bug: fix MRR and MAP calculations (#7841)

* bug: fix MRR and MAP calculations

haystack/components/evaluators/document_map.py

@@ -43,6 +43,7 @@ class DocumentMAPEvaluator:
     ```
     """
 
+    # Refer to https://www.pinecone.io/learn/offline-evaluation/ for the algorithm.
     @component.output_types(score=float, individual_scores=List[float])
     def run(
         self, ground_truth_documents: List[List[Document]], retrieved_documents: List[List[Document]]
@@ -69,25 +70,21 @@ class DocumentMAPEvaluator:
         individual_scores = []
 
         for ground_truth, retrieved in zip(ground_truth_documents, retrieved_documents):
-            score = 0.0
-            for ground_document in ground_truth:
-                if ground_document.content is None:
-                    continue
-
             average_precision = 0.0
+            average_precision_numerator = 0.0
             relevant_documents = 0
 
+            ground_truth_contents = [doc.content for doc in ground_truth if doc.content is not None]
             for rank, retrieved_document in enumerate(retrieved):
                 if retrieved_document.content is None:
                     continue
 
-                    if ground_document.content in retrieved_document.content:
+                if retrieved_document.content in ground_truth_contents:
                     relevant_documents += 1
-                        average_precision += relevant_documents / (rank + 1)
+                    average_precision_numerator += relevant_documents / (rank + 1)
             if relevant_documents > 0:
-                    score = average_precision / relevant_documents
-            individual_scores.append(score)
-
-        score = sum(individual_scores) / len(retrieved_documents)
+                average_precision = average_precision_numerator / relevant_documents
+            individual_scores.append(average_precision)
 
+        score = sum(individual_scores) / len(ground_truth_documents)
         return {"score": score, "individual_scores": individual_scores}

haystack/components/evaluators/document_mrr.py

@@ -41,6 +41,7 @@ class DocumentMRREvaluator:
     ```
     """
 
+    # Refer to https://www.pinecone.io/learn/offline-evaluation/ for the algorithm.
     @component.output_types(score=float, individual_scores=List[float])
     def run(
         self, ground_truth_documents: List[List[Document]], retrieved_documents: List[List[Document]]
@@ -67,20 +68,17 @@ class DocumentMRREvaluator:
         individual_scores = []
 
         for ground_truth, retrieved in zip(ground_truth_documents, retrieved_documents):
-            score = 0.0
-            for ground_document in ground_truth:
-                if ground_document.content is None:
-                    continue
+            reciprocal_rank = 0.0
 
+            ground_truth_contents = [doc.content for doc in ground_truth if doc.content is not None]
             for rank, retrieved_document in enumerate(retrieved):
                 if retrieved_document.content is None:
                     continue
-
-                    if ground_document.content in retrieved_document.content:
-                        score = 1 / (rank + 1)
+                if retrieved_document.content in ground_truth_contents:
+                    reciprocal_rank = 1 / (rank + 1)
                     break
-            individual_scores.append(score)
+            individual_scores.append(reciprocal_rank)
 
-        score = sum(individual_scores) / len(retrieved_documents)
+        score = sum(individual_scores) / len(ground_truth_documents)
 
         return {"score": score, "individual_scores": individual_scores}

releasenotes/notes/fix-issue-7758-d35b687ca226a707.yaml

@@ -0,0 +1,4 @@
+---
+fixes:
+  - |
+    Fixed the calculation for MRR and MAP scores.

test/components/evaluators/test_document_map.py

@@ -62,7 +62,17 @@ def test_run_with_complex_data():
             ],
         ],
     )
-    assert result == {"individual_scores": [1.0, 0.8333333333333333, 1.0, 0.5, 0.0, 1.0], "score": 0.7222222222222222}
+    assert result == {
+        "individual_scores": [
+            1.0,
+            pytest.approx(0.8333333333333333),
+            1.0,
+            pytest.approx(0.5833333333333333),
+            0.0,
+            pytest.approx(0.8055555555555555),
+        ],
+        "score": pytest.approx(0.7037037037037037),
+    }
 
 
 def test_run_with_different_lengths():