Minor updates

scribeocr · Sep 9, 2024 · 5846586 · 5846586
1 parent 4b2b43e
commit 5846586
Show file tree

Hide file tree

Showing 4 changed files with 10 additions and 19 deletions.
diff --git a/js/fontSupp.js b/js/fontSupp.js
@@ -137,7 +137,7 @@ export const calcSuppFontInfo = async (ocrArr) => {
     }
     if (fontSizeMultArr.length >= 3) {
       const fontSizeMult = quantile(fontSizeMultArr, 0.5);
-      FontProps.sizeMult[key] = fontSizeMult;
+      if (fontSizeMult && fontSizeMult > 0.9 && fontSizeMult < 1.5) FontProps.sizeMult[key] = fontSizeMult;
     }
 
     if (serifVotes > sansVotes) {

diff --git a/js/import/convertPageAbbyy.js b/js/import/convertPageAbbyy.js
@@ -3,7 +3,6 @@ import ocr from '../objects/ocrObjects.js';
 import {
   ascCharArr,
   calcBboxUnion,
-  determineSansSerif,
   mean50,
   quantile,
   round6, unescapeXml,
@@ -72,8 +71,6 @@ export async function convertPageAbbyy({ ocrStr, n }) {
       const xmlLineFormatting = xmlLinePreChar?.match(/<formatting[^>]+/)?.[0];
       const fontName = xmlLineFormatting?.match(/ff=['"]([^'"]*)/)?.[1];
 
-      const fontFamily = determineSansSerif(fontName);
-
       let dropCap = false;
       const dropCapMatch = xmlLine.match(abbyyDropCapRegex);
       if (dropCapMatch != null && parseInt(dropCapMatch[1]) > 0) {
@@ -315,9 +312,7 @@ export async function convertPageAbbyy({ ocrStr, n }) {
 
         wordObj.smallCaps = smallCapsArr[i];
 
-        if (fontFamily !== 'Default') {
-          wordObj.font = fontFamily;
-        }
+        if (fontName) wordObj.font = fontName;
 
         if (styleArr[i] === 'sup') {
           wordObj.sup = true;

diff --git a/js/import/convertPageHocr.js b/js/import/convertPageHocr.js
@@ -1,7 +1,7 @@
 import ocr from '../objects/ocrObjects.js';
 
 import {
-  determineSansSerif, getTextScript,
+  getTextScript,
   unescapeXml,
 } from '../utils/miscUtils.js';
 
@@ -166,9 +166,8 @@ export async function convertPageHocr({
 
       const wordLangRaw = match.match(/lang=['"]([^'"]*)['"]/i)?.[1];
 
-      const fontName = match.match(/^[^>]+?x_font\s*([\w-]+)/)?.[1];
-
-      const fontFamily = determineSansSerif(fontName);
+      let fontName = match.match(/^[^>]+?x_font\s*([^'";]+)/)?.[1];
+      if (fontName) fontName = fontName.trim();
 
       const it = match.matchAll(charRegex);
       const letterArr = [...it];
@@ -246,7 +245,7 @@ export async function convertPageHocr({
       if (debugMode) wordObj.raw = match;
 
       if (italic) wordObj.style = 'italic';
-      if (fontFamily !== 'Default') wordObj.font = fontFamily;
+      if (fontName) wordObj.font = fontName;
 
       wordObj.conf = wordConf;
 
@@ -297,9 +296,8 @@ export async function convertPageHocr({
         bottom: wordBox1[3],
       };
 
-      const fontName = match.match(/^[^>]+?x_font\s*([\w-]+)/)?.[1];
-
-      const fontFamily = determineSansSerif(fontName);
+      let fontName = match.match(/^[^>]+?x_font\s*([^'";]+)/)?.[1];
+      if (fontName) fontName = fontName.trim();
 
       const styleStr = match.match(/style=['"]([^'"]+)/)?.[1];
 
@@ -333,9 +331,7 @@ export async function convertPageHocr({
       }
 
       wordObj.style = fontStyle;
-      if (fontFamily !== 'Default') {
-        wordObj.font = fontFamily;
-      }
+      if (fontName) wordObj.font = fontName;
 
       wordObj.sup = wordSup;
 

diff --git a/js/import/import.js b/js/import/import.js
@@ -449,7 +449,7 @@ export async function importFiles(files, options = {}) {
     await extractInternalPDFText({
       setActive: true, extractPDFTextNative, extractPDFTextOCR, extractPDFTextImage,
     });
-    if (opt.calcSuppFontInfo) calcSuppFontInfo(ocrAll.pdf);
+    if (opt.calcSuppFontInfo) await calcSuppFontInfo(ocrAll.pdf);
   }
 }