New Word dictionary and new disambiguation algorithm

[ypp-sc-tools.db-test.git] / pctb / ocr.c

diff --git a/pctb/ocr.c b/pctb/ocr.c
index fe264e8110ce8ffce6cbb6be0282c004151dec31..f6104452dc1c1854e6ccd19aac77882caae0aba9 100644 (file)
--- a/pctb/ocr.c
+++ b/pctb/ocr.c
@@ -42,6 +42,7 @@ typedef struct DatabaseNode {
 
 typedef struct {
   OcrReader *rd;
+  OcrCellType ct;
   int w;
   Pixcol *cols;
   int x;
@@ -52,30 +53,21 @@ typedef struct {
   int rx;
 } FindCharResults;
 
+enum { ct_Lower, ct_Upper, ct_Word, ct_Digit };
 static const char *context_names[]= {
-  "Lower",
-  "Upper",
-  "Digit"
+  "Lower",  /*  bit 0, value 001 */
+  "Upper",  /*  bit 1, value 002 */
+  "Word",   /*  bit 2, value 004 */
+  "Digit",  /*  bit 3, value 010 */
 };
 struct OcrCellTypeInfo {
   /* bitmaps of indices into context_names: */
   unsigned initial, nextword, midword;
   int space_spaces;
   const char *name;
+  int (*findchar_select)(const FindCharArgs *fca,
+                        const FindCharResults results[]);
 };
-const struct OcrCellTypeInfo ocr_celltype_number= {
-  4,4,4,
-  .space_spaces= 5,
-  .name= "number"
-};
-const struct OcrCellTypeInfo ocr_celltype_text= {
-  .initial=2, /* Uppercase */
-  .nextword=3, /* Either */
-  .midword=1, /* Lower only */
-  .space_spaces= 4,
-  .name= "text"
-};
-
 
 #define NCONTEXTS (sizeof(context_names)/sizeof(context_names[0]))
 
@@ -340,9 +332,12 @@ static DatabaseNode *findchar(const FindCharArgs *fca, int *match_rx) {
     nmatches++;
   }
   if (nmatches==1) {
-    debugf( " unambiguous");
+    debugf(" unique");
   } else {
-    match=-1;
+    debugf(" ambiguous");
+    match= !fca->ct->findchar_select ? -1 :
+      fca->ct->findchar_select(fca,results);
+    debugf(" resolved %s", match<0 ? "<none>" : context_names[match]);
   }
   if (match<0)
     return 0;
@@ -351,6 +346,33 @@ static DatabaseNode *findchar(const FindCharArgs *fca, int *match_rx) {
   return results[ctxi].match;
 }
 
+static int findchar_select_text(const FindCharArgs *fca,
+                               const FindCharResults results[]) {
+  if (fca->ctxmap != 017) return -1;
+
+  dbassert(! results[ct_Digit].match );
+  if (results[ct_Word].match) return ct_Word;
+  if (results[ct_Lower].rx > results[ct_Upper].rx) return ct_Lower;
+  if (results[ct_Upper].rx > results[ct_Lower].rx) return ct_Upper;
+  return -1;
+}
+
+const struct OcrCellTypeInfo ocr_celltype_number= {
+  010,010,010,
+  .space_spaces= 5,
+  .name= "number",
+  .findchar_select= 0
+};
+const struct OcrCellTypeInfo ocr_celltype_text= {
+  .initial=  012, /* Digit|Upper */
+  .nextword= 017, /* Digit|Upper|Lower|Word */
+  .midword=  014, /* Digit|Lower */
+  .space_spaces= 4,
+  .name= "text",
+  .findchar_select= findchar_select_text
+};
+
+
 const char *ocr_celltype_name(OcrCellType ct) { return ct->name; }
 
 OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
@@ -359,6 +381,7 @@ OcrResultGlyph *ocr(OcrReader *rd, OcrCellType ct, int w, Pixcol cols[]) {
 
   FindCharArgs fca;
   fca.rd= rd;
+  fca.ct= ct;
   fca.w= w;
   fca.cols= cols;
   fca.x= -1;