Changeset 37017

Timestamp:

2022-12-10T22:04:37+13:00 (17 months ago)

Author:

davidb

Message:

OCR'd txt now stored in each <div> as a data field

Location:

gs3-installations/intermuse/trunk/sites/intermuse/collect/programmes

Files:

• js/document_viewer.js (modified) (6 diffs)
• transform/pages/document.xsl (modified) (2 diffs)

gs3-installations/intermuse/trunk/sites/intermuse/collect/programmes/js/document_viewer.js

@@ -19 +19 @@
 }
 
+function gv_ocr_paras_to_text(paras)
+{
+    var ocr_text = "";
+
+    var num_paras = paras.length;
+    for (var p=0; p<num_paras; p++) {
+
+    var para = paras[p];    
+    var words = para.words;
+    var num_words = words.length;
+    
+    for (var w=0; w<num_words; w++) {
+        var word = words[w];
+        
+        var symbols = word.symbols;
+        var num_symbols = symbols.length;
+        
+        for (var s=0; s<num_symbols; s++) {
+        
+        var symbol = symbols[s];
+
+        ocr_text += symbol.text;
+
+        // Enumerated fields listed at:
+        //   http://googleapis.github.io/googleapis/java/grpc-google-cloud-vision-v1/0.1.5/apidocs/com/google/cloud/vision/v1/TextAnnotation.DetectedBreak.BreakType.html
+        if (symbol.property) {
+            var detected_break = symbol.property.detectedBreak;
+            if ((detected_break.type == "SPACE" || detected_break.type == "EOL_SURE_SPACE")) {
+            ocr_text += ' ';
+            }
+            else if (detected_break.type == "SURE_SPACE") { // very wide space
+            ocr_text += '\t';
+            }
+            // "HYPHEN" at end end of line => assume word continues onto next line
+            else if (detected_break.type == "LINE_BREAK" ) {
+            ocr_text += '\n\n';
+            }
+        }
+        }
+    }
+    }
+    
+    return ocr_text;
+}
+
 function display_gv_ocr_bounding_boxes(gv_ocr_json,doc_sec_id)
 {
-    console.log(gv_ocr_json);
+    //console.log(gv_ocr_json);
 
     
@@ -38 +83 @@
     var num_blocks = blocks.length;
 
-    for (b=0; b<num_blocks; b++) {
+    for (var b=0; b<num_blocks; b++) {
         var block = blocks[b];
 
@@ -51 +96 @@
         var num_vertices = vertices.length;
 
-        for (v=0; v<num_vertices; v++) {
+        for (var v=0; v<num_vertices; v++) {
         var x = vertices[v].x;
         var y = vertices[v].y;
@@ -67 +112 @@
 
         var rect = { "x_org": x_org, "y_org": y_org, "x_dim": x_dim, "y_dim": y_dim};
-        bounding_box_rects.push(rect);
+
+        var paras = block.paragraphs;
+        var ocr_text = gv_ocr_paras_to_text(paras);
+        rect.text = ocr_text;
+
+        bounding_box_rects.push(rect);      
     }
-
+    
         
     display_scaled_div_bounding_boxes(doc_sec_id, bounding_box_rects, page_x_dim,page_y_dim);
@@ -100 +150 @@
 
     var $screen_div = $('#'+screen_image_id);
-    //var $screen_img = $screen_div.find("img");
     
     $screen_div.append($boundingbox_overlay);
-    //$screen_img.append($boundingbox_overlay);
-
     
     var num_bb_rects = bounding_box_rects.length;
 
-    console.log("Block bounding boxes:")
-
-    for (r=0; r<num_bb_rects; r++) {
+    //console.log("Block bounding boxes:")
+
+    for (var r=0; r<num_bb_rects; r++) {
     var rect = bounding_box_rects[r];
-        console.log("  " + JSON.stringify(rect));
+        //console.log("  " + JSON.stringify(rect));
 
     var scaled_x_org = rect.x_org * scale_x;
@@ -125 +172 @@
         .css("width", scaled_x_dim)
         .css("height",scaled_y_dim);
-    
-    $boundingbox_overlay.append($boundingbox_div)
+
+    var text = rect.text;
+    // //console.log("text = " + text);
+    // // Could use 'data' here, but then does not appear in the element itself
+    // // => for now do this 'old school' with an explicit 'data-text' attribute
+    // $boundingbox_div.attr("data-text",text);
+
+    $boundingbox_div.attr("text",text);
+    
+    $boundingbox_overlay.append($boundingbox_div)   
     }
 }

gs3-installations/intermuse/trunk/sites/intermuse/collect/programmes/transform/pages/document.xsl

@@ -33 +33 @@
     <xsl:param name="assocfilepath"/>
     
+    <gsf:variable name="GVDocumentOCRJSON"><gsf:metadata name="GVDocumentOCRJSON"/></gsf:variable>
+    <gsf:variable name="docSectionID"><xsl:value-of select="util:replace(@nodeID, '.', '_')}"/></gsf:variable>
+    
     <div style="padding-top: 10px; padding-bottom: 10px;">
       Google Vision OCR JSON:
@@ -38 +41 @@
     <xsl:attribute name="href">
       <xsl:value-of disable-output-escaping="yes" select="/page/pageResponse/collection/metadataList/metadata[@name = 'httpPath']"/>
-      <xsl:text>/index/assoc/</xsl:text><xsl:value-of disable-output-escaping="yes" select="$assocfilepath" /><xsl:text>/</xsl:text>gv_document_ocrgoogle-vision-output.json
+      <xsl:text>/index/assoc/</xsl:text><xsl:value-of disable-output-escaping="yes" select="$assocfilepath" /><xsl:text>/</xsl:text><xsl:value-of select="$GVDocumentOCRJSON"/>
     </xsl:attribute>
-    gv_document_ocrgoogle-vision-output.json
+    <xsl:value-of select="$GVDocumentOCRJSON"/>
       </a>
     </div>
-    <gsf:variable name="GVDocumentOCRJSON"><gsf:metadata name="GVDocumentOCRJSON"/></gsf:variable>
-    <gsf:variable name="docSectionID"><xsl:value-of select="util:replace(@nodeID, '.', '_')}"/></gsf:variable>
     <gsf:script>
       load_gv_dococr_json(gs.variables.GVDocumentOCRJSON,gs.variables.docSectionID,display_gv_ocr_bounding_boxes);