue(v))
{
return;
}
if (DataHelper.IsRelativeInfo(v))
{
SetRelative(v);
}
else
{
string pattern = @"\s+\p{Pd}\s+"; // UnicodeCategory.DashPunctuation
v = Regex.Replace(v, @"\d+\.\s+", "");
string[] two_parts = Regex.Split(v, pattern);
string clean_v = Regex.Replace(v, pattern, " ");
string[] words = Regex.Split(clean_v, @"[\,\s\n]+");
if (words.Length >= 3 && TextHelpers.CanBePatronymic(words[2]) &&
!TextHelpers.MayContainsRole(words[0]) &&
!TextHelpers.MayContainsRole(words[1]))
{
// ex: "Рутенберг Дмитрий Анатольевич начальник управления"
PersonName = String.Join(" ", words.Take(3)).Trim();
Occupation = String.Join(" ", words.Skip(3)).Trim();
}
else if (TextHelpers.CanBePatronymic(words.Last()))
{
// ex: "начальник управления Рутенберг Дмитрий Анатольевич"
PersonName = String.Join(" ", words.Skip(words.Length - 3)).Trim();
Occupation = String.Join(" ", words.Take(words.Length - 3)).Trim();
}
else if (words.Length >= 2 && TextHelpers.CanBeInitials(words[1]) && TextHelpers.MayContainsRole(String.Join(" ", words.Skip(2)).Trim()))
{
// ex: "Головачева Н.В., заместитель"
PersonName = String.Join(" ", words.Take(2)).Trim();
Occupation = String.Join(" ", words.Skip(2)).Trim();
}
else if (two_parts.Length == 2)
{
PersonName = two_parts[0].Trim();
Occupation = String.Join(" - ", two_parts.Skip(1)).Trim();
}
else
{
throw new SmartParserException(
string.Format("Cannot parse name+occupation value {0} at row {1}", v, GetRowIndex()));
}
}
}
public static DeclarationField PredictEmptyColumnTitle(IAdapter adapter, Cell headerCell)
{
List <string> texts = new List <string>();
int rowIndex = headerCell.Row + headerCell.MergedRowsCount;
const int maxRowToCollect = 10;
int numbers = 0;
int not_numbers = 0;
for (int i = 0; i < maxRowToCollect; i++)
{
var cells = adapter.GetDataCells(rowIndex, IAdapter.MaxColumnsCount);
string dummy;
if (adapter.IsSectionRow(rowIndex, cells, adapter.GetColsCount(), false, out dummy))
{
rowIndex += 1;
}
else if (IAdapter.IsNumbersRow(cells))
{
rowIndex += 1;
}
else
{
var c = adapter.GetCell(rowIndex, headerCell.Col);
if (c != null)
{
var txt = c.GetText(true);
if (txt.Length > 0)
{
texts.Add(txt);
int d;
if (int.TryParse(txt, out d))
{
numbers += 1;
}
else
{
not_numbers += 1;
}
}
rowIndex += c.MergedRowsCount;
}
else
{
rowIndex += 1;
}
}
if (rowIndex >= adapter.GetRowsCount())
{
break;
}
}
var field = DeclarationField.None;
if (texts.Count == 1 && headerCell.Col == 0 && TextHelpers.CanBePatronymic(texts[0]))
{
// not enough data, if texts.Count == 1
field = DeclarationField.NameOrRelativeType;
}
else if (headerCell.Col == 0 && numbers > not_numbers)
{
field = DeclarationField.DeclarantIndex;
}
else
{
field = PredictByStrings(texts);
if (field == DeclarationField.NameOrRelativeType && String.Join(" ", texts).Contains(","))
{
field = DeclarationField.NameAndOccupationOrRelativeType;
}
}
if (headerCell.TextAbove != null && ((field & DeclarationField.AllOwnTypes) > 0))
{
string h = headerCell.TextAbove;
// AllOwnTypes defined from
field &= ~DeclarationField.AllOwnTypes;
if (HeaderHelpers.IsMixedColumn(h))
{
field |= DeclarationField.Mixed;
}
else if (HeaderHelpers.IsStateColumn(h))
{
field |= DeclarationField.State;
}
else if (HeaderHelpers.IsOwnedColumn(h))
{
field |= DeclarationField.Owned;
}
}
if (field == DeclarationField.NameOrRelativeType)
{
if (TextHelpers.MayContainsRole(String.Join(" ", texts)))
{
field = DeclarationField.NameAndOccupationOrRelativeType;
}
}
Logger.Debug(string.Format("predict by {0} -> {1}",
String.Join("\\n", texts), field));
return(field);
}
public static bool CheckMergeRow(List <string> row1, List <string> row2)
{
if (row1.Count != row2.Count)
{
return(false);
}
for (int i = 0; i < row1.Count; ++i)
{
var tokens1 = TokenizeCellText(row1[i]);
var tokens2 = TokenizeCellText(row2[i]);
if (tokens1.Count > 0 && tokens2.Count > 0)
{
string lastWord = tokens1.Last();
string firstWord = tokens2.First();
if (lastWord.Length > 0 && firstWord.Length > 0)
{
string joinExplanation = "";
if (Bigrams.ContainsKey(lastWord + " " + firstWord))
{
joinExplanation = "frequent bigram";
}
if (Regex.Matches(lastWord, @".+\p{Pd}$").Count > 0 &&
Char.IsLower(firstWord[0])
)
{
joinExplanation = "word break regexp";
}
if (tokens1.Count + tokens2.Count == 3 &&
TextHelpers.CanBePatronymic(tokens2[tokens2.Count - 1]) &&
!tokens2[tokens2.Count - 1].Contains('.') &&
Char.IsUpper(tokens1[0][0])
)
{
joinExplanation = "person regexp";
}
if (TextHelpers.MayContainsRole(string.Join(" ", tokens1)) &&
TextHelpers.CanBePatronymic(tokens2.Last()) &&
Char.IsUpper(tokens2[0][0]) &&
tokens1.All(x => !TextHelpers.CanBePatronymic(x))
)
{
joinExplanation = "role and person regexp";
}
if (Regex.Match(string.Join(" ", tokens1), @".+\([^\)]+", RegexOptions.Singleline).Success&&
Regex.Match(string.Join(" ", tokens2), @"^[^\(]+\).*", RegexOptions.Singleline).Success)
{
joinExplanation = "non-closed ) regexp";
}
if (firstWord.Trim()[0] == '(')
{
joinExplanation = "open ( regexp";
}
if (joinExplanation != "")
{
Logger.Debug(string.Format(
"Join rows using {0} on cells \"{1}\" and \"{2}\"",
joinExplanation,
row1[i].ReplaceEolnWithSpace(),
row2[i].ReplaceEolnWithSpace()));
return(true);
}
}
}
}
return(false);
}
bool DivideNameAndOccupation(Cell nameCell)
{
NameDocPosition = adapter.GetDocumentPosition(GetRowIndex(), nameCell.Col);
string v = nameCell.GetText(true);
if (DataHelper.IsEmptyValue(v))
{
return(true);
}
if (DataHelper.IsRelativeInfo(v))
{
SetRelative(v);
}
else
{
string pattern = @"\s+\p{Pd}\s+"; // UnicodeCategory.DashPunctuation
v = Regex.Replace(v, @"\d+\.\s+", "");
string[] two_parts = Regex.Split(v, pattern);
string clean_v = Regex.Replace(v, pattern, " ");
string[] words = Regex.Split(clean_v, @"[\,\s\n]+");
if (words.Length >= 3 && TextHelpers.CanBePatronymic(words[2]) &&
!TextHelpers.MayContainsRole(words[0]) &&
!TextHelpers.MayContainsRole(words[1]))
{
// ex: "Рутенберг Дмитрий Анатольевич начальник управления"
PersonName = String.Join(" ", words.Take(3)).Trim();
Occupation = String.Join(" ", words.Skip(3)).Trim();
}
else if (TextHelpers.CanBePatronymic(words.Last()))
{
// ex: "начальник управления Рутенберг Дмитрий Анатольевич"
PersonName = String.Join(" ", words.Skip(words.Length - 3)).Trim();
Occupation = String.Join(" ", words.Take(words.Length - 3)).Trim();
}
else if (Regex.Match(v, @"\w\.\w\.,").Success)
{
// ex: "Головачева Н.В., заместитель"
var match = Regex.Match(v, @"\w\.\w\.,");
PersonName = v.Substring(0, match.Index + match.Length - 1).Trim();
Occupation = v.Substring(match.Index + match.Length).Trim();
}
else if (words.Length >= 2 && TextHelpers.CanBeInitials(words[1]) &&
TextHelpers.MayContainsRole(String.Join(" ", words.Skip(2)).Trim()))
{
// ex: "Головачева Н.В., заместитель"
PersonName = String.Join(" ", words.Take(2)).Trim();
Occupation = String.Join(" ", words.Skip(2)).Trim();
}
else if (two_parts.Length == 2)
{
PersonName = two_parts[0].Trim();
Occupation = String.Join(" - ", two_parts.Skip(1)).Trim();
}
else
{
// maybe PDF has split cells (table on different pages)
// example file: "5966/14 Upravlenie delami.pdf" converted to docx
Logger.Error(string.Format("Cannot parse name+occupation value {0} at row {1}", v, GetRowIndex()));
return(false);
}
}
return(true);
}