Update annotations for Iraa/paper_98.txt

annotations/Iraa/paper_98.txt.json

@@ -6,5 +6,13 @@
     "label": "Format",
     "user": "Iraa",
     "text": "(Karpukhin et al., 2020;. "
+  },
+  {
+    "file": "paper_98.txt",
+    "start": 1168,
+    "end": 1676,
+    "label": "Coherence",
+    "user": "Iraa",
+    "text": "First, dense retrieval models need a large amount of labeled training data for a decent performance. However, as Figure 1 shows, the proportion of labeled query-document pairs is extremely small since it is almost impossible to rely on humans for the annotations of a large document corpus. Second, in order to adapt a retrieval model to the real world, where documents constantly emerge, handling unlabeled documents that are not seen during training should obviously be considered, but remains challenging."
   }
 ]