\file GikoBayesian.pas
\brief \83x\83C\83W\83A\83\93\83t\83B\83\8b\83^
-$Id: GikoBayesian.pas,v 1.2 2004/10/21 01:20:34 yoffy Exp $
+
+$Id: GikoBayesian.pas,v 1.22 2009/01/31 15:47:15 h677 Exp $
}
+//! \95½\89¼\96¼\82ð\8e«\8f\91\82É\8aÜ\82ß\82È\82¢
+{$DEFINE GIKO_BAYESIAN_NO_HIRAGANA_DIC}
+
interface
//==================================================
uses
//==================================================
- Classes, IniFiles;
+ Classes;
//==================================================
type
\brief \89ð\90Í\8dÏ\82Ý\92P\8cê\83\8a\83X\83g
************************************************************}
// TWordCount = class( THashedStringList ) // \8c\83\92x
- TWordCount = class( TStringList ) // \92x
+ TWordCount = class( TStringList )
public
+ constructor Create;
destructor Destroy; override;
end;
\brief \83t\83B\83\8b\83^\83A\83\8b\83S\83\8a\83Y\83\80
************************************************************}
TGikoBayesianAlgorithm =
- (gbaPaulGraham, gbaGaryRonbinson{, gbaGaryRonbinsonFisher});
+ (gbaPaulGraham, gbaGaryRobinson, gbaGaryRobinsonFisher);
{!***********************************************************
\brief \83x\83C\83W\83A\83\93\83t\83B\83\8b\83^
************************************************************}
- TGikoBayesian = class( THashedStringList )
+// TGikoBayesian = class( THashedStringList ) // \8c\83\92x
+ TGikoBayesian = class( TStringList )
private
FFilePath : string; //!< \93Ç\82Ý\8d\9e\82ñ\82¾\83t\83@\83C\83\8b\83p\83X
function GetObject( const name : string ) : TWordInfo;
}
function CalcGaryRobinson( wordCount : TWordCount ) : Extended;
-// function CalcGaryRobinsonFisher( wordCount : TWordCount ) : Extended;
+ {!
+ \brief GaryRobinson-Fisher \96@\82É\8aî\82Ã\82¢\82Ä\95¶\8fÍ\82Ì\92\8d\96Ú\93x\82ð\8c\88\92è\82µ\82Ü\82·
+ \return \95¶\8fÍ\82Ì\92\8d\96Ú\93x (\92\8d\96Ú\82É\92l\82µ\82È\82¢ 0.0\81`1.0 \92\8d\96Ú\82·\82×\82«)
+ }
+ function CalcGaryRobinsonFisher( wordCount : TWordCount ) : Extended;
{!
\brief \95¶\8fÍ\82ð\89ð\90Í
function Parse(
const text : string;
wordCount : TWordCount;
- algorithm : TGikoBayesianAlgorithm = gbaGaryRonbinson
+ algorithm : TGikoBayesianAlgorithm = gbaGaryRobinsonFisher
) : Extended;
{!
//==================================================
uses
- SysUtils, Math
-{$IFDEF BENCHMARK}
- , Windows, Dialogs
-{$ENDIF}
- ;
-
-{$IFDEF BENCHMARK}
-var
- b1, b2, b3, b4, b5, b6, b7, b8, b9, b10 : Int64; // benchmark
-{$ENDIF}
+ SysUtils, Math, Windows,
+ MojuUtils;
const
GIKO_BAYESIAN_FILE_VERSION = '1.0';
- kYofKanji : TSysCharSet = [#$80..#$A0, #$E0..#$ff];
+{
+ Modes = (ModeWhite, ModeGraph, ModeAlpha, ModeHanKana, ModeNum,
+ ModeWGraph, ModeWAlpha, ModeWNum,
+ ModeWHira, ModeWKata, ModeWKanji);
+}
+ CharMode1 : array [ 0..255 ] of Byte =
+ (
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1,
+ 1, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 1, 1, 1, 1, 1,
+ 1, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 1, 1, 1, 1, 0,
+
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 1, 1, 1, 1, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+ 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+ 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+ 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+ );
//************************************************************
// misc
//************************************************************
// TWordCount class
//************************************************************
+constructor TWordCount.Create;
+begin
+
+ Duplicates := dupIgnore;
+ CaseSensitive := True;
+ Sorted := True;
+
+end;
+
destructor TWordCount.Destroy;
var
i : Integer;
constructor TGikoBayesian.Create;
begin
-{$IFDEF BENCHMARK}
- b1:=0; b2:=0; b3:=0; b4:=0; b5:=0; b6:=0; b7:=0; b8:=0; b9:=0; b10:=0;
-{$ENDIF}
-
- Duplicates := dupIgnore;
- Sorted := True;
+ Duplicates := dupIgnore;
+ CaseSensitive := True;
+ Sorted := True;
end;
for i := 1 to sl.Count - 1 do begin
s := sl[ i ];
- name := RemoveToken( s, #1 );
+ name := GikoBayesian.RemoveToken( s, #1 );
info := TWordInfo.Create;
- info.NormalWord := StrToIntDef( '$' + RemoveToken( s, #1 ), 0 );
- info.ImportantWord := StrToIntDef( '$' + RemoveToken( s, #1 ), 0 );
- info.NormalText := StrToIntDef( '$' + RemoveToken( s, #1 ), 0 );
- info.ImportantText := StrToIntDef( '$' + RemoveToken( s, #1 ), 0 );
+ info.NormalWord := StrToIntDef( '$' + GikoBayesian.RemoveToken( s, #1 ), 0 );
+ info.ImportantWord := StrToIntDef( '$' + GikoBayesian.RemoveToken( s, #1 ), 0 );
+ info.NormalText := StrToIntDef( '$' + GikoBayesian.RemoveToken( s, #1 ), 0 );
+ info.ImportantText := StrToIntDef( '$' + GikoBayesian.RemoveToken( s, #1 ), 0 );
AddObject( name, info );
end;
info : TWordInfo;
begin
-{$IFDEF BENCHMARK}
- ShowMessage(IntToStr(b1)+'/'+IntToStr(b2)+'/'+IntToStr(b3)+'/'+IntToStr(b4)+
- '/'+IntToStr(b5)+'/'+IntToStr(b6));
-{$ENDIF}
-
FFilePath := filePath;
sl := TStringList.Create;
idx : Integer;
begin
- idx := IndexOf( name );
- if idx < 0 then
- Result := nil
+ if Find( name, idx ) then
+ Result := TWordInfo( inherited Objects[ idx ] )
else
- Result := TWordInfo( inherited Objects[ idx ] );
+ Result := nil;
end;
idx : Integer;
begin
- idx := IndexOf( name );
- if idx < 0 then
- AddObject( name, value )
+ if Find( name, idx ) then
+ inherited Objects[ idx ] := value
else
- inherited Objects[ idx ] := value;
+ AddObject( name, value );
end;
ModeWGraph, ModeWAlpha, ModeWNum,
ModeWHira, ModeWKata, ModeWKanji);
var
- p, tail, last : PChar;
- mode, newMode : Modes;
- aWord : string;
- ch : Longword;
- chSize : Integer;
- delimiter : TStringList;
- delimited : Boolean;
- i, idx : Integer;
- countInfo : TWordCountInfo;
-{$IFDEF BENCHMARK}
- t1, t2 : Int64;
+ p, tail, last : PChar;
+ mode, newMode : Modes;
+ ch : Longword;
+ chSize : Integer;
+ wHiraDelimiter : TStringList;
+ wHiraFinalDelimiter : TStringList;
+ wKanjiDelimiter : TStringList;
+ words : TStringList;
+ aWord : string;
+// countInfo : TWordCountInfo;
+
+ function cutBoth( _aWord : string; _delim : TStringList ) : string;
+ var
+ _i : Integer;
+ begin
+ for _i := 0 to _delim.Count - 1 do begin
+ _aWord := CustomStringReplace(
+ _aWord,
+ _delim[ _i ],
+ #10 + _delim[ _i ] + #10, False );
+ end;
+ Result := _aWord;
+ end;
+
+ function cutFirst( _aWord : string; _delim : TStringList ) : string;
+ var
+ _i : Integer;
+ begin
+ for _i := 0 to _delim.Count - 1 do begin
+ _aWord := CustomStringReplace(
+ _aWord,
+ _delim[ _i ],
+ #10 + _delim[ _i ], False );
+ end;
+ Result := _aWord;
+ end;
+
+ function cutFinal( _aWord : string; _delim : TStringList ) : string;
+ var
+ _i : Integer;
+ begin
+ for _i := 0 to _delim.Count - 1 do begin
+ _aWord := CustomStringReplace(
+ _aWord,
+ _delim[ _i ],
+ _delim[ _i ] + #10, False );
+ end;
+ Result := _aWord;
+ end;
+
+ procedure addWord( _dst : TWordCount; _words : TStringList );
+ var
+ _aWord : string;
+ _i, _idx : Integer;
+ _countInfo : TWordCountInfo;
+ begin
+ for _i := 0 to _words.Count - 1 do begin
+ _aWord := _words[ _i ];
+ if Length( _aWord ) > 0 then begin
+ if _dst.Find( _aWord, _idx ) then begin
+ _countInfo := TWordCountInfo( _dst.Objects[ _idx ] );
+ end else begin
+ _countInfo := TWordCountInfo.Create;
+ _dst.AddObject( _aWord, _countInfo );
+ end;
+ _countInfo.WordCount := _countInfo.WordCount + 1;
+ end;
+ end;
+ end;
+
+ function changeMode( _aWord : string; _mode : Modes ) : string;
+ var
+ _i : Integer;
+ _aWord2 : string;
+ _pWord, _pWord2 : PChar;
+ _pWordTail, _pFound : PChar;
+ const
+ _delim : string = #10;
+ begin
+{$IFDEF GIKO_BAYESIAN_NO_HIRAGANA_DIC}
+ if mode = ModeWHira then begin
+ Result := '';
+ Exit;
+ end;
{$ENDIF}
+ if Ord( _mode ) >= Ord( ModeWGraph ) then begin
+ // \93ú\96{\8cê
+ // \83X\83y\81[\83X\82ð\8bl\82ß\82é
+ _aWord := CustomStringReplace( _aWord, ' ', '', False );
+ _aWord := CustomStringReplace( _aWord, '\81@', '', False );
+
+ // \83f\83\8a\83~\83^\82Å\92P\8cê\95ª\82¯
+ case mode of
+ ModeWHira:
+ begin
+ _aWord := cutFinal( _aWord, wHiraFinalDelimiter );
+ Result := cutBoth( _aWord, wHiraDelimiter );
+ end;
+
+ ModeWKanji:
+ begin
+ // \83f\83\8a\83~\83^\82Å\92P\8cê\95ª\82¯
+ _aWord := cutBoth( _aWord, wKanjiDelimiter );
+ // 4 byte (2 \8e\9a) \82¸\82Â\82Å\92P\8cê\95ª\82¯
+ _pWord := PChar( _aWord );
+ _i := Length( _aWord );
+ _pWordTail := _pWord + _i;
+ SetLength( _aWord2, _i + (_i shr 2) );
+ _pWord2 := PChar( _aWord2 );
+
+ while _pWord < _pWordTail do begin
+ _pFound := AnsiStrPos( _pWord, PChar( _delim ) );
+ if _pFound = nil then
+ _pFound := _pWordTail;
+ _pFound := _pFound - 3;
+
+ while _pWord <= _pFound do begin
+ CopyMemory( _pWord2, _pWord, 4 ); _pWord2[ 4 ] := #10;
+ _pWord2 := _pWord2 + 5; _pWord := _pWord + 4;
+ end;
+ _i := _pFound + 4 - _pWord; // 4 = 3 + #10
+ CopyMemory( _pWord2, _pWord, _i );
+ _pWord2 := _pWord2 + _i; _pWord := _pWord + _i;
+ end;
+ if _pWord < _pWordTail then begin
+ _i := _pWordTail - _pWord;
+ CopyMemory( _pWord2, _pWord, _i );
+ _pWord2 := _pWord2 + _i;
+ end;
+ SetLength( _aWord2, _pWord2 - PChar( _aWord2 ) );
+
+ Result := _aWord2;
+ end;
+
+ else
+ Result := _aWord;
+ end;
+ end else begin
+ Result := _aWord;
+ end;
+ end;
const
- KAKUJOSI = '\82ð' + #10 + '\82É' + #10 + '\82ª' + #10 + '\82Æ' + #10 + '\82©\82ç' +
- #10 + '\82Å' + #10 + '\82Ö' + #10 + '\82æ\82è' + #10 + '\82Ü\82Å';
+ WHIRA_DELIMITER = '\82ð' + #10 + '\82É' + #10 + '\82ª' + #10 + '\82Æ' + #10 + '\82©\82ç'
+ + #10 + '\82Ö' + #10 + '\82æ\82è' + #10 + '\82Ü\82Å'+ #10 + '\82Å'
+ + #10 + '\82±\82±' + #10 + '\82»\82±' + #10 + '\82Ç\82±'
+ + #10 + '\82±\82ê' + #10 + '\82»\82ê' + #10 + '\82 \82ê' + #10 + '\82Ç\82ê'
+ + #10 + '\82±\82Ì' + #10 + '\82»\82Ì' + #10 + '\82 \82Ì' + #10 + '\82Ç\82Ì'
+ + #10 + '\82±\82¤' + #10 + '\82»\82¤' + #10 + '\82 \82 ' + #10 + '\82Ç\82¤'
+ + #10 + '\82±\82ñ\82È' + #10 + '\82»\82ñ\82È' + #10 + '\82 \82ñ\82È' + #10 + '\82Ç\82ñ\82È'
+ + #10 + '\82ê\82½' + #10 + '\82ê\82Ä' + #10 + '\82ê\82ê' + #10 + '\82ê\82ë'
+ + #10 + '\82ê\82é' + #10 + '\82ç\82ê\82é'
+ + #10 + '\82Å\82·' + #10 + '\82Ü\82·' + #10 + '\82Ü\82¹\82ñ'
+ + #10 + '\82Å\82µ\82½' + #10 + '\82Ü\82µ\82½'
+ + #10 + '\82·\82é' + #10 + '\82µ\82È\82¢' + #10 + '\82³\82ê\82é' + #10 + '\82³\82ê\82È\82¢'
+ ;
+ WKANJI_DELIMITER = '\93I' + #10 + '\90«' + #10 + '\8e®' + #10 + '\89»' + #10 + '\96@'
+ + #10 + '\95s' + #10 + '\96³' + #10 + '\94ñ' + #10 + '\94½'
+ ;
+ WHIRA_FINAL_DELIMITER = '\82Á\82½' + #10 + '\82Á\82Ä'
+ ;{
+ + #10 + '\82æ\82Á\82Ä' + #10 + '\82µ\82½\82ª\82Á\82Ä' + #10 + '\82È\82Ì\82Å'
+ + #10 + '\82¾\82©\82ç' + #10 + '\82Å\82·\82©\82ç'
+ + #10 + '\82Ü\82½'
+ + #10 + '\82µ\82©\82µ' + #10 + '\82¾\82ª' + #10 + '\82¯\82Ç' + #10 + '\82¯\82ê\82Ç'
+ + #10 + '\82â\82Í\82è' + #10 + '\82â\82Á\82Ï\82è'
+ + #10 + '\82Å\82µ' + #10 + '\82¾\82ë'
+ + #10 + '\82·\82é' + #10 + '\82µ\82È\82¢' + #10 + '\82µ\82½' + #10 + '\82µ\82È\82¢'
+ ;}
+ // '\81[' \82ð '\82\9f\82¡\82£\82¥\82§' \82É\81B
+ HA_LINE = '\82 \82©\82³\82½\82È\82Í\82Ü\82â\82ç\82í\82ª\82´\82¾\82Î\82Ï\82\9f\82ì';
+ HI_LINE = '\82¢\82«\82µ\82¿\82É\82Ð\82Ý\82è\82î\82¬\82¶\82Ñ\82Ò\82¡';
+ HU_LINE = '\82¤\82\82·\82Â\82Ê\82Ó\82Þ\82ä\82é\82®\82Ô\82Õ\82£';
+ HE_LINE = '\82¦\82¯\82¹\82Ä\82Ë\82Ö\82ß\82ê\82ï\82°\82×\82Ø\82¥';
+ HO_LINE = '\82¨\82±\82»\82Æ\82Ì\82Ù\82à\82æ\82ë\82ð\82²\82Ú\82Û\82§';
+ KA_LINE = '\83A\83J\83T\83^\83i\83n\83}\83\84\83\89\83\8f\83K\83U\83_\83o\83p\83@\83\95\83\8e';
+ KI_LINE = '\83C\83L\83V\83`\83j\83q\83~\83\8a\83\90\83M\83W\83r\83s\83B';
+ KU_LINE = '\83E\83N\83X\83c\83k\83t\83\80\83\86\83\8b\83O\83u\83v\83D\83\94';
+ KE_LINE = '\83G\83P\83Z\83e\83l\83w\83\81\83\8c\83\91\83Q\83x\83y\83F\83\96';
+ KO_LINE = '\83I\83R\83\\83g\83m\83z\83\82\83\88\83\8d\83\92\83S\83{\83|\83H';
+ kKanji = [$80..$A0, $E0..$ff];
begin
- delimiter := TStringList.Create;
+ wHiraDelimiter := TStringList.Create;
+ wHiraFinalDelimiter := TStringList.Create;
+ wKanjiDelimiter := TStringList.Create;
+ words := TStringList.Create;
try
- //*** \91¬\93x\83e\83X\83g\92\86
- wordCount.Duplicates := dupIgnore;
- wordCount.CaseSensitive := True;
- wordCount.Capacity := 1000;
- wordCount.Sorted := True;
- //***
-
mode := ModeWhite;
- delimiter.Text := KAKUJOSI;
- SetLength( aWord, 256 );
+{$IFNDEF GIKO_BAYESIAN_NO_HIRAGANA_DIC}
+ wHiraDelimiter.Text := WHIRA_DELIMITER;
+ wHiraFinalDelimiter.Text := WHIRA_FINAL_DELIMITER;
+{$ENDIF}
+ wKanjiDelimiter.Text := WKANJI_DELIMITER;
p := PChar( text );
tail := p + Length( text );
last := p;
while p < tail do begin
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t1 );
-{$ENDIF}
- delimited := False;
// \95¶\8e\9a\82Ì\83^\83C\83v\82ð\94»\95Ê
// \81¦\8bå\93Ç\93_\82Í ModeGraph \82É\82È\82é\82Ì\82Å\8cÂ\95Ê\82É\91Î\89\9e\82µ\82È\82\82Ä\82à\82¢\82¢
- if p^ in kYofKanji then begin
+// if Byte(Byte( p^ ) - $a1) < $5e then begin
+ if Byte( p^ ) in kKanji then begin
if p + 1 < tail then begin
ch := (PByte( p )^ shl 8) or PByte( p + 1 )^;
case ch of
- $8140: newMode := ModeWhite;
+ // \83X\83y\81[\83X\82Å\92P\8cê\95ª\82¯\82¹\82¸\82É\8bl\82ß\82é
+ //$8140: newMode := ModeWhite;
$8141..$824e: newMode := ModeWGraph;
$824f..$8258: newMode := ModeWNum;
$8260..$829a: newMode := ModeWAlpha;
$8340..$8396: newMode := ModeWKata;
else newMode := ModeWKanji;
end;
+ // '\81J\81K\81[' \82Í\95½\89¼\96¼\81A\82Ü\82½\82Í\83J\83^\83J\83i\82É\8aÜ\82Ü\82ê\82é
+ if (mode = ModeWHira) or (mode = ModeWKata) then
+ if (ch = $814a) or (ch = $814b) or (ch = $815b) then
+ newMode := mode;
end else begin
newMode := ModeWhite;
end;
chSize := 2;
-
- // \8bæ\90Ø\82è\82É\82È\82é\95¶\8e\9a\82ª\82 \82é\82©\8c\9f\8d¸\82·\82é
- if p + 3 < tail then begin // 3 = delimiter \82Ì\8dÅ\91å\8e\9a\90\94 - 1
- for i := 0 to delimiter.Count - 1 do begin
- if CompareMem(
- p, PChar( delimiter[ i ] ), Length( delimiter[ i ] ) ) then begin
- delimited := True;
- chSize := Length( delimiter[ i ] );
- Break;
- end;
- end;
- end;
end else begin
- case p^ of
- #$0..#$20, #$7f: newMode := ModeWhite;
- '0'..'9': newMode := ModeNum;
- 'a'..'z', 'A'..'Z': newMode := ModeAlpha;
- #$A6..#$DD: newMode := ModeHanKana;
- else newMode := ModeGraph;
+ newMode := Modes( CharMode1[ Byte( p^ ) ] );
+ if (p^ = ' ') and (Ord( mode ) >= Ord( ModeWGraph )) then begin
+ // \8d¡\82Ü\82Å\93ú\96{\8cê\82Å\8d¡\83X\83y\81[\83X
+ // \92P\8cê\82ð\8cq\82°\82Ä\8cã\82Å\83X\83y\81[\83X\82ð\8bl\82ß\82é
+ // \81¦\94¼\8ap\83J\83i\82Í\92Ê\8fí\83X\83y\81[\83X\82Å\8bæ\90Ø\82é\82¾\82ë\82¤\82©\82ç\8bl\82ß\82È\82¢
+ newMode := mode;
end;
chSize := 1;
end;
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t2 ); b1 := b1 + (t2 - t1);
-{$ENDIF}
- if (mode <> newMode) or delimited then begin
+ if mode <> newMode then begin
// \95¶\8e\9a\82Ì\83^\83C\83v\82ª\95Ï\8dX\82³\82ê\82½
- // \82à\82µ\82\82Í\8bæ\90Ø\82è\82É\82È\82é\95¶\8e\9a\82É\91\98\8bö\82µ\82½
if mode <> ModeWhite then begin
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t1 );
-{$ENDIF}
- aWord := Copy( last, 0, p - last ); // \8c\83\92x
-// SetLength( aWord, p - last );
-// CopyMemory( PChar( aWord ), last, p - last );
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t2 ); b2 := b2 + (t2 - t1);
-{$ENDIF}
- idx := wordCount.IndexOf( aWord ); // \8c\83\92x
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t1 ); b3 := b3 + (t1 - t2);
-{$ENDIF}
- if idx < 0 then begin
- countInfo := TWordCountInfo.Create;
- wordCount.AddObject( aWord, countInfo );
- end else begin
- countInfo := TWordCountInfo( wordCount.Objects[ idx ] );
- end;
- countInfo.WordCount := countInfo.WordCount + 1;
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t2 ); b4 := b4 + (t2 - t1);
-{$ENDIF}
+ SetLength( aWord, p - last );
+ CopyMemory( PChar( aWord ), last, p - last );
+
+ words.Text := changeMode( aWord, mode );
+
+ // \92P\8cê\93o\98^
+ addWord( wordCount, words );
end;
last := p;
end; // while
if mode <> ModeWhite then begin
- aWord := Copy( last, 0, p - last );
- idx := wordCount.IndexOf( aWord );
- if idx < 0 then begin
- countInfo := TWordCountInfo.Create;
- wordCount.AddObject( aWord, countInfo );
- end else begin
- countInfo := TWordCountInfo( wordCount.Objects[ idx ] );
- end;
- countInfo.WordCount := countInfo.WordCount + 1;
+ SetLength( aWord, p - last );
+ CopyMemory( PChar( aWord ), last, p - last );
+
+ words.Text := changeMode( aWord, mode );
+
+ // \92P\8cê\93o\98^
+ addWord( wordCount, words );
end;
finally
- delimiter.Free;
+ words.Free;
+ wKanjiDelimiter.Free;
+ wHiraFinalDelimiter.Free;
+ wHiraDelimiter.Free;
end;
end;
begin
info := Objects[ aWord ];
if info = nil then
- Result := 0.4
+ Result := 0.415
else if info.NormalWord = 0 then
Result := 0.99
else if info.ImportantWord = 0 then
Result := 0.01
- else
- Result := ( info.ImportantWord / info.ImportantText ) /
- ((info.NormalWord * 2 / info.NormalText ) +
- (info.ImportantWord / info.ImportantText));
+ else if info.ImportantWord + info.NormalWord * 2 < 5 then
+ Result := 0.5
+ else begin
+ try
+ Result := ( info.ImportantWord / info.ImportantText ) /
+ ((info.NormalWord * 2 / info.NormalText ) +
+ (info.ImportantWord / info.ImportantText));
+ except
+ on EZeroDivide do Result := 0.99;
+ end;
+ end;
end;
var
i := min( SAMPLE_COUNT, narray.Count );
while i > 0 do begin
Dec( i );
+
s := s * Single( narray[ i ] );
q := q * (1 - Single( narray[ i ] ));
end;
-
- Result := s / (s + q);
+ try
+ Result := s / (s + q);
+ except
+ Result := 0.5;
+ end;
finally
narray.Free;
end;
if info = nil then
Result := 0.415
else if info.ImportantWord = 0 then
- Result := 0.0001
+ Result := 0.01
else if info.NormalWord = 0 then
- Result := 0.9999
+ Result := 0.99
else
+ {
Result := ( info.ImportantWord / info.ImportantText ) /
((info.NormalWord / info.NormalText ) +
(info.ImportantWord / info.ImportantText));
+ }
+ try
+ Result := (info.ImportantWord * info.NormalText) /
+ (info.NormalWord * info.ImportantText +
+ info.ImportantWord * info.NormalText);
+ except
+ Result := 0.5;
+ end;
end;
function f( cnt : Integer; n, mean : Single ) : Extended;
const
- k = 0.00001;
+ k = 0.001;
begin
Result := ( (k * mean) + (cnt * n) ) / (k + cnt);
end;
mean : Extended;
countInfo : TWordCountInfo;
i : Integer;
- normal : Extended;
- important : Extended;
+ P1, Q1{, R1} : Extended;
cnt : Extended;
begin
end;
mean := mean / wordCount.Count;
- cnt := 0;
- normal := 1;
- important := 1;
+ P1 := 1;
+ Q1 := 1;
for i := 0 to wordCount.Count - 1 do begin
countInfo := TWordCountInfo( wordCount.Objects[ i ] );
n := f( countInfo.WordCount, narray[ i ], mean );
- normal := normal * n;
- important := important * (1 - n);
- if countInfo <> nil then
- cnt := cnt + countInfo.WordCount;
+ P1 := P1 * ( 1 - n );
+ Q1 := Q1 * n;
end;
+ cnt := wordCount.Count;
if cnt = 0 then
cnt := 1;
- normal := 1 - Exp( Ln( normal ) * (1 / cnt) );
- important := 1 - Exp( Ln( important ) * (1 / cnt) );
+ try
+ P1 := 1 - Power( P1, 1 / cnt );
+ except
+ end;
+ try
+ Q1 := 1 - Power( Q1, 1 / cnt );
+ except
+ end;
- n := (important - normal+ 0.00001) / (important + normal + 0.00001);
- Result := (1 + n) / 2;
+ if P1 + Q1 = 0 then begin
+ Result := 0.5
+ end else begin
+ n := (P1 - Q1) / (P1 + Q1);
+ Result := (1 + n) / 2;
+ end;
+
+end;
+
+//==============================
+// CalcGaryRobinsonFisher
+//==============================
+function TGikoBayesian.CalcGaryRobinsonFisher(
+ wordCount : TWordCount
+) : Extended;
+
+ function p( const aWord : string ) : Single;
+ var
+ info : TWordInfo;
+ begin
+ info := Objects[ aWord ];
+ if info = nil then
+ Result := 0.415
+ else if info.ImportantWord = 0 then
+ Result := 0.01
+ else if info.NormalWord = 0 then
+ Result := 0.99
+ else
+ {
+ Result := ( info.ImportantWord / info.ImportantText ) /
+ ((info.NormalWord / info.NormalText ) +
+ (info.ImportantWord / info.ImportantText));
+ }
+ Result := (info.ImportantWord * info.NormalText) /
+ (info.NormalWord * info.ImportantText +
+ info.ImportantWord * info.NormalText);
+ end;
+
+ function f( cnt : Integer; n, mean : Single ) : Extended;
+ const
+ k = 0.001;
+ begin
+ Result := ( (k * mean) + (cnt * n) ) / (k + cnt);
+ end;
+
+ function prbx( x2, degree : Extended ) : Extended;
+ begin
+
+ Result := 0.5;
+
+ end;
+
+var
+ n : Extended;
+ narray : array of Single;
+ mean : Extended;
+ countInfo : TWordCountInfo;
+ i : Integer;
+// normal : Extended;
+// important : Extended;
+ P1, Q1 : Extended;
+ cnt : Extended;
+begin
+
+ if wordCount.Count = 0 then begin
+ Result := 1;
+ Exit;
+ end;
+
+ SetLength( narray, wordCount.Count );
+ mean := 0;
+ for i := 0 to wordCount.Count - 1 do begin
+ n := p( wordCount[ i ] );
+ narray[ i ] := n;
+ mean := mean + n;
+ end;
+ mean := mean / wordCount.Count;
+
+ P1 := 1;
+ Q1 := 1;
+ for i := 0 to wordCount.Count - 1 do begin
+ countInfo := TWordCountInfo( wordCount.Objects[ i ] );
+ n := f( countInfo.WordCount, narray[ i ], mean );
+ P1 := P1 * ( 1 - n );
+ Q1 := Q1 * n;
+ end;
+ cnt := wordCount.Count;
+ if cnt = 0 then
+ cnt := 1;
+ try
+ P1 := Power( P1, 1 / cnt );
+ except
+ end;
+ try
+ Q1 := Power( Q1, 1 / cnt );
+ except
+ end;
+
+ P1 := 1 - prbx( -2 * Ln( P1 ), 2 * cnt );
+ Q1 := 1 - prbx( -2 * Ln( Q1 ), 2 * cnt );
+
+ Result := (1 + P1 - Q1) / 2;
end;
function TGikoBayesian.Parse(
const text : string;
wordCount : TWordCount;
- algorithm : TGikoBayesianAlgorithm = gbaGaryRonbinson
+ algorithm : TGikoBayesianAlgorithm
) : Extended;
-{$IFDEF BENCHMARK}
-var
- t1, t2 : Int64;
-{$ENDIF}
begin
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t1 );
-{$ENDIF}
CountWord( text, wordCount );
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t2 ); b5 := b5 + (t2 - t1);
-{$ENDIF}
case algorithm of
gbaPaulGraham: Result := CalcPaulGraham( wordCount );
- gbaGaryRonbinson: Result := CalcGaryRobinson( wordCount );
+ gbaGaryRobinson: Result := CalcGaryRobinson( wordCount );
+ gbaGaryRobinsonFisher:
+ Result := CalcGaryRobinsonFisher( wordCount );
else Result := 0;
end;
-{$IFDEF BENCHMARK}
- QueryPerformanceCounter( t1 ); b6 := b6 + (t1 - t2);
-{$ENDIF}
end;
aWord : string;
wordinfo : TWordInfo;
countinfo : TWordCountInfo;
- i : Integer;
+ i : Integer;
begin
for i := 0 to wordCount.Count - 1 do begin
aWord := wordCount[ i ];
wordinfo := Objects[ aWord ];
+ countinfo := TWordCountInfo( wordCount.Objects[ i ] );
if wordinfo = nil then begin
wordinfo := TWordInfo.Create;
Objects[ aWord ] := wordinfo;
end;
- countinfo := TWordCountInfo( wordCount.Objects[ i ] );
if isImportant then begin
wordinfo.ImportantWord := wordinfo.ImportantWord + countinfo.WordCount;
wordinfo.ImportantText := wordinfo.ImportantText + 1;
countinfo := TWordCountInfo( wordCount.Objects[ i ] );
if isImportant then begin
- wordinfo.ImportantWord := wordinfo.ImportantWord - countinfo.WordCount;
- wordinfo.ImportantText := wordinfo.ImportantText - 1;
+ if wordInfo.ImportantText > 0 then begin
+ wordinfo.ImportantText := wordinfo.ImportantText - 1;
+ wordinfo.ImportantWord := wordinfo.ImportantWord - countinfo.WordCount;
+ end;
end else begin
- wordinfo.NormalWord := wordinfo.NormalWord - countinfo.WordCount;
- wordinfo.NormalText := wordinfo.NormalText - 1;
+ if wordinfo.NormalText > 0 then begin
+ wordinfo.NormalText := wordinfo.NormalText - 1;
+ wordinfo.NormalWord := wordinfo.NormalWord - countinfo.WordCount;
+ end;
end;
end;
OSDN Git Service
