fix: Updates to hOCR Template to follow hOCR Spec

- Added validation in testing with https://github.com/kba/hocr-spec-python
googleapis · Nov 7, 2023 · 8d32327 · 8d32327
1 parent e05cf50
commit 8d32327
Show file tree

Hide file tree

Showing 4 changed files with 50 additions and 37 deletions.
diff --git a/google/cloud/documentai_toolbox/templates/hocr_document_template.xml.j2 b/google/cloud/documentai_toolbox/templates/hocr_document_template.xml.j2
@@ -6,8 +6,9 @@
 <meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
 <meta name="ocr-system" content="Document AI OCR" />
 <meta name="ocr-langs" content="unknown" />
+<meta name="ocr-scripts" content="unknown" />
 <meta name="ocr-number-of-pages" content="{{ pages|length }}" />
-<meta name="ocr-capabilities" content="ocr_page ocr_carea ocr_par ocr_line ocrx_word" />
+<meta name="ocr-capabilities" content="ocrp_lang ocr_page ocr_carea ocr_par ocr_line ocrx_word" />
 </head>
 <body>
 {% for page in pages -%}
@@ -16,13 +17,13 @@
         {% set bidx = loop.index0 -%}
         <span class='ocr_carea' id='block_{{ page_number }}_{{ bidx }}' title='{{ docai_block.hocr_bounding_box -}}'>{% for paragraph in docai_block.paragraphs -%}
             {% set paridx = loop.index0 -%}
-            <span class='ocr_par' id='par_{{ page_number }}_{{ bidx }}_{{ paridx }}' title='{{ paragraph.hocr_bounding_box -}}'>{% for line in paragraph.lines -%}
+            <p class='ocr_par' id='par_{{ page_number }}_{{ bidx }}_{{ paridx }}' title='{{ paragraph.hocr_bounding_box -}}'>{% for line in paragraph.lines -%}
                 {% set lidx = loop.index0 -%}
                 <span class='ocr_line' id='line_{{ page_number }}_{{ bidx }}_{{ paridx }}_{{ lidx }}' title='{{ line.hocr_bounding_box }}'>{{ line.text }}{% for token in line.tokens -%}
                     {% set tidx = loop.index0 -%}
                     <span class='ocrx_word' id='word_{{ page_number }}_{{ bidx }}_{{ paridx }}_{{ lidx }}_{{ tidx }}' title='{{ token.hocr_bounding_box }}'>{{ token.text }}</span>{% endfor -%}
                 </span>{% endfor -%}
-            </span>{% endfor -%}
+            </p>{% endfor -%}
         </span>{% endfor -%}
     </div>
 {% endfor -%}

diff --git a/testing/constraints-3.7.txt b/testing/constraints-3.7.txt
@@ -14,3 +14,4 @@ google-cloud-documentai==2.20.0
 google-cloud-storage==2.7.0
 numpy==1.19.5
 pikepdf==6.2.9
+hocr-spec==0.2.0