update ch02

rag-advanced/notebooks/Chapter02.ipynb

@@ -204,7 +204,15 @@
     "\n",
     "We can evaluate a retriever using traditional ML metrics. We can also evaluate by using a powerful LLM (next section).\n",
     "\n",
-    "Below we are importing both traditional metrics and LLM as a judge metric from the `scripts/retrieval_metrics.py` file."
+    "Below we are importing both traditional metrics and LLM as a judge metric from the `scripts/retrieval_metrics.py` file.\n",
+    "\n",
+    "* **Hit Rate**: Measures the proportion of queries where the retriever successfully returns at least one relevant document.\n",
+    "* **MRR (Mean Reciprocal Rank)**: Evaluates how quickly the retriever returns the first relevant document, based on the reciprocal of its rank.\n",
+    "* **NDCG (Normalized Discounted Cumulative Gain)**: Assesses the quality of the ranked retrieval results, giving more importance to relevant documents appearing earlier.\n",
+    "* **MAP (Mean Average Precision)**: Computes the mean precision across all relevant documents retrieved, considering the rank of each relevant document.\n",
+    "* **Precision**: Measures the ratio of relevant documents retrieved to the total documents retrieved by the retriever.\n",
+    "* **Recall**: Evaluates the ratio of relevant documents retrieved to the total relevant documents available for the query.\n",
+    "* **F1 Score**: The harmonic mean of precision and recall, providing a balance between both metrics to gauge retriever performance."
    ]
   },
   {
@@ -396,7 +404,7 @@
    "source": [
     "## Exercise\n",
     "\n",
-    "1. Implement the `Relevance` and `Faithfulness` evaluators and evaluate the pipeline on all the dimensions.\n",
+    "1. Implement the `Relevance` and `Faithfulness` LLM evaluators (LLM as a Judge) and evaluate the pipeline on all the dimensions.\n",
     "2. Generate and share a W&B report with the following sections in the form of tables and charts:\n",
     "    \n",
     "    - Summary of the evaluation\n",