lab3

MaastrichtU-IDS · Feb 19, 2024 · 2d3fdcd · 2d3fdcd
1 parent 420054d
commit 2d3fdcd
Showing 1 changed file with 56 additions and 0 deletions.
diff --git a/2024/lab3/Lab3 - KG from unstructured data.ipynb b/2024/lab3/Lab3 - KG from unstructured data.ipynb
@@ -1082,6 +1082,62 @@
    "source": [
     "## 3. 🕸️ Generate RDF"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from rdflib import URIRef, Literal, RDFS, Namespace, RDF, XSD, DCTERMS\n",
+    "\n",
+    "score_threshold = 0.8\n",
+    "\n",
+    "BASE_NS = Namespace(\"https://w3id.org/myannotations/\")\n",
+    "TAO = Namespace(\"http://pubannotation.org/ontology/tao.owl#\")\n",
+    "\n",
+    "g = Graph(store=\"Oxigraph\")\n",
+    "g.bind(\"\", BASE_NS)\n",
+    "g.bind(\"tao\", TAO)\n",
+    "g.bind(\"doid\", Namespace(\"http://purl.obolibrary.org/obo/DOID_\"))\n",
+    "\n",
+    "annotated_text_uri = BASE_NS[\"document0\"]\n",
+    "g.add((annotated_text_uri, RDF.type, TAO.text_document))\n",
+    "g.add((annotated_text_uri, RDFS.label, Literal(annotate_text)))\n",
+    "\n",
+    "sentence_uris = {}\n",
+    "for i, ent in enumerate(doc.ents):\n",
+    "    hits = search_disease(ent.text)\n",
+    "    # Only consider the first match\n",
+    "    if len(hits) > 0 and hits[0].score > score_threshold:\n",
+    "        print(f\"{ent.text} > {hits[0].payload['label']} ({hit.payload['uri']}) [{hits[0].score}]\")\n",
+    "        # Add the text annotation to the RDF graph\n",
+    "        annot_uri = BASE_NS[f\"match{i}\"]\n",
+    "        g.add((annot_uri, RDF.type, TAO.text_span))\n",
+    "        g.add((annot_uri, TAO.has_value, Literal(ent.text)))\n",
+    "        g.add((annot_uri, TAO.begins_at, Literal(ent.start_char, datatype=XSD.integer)))\n",
+    "        g.add((annot_uri, TAO.ends_at, Literal(ent.end_char, datatype=XSD.integer)))\n",
+    "        g.add((annot_uri, TAO.parts_of, annotated_text_uri))\n",
+    "\n",
+    "        # Create or retrieve the URI for the sentence\n",
+    "        sentence_text = ent.sent.text\n",
+    "        if sentence_text not in sentence_uris:\n",
+    "            sentence_uri = BASE_NS[f\"sentence{len(sentence_uris)}\"]\n",
+    "            sentence_uris[sentence_text] = sentence_uri\n",
+    "            g.add((sentence_uri, RDF.type, TAO.text_span))\n",
+    "            g.add((sentence_uri, TAO.has_value, Literal(sentence_text)))\n",
+    "            g.add((sentence_uri, TAO.parts_of, annotated_text_uri))\n",
+    "        # We use a different predicate to link the annotation to the sentence\n",
+    "        g.add((annot_uri, DCTERMS.isPartOf, sentence_uris[sentence_text]))\n",
+    "\n",
+    "        # Add link to the matching concept\n",
+    "        concept_uri = URIRef(hits[0].payload['uri'])\n",
+    "        g.add((annot_uri, TAO.denotes, concept_uri))\n",
+    "        g.add((concept_uri, RDFS.label, Literal(hits[0].payload['label'])))\n",
+    "\n",
+    "g.serialize(\"diseases_annotations_kg.ttl\", format=\"ttl\")\n",
+    "# print(g.serialize(format=\"ttl\"))"
+   ]
   }
  ],
  "metadata": {