Updated blab dataset

README.md

@@ -26,16 +26,14 @@ configs:
         dtype: string
       - name: question
         dtype: string
-      - name: answer_type
-        dtype: string
       - name: groundtruth
         dtype: LargeList
         inner_dtype:
           - name: word
             dtype: string
-          - name: start
+          - name: start_timestamp
             dtype: float32
-          - name: end
+          - name: end_timestamp
             dtype: float32
 - config_name: advertisement_localization
   features:
@@ -45,8 +43,6 @@ configs:
         dtype: string
       - name: question
         dtype: string
-      - name: answer_type
-        dtype: string
       - name: groundtruth
         dtype: Struct
         fields:
@@ -77,8 +73,6 @@ configs:
         dtype: string
       - name: question
         dtype: string
-      - name: answer_type
-        dtype: string
       - name: groundtruth
         dtype: Struct
         fields:
@@ -98,9 +92,9 @@ configs:
             inner_dtype:
               - name: word
                 dtype: string
-              - name: start
+              - name: start_time
                 dtype: float32
-              - name: end
+              - name: end_time
                 dtype: float32
 - config_name: speaker_number_estimation
   features:
@@ -136,6 +130,13 @@ configs:
         dtype: string
       - name: groundtruth
         dtype: float32
+      - name: event_timestamps
+        dtype: LargeList
+        inner_dtype:
+          - name: start_timestamp
+            dtype: float32
+          - name: end_timestamp
+            dtype: float32
 - config_name: emotion_ranking
   features:
       - name: video_url
@@ -158,8 +159,13 @@ configs:
         dtype: string
       - name: option_E
         dtype: string
-      - name: correct_answer
+      - name: groundtruth
         dtype: string
+      - name: event_timestamps
+        dtype: LargeList
+        inner_dtype:
+          - name: start_timestamp
+            dtype: float32
 - config_name: emotion_reasoning
   features:
       - name: video_url
@@ -180,8 +186,15 @@ configs:
         dtype: string
       - name: option_D
         dtype: string
-      - name: correct_answer
+      - name: groundtruth
         dtype: string
+      - name: event_timestamps
+        dtype: LargeList
+        inner_dtype:
+          - name: start_timestamp
+            dtype: float32
+          - name: event
+            dtype: string
 ---
 
 
@@ -193,6 +206,8 @@ Brutally Long Audio Bench (BLAB) is a challenging long-form audio benchmark that
 
 NB: This data should only be used for evaluation purposes and not for model training.
 
+We provide time-aligned metadata (`event_timestamps`) for the **Event Duration**, **Emotion Reasoning**, and **Emotion Ranking** tasks; time-aligned metadata for **Speaker Number Estimation** is coming soon. This metadata can also be used to evaluate long-form reasoning across timestamps. All timestamps in the dataset, including `event_timestamps`, are expressed as float seconds with millisecond precision (e.g., `200.349`).
+
 
 
 ## Tasks Covered in BLAB
@@ -203,7 +218,7 @@ NB: This data should only be used for evaluation purposes and not for model trai
 * **Advertisement Localization:** Locate and transcribe advertisement segments within a podcast.
 
 ### Counting
-* **Speaker Number Estimation:** Determine the number of unique speakers present in the full audio segment.
+* **Speaker Number Estimation:** Determine the number of unique speakers present in the full audio segment. The groundtruth is provided as a range `[min, max]`.
 
 ### Duration
 * **Event Duration:** Calculate the duration of specific acoustic events (e.g., laughter in a comedy special, question-and-answer segments in a panel session, or a particular speaker’s total speaking time in a meeting) within an audio sample,.

blab_long_audio.py

@@ -47,8 +47,8 @@ class BLAB(datasets.GeneratorBasedBuilder):
                     "groundtruth": LargeList(
                         feature=Features({
                             "word": Value("string"),
-                            "start": Value("float32"),
-                            "end": Value("float32"),
+                            "start_timestamp": Value("float32"),
+                            "end_timestamp": Value("float32"),
                         })
                     )
                 }),
@@ -101,8 +101,8 @@ class BLAB(datasets.GeneratorBasedBuilder):
                         "word_timestamp": LargeList(
                             feature=Features({
                                 "word": Value("string"),
-                                "start": Value("float32"),
-                                "end": Value("float32"),
+                                "start_time": Value("float32"),
+                                "end_time": Value("float32"),
                             }),
                         ),
                     })
@@ -143,6 +143,12 @@ class BLAB(datasets.GeneratorBasedBuilder):
                     "question": Value("string"),
                     "groundtruth": Value("float32"),
                     "answer_type": Value("string"),
+                    "event_timestamps": LargeList(
+                        feature=Features({
+                            "start_timestamp": Value("float32"),
+                            "end_timestamp": Value("float32"),
+                        })
+                    ),
                 }),
                 description="Schema for Event Duration task: identifying and timing specific events.",
                 # ... (other metadata)
@@ -161,7 +167,12 @@ class BLAB(datasets.GeneratorBasedBuilder):
                     "option_C": Value("string"),
                     "option_D": Value("string"),
                     "option_E": Value("string"),
-                    "correct_answer": Value("string"), # Stores the correct_answer string
+                    "groundtruth": Value("string"),
+                    "event_timestamps": LargeList(
+                        feature=Features({
+                            "start_timestamp": Value("float32"),
+                        })
+                    ),
                 }),
                 description="Schema for Emotion Ranking task: selecting the best emotion option.",
                 # ... (other metadata)
@@ -179,7 +190,13 @@ class BLAB(datasets.GeneratorBasedBuilder):
                     "option_B": Value("string"),
                     "option_C": Value("string"),
                     "option_D": Value("string"),
-                    "correct_answer": Value("string"), # Stores the correct_answer string
+                    "groundtruth": Value("string"),
+                    "event_timestamps": LargeList(
+                        feature=Features({
+                            "start_timestamp": Value("float32"),
+                            "event": Value("string"),
+                        })
+                    ),
                 }),
                 description="Schema for Emotion Reasoning task: explaining emotional context.",
                 # ... (other metadata)
@@ -250,8 +267,8 @@ class BLAB(datasets.GeneratorBasedBuilder):
                             if isinstance(item, dict):
                                 processed_groundtruth.append({
                                     "word": item.get("word", None),
-                                    "start": item.get("start", None),
-                                    "end": item.get("end", None),
+                                    "start_timestamp": item.get("start_timestamp", None),
+                                    "end_timestamp": item.get("end_timestamp", None),
                                 })
                         example["groundtruth"] = processed_groundtruth
 
@@ -298,8 +315,8 @@ class BLAB(datasets.GeneratorBasedBuilder):
                             if isinstance(word_item, dict):
                                 processed_word_timestamps.append({
                                     "word": word_item.get("word", None),
-                                    "start": word_item.get("start", None),
-                                    "end": word_item.get("end", None),
+                                    "start_time": word_item.get("start_time", None),
+                                    "end_time": word_item.get("end_time", None),
                                 })
                         example["groundtruth"] = {
                             "entities": processed_entities,
@@ -323,6 +340,11 @@ class BLAB(datasets.GeneratorBasedBuilder):
                     elif self.config.name == "event_duration":
                         example["groundtruth"] = data.get("groundtruth", None)
                         example["answer_type"] = data.get("answer_type", None)
+                        raw_event_ts = data.get("event_timestamps", [])
+                        example["event_timestamps"] = [
+                            {"start_timestamp": et.get("start_timestamp", None), "end_timestamp": et.get("end_timestamp", None)}
+                            for et in raw_event_ts if isinstance(et, dict)
+                        ]
 
                     elif self.config.name == "emotion_ranking":
                         example["type"] = data.get("type", None)
@@ -332,7 +354,12 @@ class BLAB(datasets.GeneratorBasedBuilder):
                         example["option_C"] = data.get("option_C", None)
                         example["option_D"] = data.get("option_D", None)
                         example["option_E"] = data.get("option_E", None)
-                        example["correct_answer"] = data.get("correct_answer", None)
+                        example["groundtruth"] = data.get("groundtruth", None)
+                        raw_event_ts = data.get("event_timestamps", [])
+                        example["event_timestamps"] = [
+                            {"start_timestamp": et.get("start_timestamp", None)}
+                            for et in raw_event_ts if isinstance(et, dict)
+                        ]
 
                     elif self.config.name == "emotion_reasoning":
                         example["type"] = data.get("type", None)
@@ -341,7 +368,12 @@ class BLAB(datasets.GeneratorBasedBuilder):
                         example["option_B"] = data.get("option_B", None)
                         example["option_C"] = data.get("option_C", None)
                         example["option_D"] = data.get("option_D", None)
-                        example["correct_answer"] = data.get("correct_answer", None)
+                        example["groundtruth"] = data.get("groundtruth", None)
+                        raw_event_ts = data.get("event_timestamps", [])
+                        example["event_timestamps"] = [
+                            {"start_timestamp": et.get("start_timestamp", None), "event": et.get("event", None)}
+                            for et in raw_event_ts if isinstance(et, dict)
+                        ]
 
                     else:
                         raise ValueError(f"Unknown config name: {self.config.name}. This should not happen if BUILDER_CONFIGS and _info are consistent.")

blab_long_audio/advertisement_localization.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd9361ec91fa14691b3a20cfc96cae235d6bcbb20c58a583a515a828ac0beb5d
-size 387057817
+oid sha256:6329627fa291e052838f4f3d62a4f10e0b70b6853e4d02a1a8b7e3bb8ceffcf1
+size 188071676

blab_long_audio/emotion_ranking.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6779de03411d7d73c5587cd75afb6bd12397f2868e43c2a97d6fc3c60d23ca6f
-size 87283
+oid sha256:c23842b90d187b0d20b556bae8aa6e45e32b9f813641c5f67841e4b5e41c154d
+size 108514

blab_long_audio/emotion_reasoning.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:988caa7276998f1bff907c95d0e79a616cbb273586a77b795187ac53f8d7c368
-size 31796
+oid sha256:34ddb2af6e4a15641061f02ef4fc76262dcc7823a57c8f5da03c54ec27cc754b
+size 56911

blab_long_audio/entire_duration.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:75721fa1c6b04d188d5ce583d7ab015fd28ba9e05c687a33e52c32406be087cc
+oid sha256:39071d6440fbcf51b178fb0ed8dbae0c84817904b83a7397c3c980d414c53c22
 size 45631

blab_long_audio/event_duration.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ecf59f114e3660ca8cc11bbc64a3309481af1284b4524d55a2d0f7db4dc49c93
-size 62680
+oid sha256:0feb492947d698098efe4cb44100c1daeb23bce8133f3419b0852ede8f6354d0
+size 168378

blab_long_audio/named_entity_localization.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:834fd0b3f150829421c45121ffdf3804039c92202aeeec32eb090eb4f6483d5c
-size 308743971
+oid sha256:60baae7931dc5f5d1ee1f17b704dc90bfa9cb8ba0ee29e41eb2db648cea57940
+size 162026769

blab_long_audio/speaker_number_estimation.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:820dd675d315385c536dd4998ff7358ed022f0c7690bacaa9c90ed6550fe4bbe
+oid sha256:9cab2ddaa5de50a16858e3d8361f352c854acba0bb3efd48a890b8c46b8f3b29
 size 49593

blab_long_audio/word_localization.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7e9082228785306a544af3b7ba753f308e1f5b12e49c2a1808e68ee11461190a
-size 268796546
+oid sha256:b43dfa1813ba87db0458792691bf51456d2f26e09fe8f039ed2f07f075e5cc2e
+size 184463350