Update README.md

README.md

@@ -65,7 +65,7 @@ BEANS-Zero combines data from several well-known sources. There are total of 113
 
 ### Already Part of the BEANS Benchmark
 - **esc-50:** A labeled collection of 2000 environmental audio recordings (5 seconds each) spanning 50 classes. ([CC-BY-NC](http://dx.doi.org/10.1145/2733373.2806390))
-- **Watkins:** Marine mammal sound recordings.
+- **Watkins:** Marine mammal sound recordings. (free for personal/academic uses)
 - **CBI:** Cornell Birdcall Identification dataset. ([CC-BY-NC-SA](https://www.kaggle.com/competitions/birdsong-recognition/overview))
 - **HumBugDB:** A large-scale, multi-species dataset of mosquito sounds. ([CC-BY](https://openreview.net/forum?id=vhjsBtq9OxO))
 - **Enabirds:** Bird dawn chorus detection data. ([CC0](https://esajournals.onlinelibrary.wiley.com/doi/full/10.1002/ecy.3329))
@@ -80,7 +80,7 @@ BEANS-Zero combines data from several well-known sources. There are total of 113
 - **Unseen-genus-cmn / sci / tax:** Tasks for classifying genus-level names (common, scientific, taxonomic) of unseen species.
 - **Unseen-family-cmn / sci / tax:** Tasks for classifying family-level names (common, scientific, taxonomic) of unseen species.
 - **Captioning:** English captions for bioacoustic recordings.
-- **ZF-Indiv:** A zebra finch dataset with individual and call-type annotations.
+- **ZF-Indiv:** A zebra finch dataset with individual and call-type annotations, from [Elie et al 2020 (https://doi.org/10.6084/m9.figshare.11905533.v1)]
 
 Each subset has its own metadata and licensing details. Each sample
 
@@ -112,10 +112,10 @@ These tasks are particularly useful for exploring zero-shot learning application
 ## Data Fields
 The following fields are present in each example:
 - **source_dataset** (str): One of the source datasets mentioned above
-- **audio** (Sequence[float]): The audio data in float32 format. The audio is not decoded, and the sample rate is 16000 Hz.
+- **audio** (Sequence[float]): The audio data in float32 format. The audio is not decoded.
 - **id** (str): Sample uuid.
 - **created_at** (str): Sample creation datetime in utc
-- **metadata** (str): Each sample can have a different duration
+- **metadata** (str): Each sample can have a different duration and a different sample rate. `sample_rate = json.loads(sample["metadata"])["sample_rate"]`
 - **file_name** (str): Sample file_name
 - **instruction** (str): A prompt (a query) corresponding to the audio for your audio-text model with a placeholder for audio tokens. E.g. '<Audio><AudioHere></Audio> What is the scientific name for the focal species in the audio?'
 - **instruction_text** (str): Same as **instruction** but without the placeholder for audio tokens.