Use MiniOCR format to store fulltext to ocr-highlighting plugin.

This supports only ALTO 2.x. There should be found a solution to support different namespaces. See #488
kitodo · Sep 21, 2021 · 6c92af8 · 6c92af8
1 parent fdad413
commit 6c92af8
Show file tree

Hide file tree

Showing 3 changed files with 64 additions and 4 deletions.
diff --git a/Classes/Common/Document.php b/Classes/Common/Document.php
@@ -702,13 +702,13 @@ class_exists($class)
                 ) {
                     // Load XML from file.
                     $rawTextXml = simplexml_load_string($fileContent);
-                    $rawText = $obj->getRawText($rawTextXml);
-                    $this->rawTextArray[$id] = $rawText;
+                    $textMiniOcr = $obj->getTextAsMiniOcr($rawTextXml);
+                    $this->rawTextArray[$id] = $textMiniOcr;
                 } else {
                     $this->logger->warning('Invalid class/method "' . $class . '->getRawText()" for text format "' . $textFormat . '"');
                 }
             }
-            $fullText = $rawText;
+            $fullText = $textMiniOcr;
         } else {
             $this->logger->warning('Unsupported text format "' . $textFormat . '" in physical node with @ID "' . $id . '"');
         }

diff --git a/Classes/Common/FulltextInterface.php b/Classes/Common/FulltextInterface.php
@@ -21,7 +21,7 @@
  * @access public
  * @abstract
  */
-//TODO: check if this is still needed when actually full text xml is indexed
+
 interface FulltextInterface
 {
     /**
@@ -34,4 +34,15 @@ interface FulltextInterface
      * @return string The raw unformatted fulltext
      */
     public function getRawText(\SimpleXMLElement $xml);
+
+    /**
+     * This extracts the fulltext data from ALTO XML and returns it in MiniOCR format
+     *
+     * @access public
+     *
+     * @param \SimpleXMLElement $xml: The XML to extract the raw text from
+     *
+     * @return string The unformatted fulltext in MiniOCR format
+     */
+    public function getTextAsMiniOcr(\SimpleXMLElement $xml);
 }
diff --git a/Classes/Format/Alto.php b/Classes/Format/Alto.php
@@ -12,9 +12,13 @@
 
 namespace Kitodo\Dlf\Format;
 
+use Kitodo\Dlf\Common\Solr;
+
 /**
  * Fulltext ALTO format class for the 'dlf' extension
  *
+ * ** This currently supports only ALTO 2.x **
+ *
  * @author Sebastian Meyer <sebastian.meyer@slub-dresden.de>
  * @package TYPO3
  * @subpackage dlf
@@ -42,4 +46,49 @@ public function getRawText(\SimpleXMLElement $xml)
         }
         return $rawText;
     }
+
+
+
+    /**
+     * This extracts the fulltext data from ALTO XML and returns it in MiniOCR format
+     *
+     * @access public
+     *
+     * @param \SimpleXMLElement $xml: The XML to extract the raw text from
+     *
+     * @return string The unformatted fulltext in MiniOCR format
+     */
+    public function getTextAsMiniOcr(\SimpleXMLElement $xml)
+    {
+        $rawText = '';
+        $xml->registerXPathNamespace('alto', 'http://www.loc.gov/standards/alto/ns-v2#');
+
+        // get all text blocks
+        $blocks = $xml->xpath('./alto:Layout/alto:Page/alto:PrintSpace//alto:TextBlock');
+
+        if (empty($blocks)) {
+            return '';
+        }
+
+        $miniOcr = new \SimpleXMLElement("<ocr></ocr>");
+
+        foreach ($blocks as $block) {
+            $newBlock = $miniOcr->addChild('b');
+            foreach ($block->children() as $key => $value) {
+                if ($key === "TextLine") {
+                    $newLine = $newBlock->addChild('l');
+                    foreach ($value->children() as $lkey => $word) {
+                        if ($lkey == "String") {
+                            $attributes = $word->attributes();
+                            $la =  Solr::escapeQuery((string) $attributes['CONTENT']);
+                            $newWord = $newLine->addChild('w',  htmlspecialchars(Solr::escapeQuery((string) $attributes['CONTENT'])) . ' ');
+                            $newWord->addAttribute('x', (string) $attributes['HPOS'] . ' ' . (string) $attributes['VPOS'] . ' ' . (string) $attributes['WIDTH'] . ' ' . (string) $attributes['HEIGHT']);
+                        }
+                    }
+                }
+            }
+        }
+
+        return $miniOcr->asXml();
+    }
 }