Add Module 2 reports

reports/module_2_emotion_classification/README.md

@@ -1,11 +1,12 @@
 # Module 2 Reports
 
-This folder is populated after running `notebooks/module_2_emotion_training.ipynb` in Colab.
+These files summarize the DistilBERT emotion classifier trained in `notebooks/module_2_emotion_training.ipynb`.
 
-Expected generated files:
+Included reports:
 
 - `metrics_summary.json`
 - `test_classification_report.txt`
 - `test_classification_report.csv`
-- `test_confusion_matrix.csv`
-- `explanation_examples.json`
+- `explanation_examples.json` when explanation outputs are available in the notebook
+
+The notebook also generates `test_confusion_matrix.csv` during a full Colab run. The exact matrix is not reconstructed here unless that generated CSV is copied back from Colab, because the matrix values were not printed in the saved notebook output.

reports/module_2_emotion_classification/explanation_examples.json

@@ -0,0 +1,122 @@
+[
+  {
+    "text": "I feel anxious and overwhelmed and I cannot sleep.",
+    "prediction": {
+      "emotion": "fear",
+      "confidence": 0.9993083477020264
+    },
+    "top_evidence": [
+      {
+        "word": "anxious",
+        "impact": 0.4232
+      },
+      {
+        "word": "feel",
+        "impact": 0.0002
+      },
+      {
+        "word": "overwhelmed",
+        "impact": 0.0001
+      },
+      {
+        "word": "I",
+        "impact": -0.0
+      },
+      {
+        "word": "and",
+        "impact": -0.0
+      },
+      {
+        "word": "and",
+        "impact": -0.0
+      }
+    ]
+  },
+  {
+    "text": "I finally feel hopeful and proud of myself.",
+    "prediction": {
+      "emotion": "joy",
+      "confidence": 0.9997496008872986
+    },
+    "top_evidence": [
+      {
+        "word": "I",
+        "impact": -0.0
+      },
+      {
+        "word": "finally",
+        "impact": 0.0
+      },
+      {
+        "word": "feel",
+        "impact": -0.0
+      },
+      {
+        "word": "hopeful",
+        "impact": 0.0
+      },
+      {
+        "word": "and",
+        "impact": 0.0
+      },
+      {
+        "word": "proud",
+        "impact": 0.0
+      }
+    ]
+  },
+  {
+    "text": "I am really sad because no one listens to me.",
+    "prediction": {
+      "emotion": "sadness",
+      "confidence": 0.9994388222694397
+    },
+    "top_evidence": [
+      {
+        "word": "sad",
+        "impact": 0.9994
+      },
+      {
+        "word": "really",
+        "impact": 0.0077
+      },
+      {
+        "word": "I",
+        "impact": 0.0002
+      },
+      {
+        "word": "me.",
+        "impact": 0.0001
+      },
+      {
+        "word": "am",
+        "impact": 0.0
+      },
+      {
+        "word": "to",
+        "impact": -0.0
+      }
+    ]
+  },
+  {
+    "text": "Amazing we won!",
+    "prediction": {
+      "emotion": "surprise",
+      "confidence": 0.8034693598747253
+    },
+    "top_evidence": [
+      {
+        "word": "Amazing",
+        "impact": 0.8035
+      },
+      {
+        "word": "we",
+        "impact": 0.8035
+      },
+      {
+        "word": "won!",
+        "impact": -0.1771
+      }
+    ]
+  }
+]

reports/module_2_emotion_classification/metrics_summary.json

@@ -0,0 +1,30 @@
+{
+  "base_model": "distilbert-base-uncased",
+  "dataset": "dair-ai/emotion",
+  "labels": [
+    "sadness",
+    "joy",
+    "love",
+    "anger",
+    "fear",
+    "surprise"
+  ],
+  "training": {
+    "epochs": 5,
+    "learning_rate": 2e-05,
+    "train_batch_size": 16,
+    "eval_batch_size": 32,
+    "best_model_metric": "macro_f1"
+  },
+  "validation_metrics": {
+    "eval_loss": 0.1850450038909912,
+    "accuracy": 0.944,
+    "macro_f1": 0.9217609879676768
+  },
+  "test_metrics": {
+    "accuracy": 0.924,
+    "macro_f1": 0.8747072046581201
+  },
+  "explainability_method": "word removal: compare predicted confidence before and after removing each word",
+  "source": "Reconstructed from saved notebook outputs in notebooks/module_2_emotion_training.ipynb"
+}

reports/module_2_emotion_classification/test_classification_report.csv

@@ -0,0 +1,10 @@
+label,precision,recall,f1-score,support
+sadness,0.97,0.96,0.97,581
+joy,0.95,0.94,0.94,695
+love,0.81,0.82,0.82,159
+anger,0.91,0.93,0.92,275
+fear,0.88,0.92,0.90,224
+surprise,0.77,0.65,0.70,66
+accuracy,,,0.92,2000
+macro avg,0.88,0.87,0.87,2000
+weighted avg,0.92,0.92,0.92,2000

reports/module_2_emotion_classification/test_classification_report.txt

@@ -0,0 +1,12 @@
+              precision    recall  f1-score   support
+
+sadness       0.97      0.96      0.97       581
+         joy       0.95      0.94      0.94       695
+        love       0.81      0.82      0.82       159
+       anger       0.91      0.93      0.92       275
+        fear       0.88      0.92      0.90       224
+    surprise       0.77      0.65      0.70        66
+
+    accuracy                           0.92      2000
+   macro avg       0.88      0.87      0.87      2000
+weighted avg       0.92      0.92      0.92      2000