Update description of Squim pipelines (#3254)

Summary:
- Add citations of [`TorchAudio-Squim`](https://arxiv.org/abs/2304.01448) publication.
- Update descriptions in the `SQUIM_OBJECTIVE` and `SQUIM_SUBJECTIVE` pipelines.

Pull Request resolved: https://github.com/pytorch/audio/pull/3254

Reviewed By: hwangjeff

Differential Revision: D44802015

Pulled By: nateanl

fbshipit-source-id: ca08298ec1eafefdd671ff2e010ef18f7372f9f8

docs/source/refs.bib

@@ -556,3 +556,9 @@ abstract = {End-to-end spoken language translation (SLT) has recently gained pop
   year={2014},
   publisher={IEEE}
 }
+@article{kumar2023torchaudio,
+  title={TorchAudio-Squim: Reference-less Speech Quality and Intelligibility measures in TorchAudio},
+  author={Kumar, Anurag and Tan, Ke and Ni, Zhaoheng and Manocha, Pranay and Zhang, Xiaohui and Henderson, Ethan and Xu, Buye},
+  journal={arXiv preprint arXiv:2304.01448},
+  year={2023}
+}

torchaudio/prototype/pipelines/squim_pipeline.py

@@ -79,8 +79,8 @@ SQUIM_OBJECTIVE = SquimObjectiveBundle(
     "squim_objective_dns2020.pth",
     _sample_rate=16000,
 )
-SQUIM_OBJECTIVE.__doc__ = """SquimObjective pipeline, trained on the *DNS 2020 Dataset*
-    :cite:`reddy2020interspeech`.
+SQUIM_OBJECTIVE.__doc__ = """SquimObjective pipeline trained using approach described in
+    :cite:`kumar2023torchaudio` on the *DNS 2020 Dataset* :cite:`reddy2020interspeech`.
 
     The underlying model is constructed by :py:func:`torchaudio.prototype.models.squim_objective_base`.
     The weights are under `Creative Commons Attribution 4.0 International License
@@ -166,8 +166,9 @@ SQUIM_SUBJECTIVE = SquimSubjectiveBundle(
     "squim_subjective_bvcc_daps.pth",
     _sample_rate=16000,
 )
-SQUIM_SUBJECTIVE.__doc__ = """SquimSubjective pipeline, trained on the *BVCC*
-    :cite:`cooper2021voices` and *DAPS* :cite:`mysore2014can` datasets.
+SQUIM_SUBJECTIVE.__doc__ = """SquimSubjective pipeline trained
+    as described in :cite:`manocha2022speech` and :cite:`kumar2023torchaudio`
+    on the *BVCC* :cite:`cooper2021voices` and *DAPS* :cite:`mysore2014can` datasets.
 
     The underlying model is constructed by :py:func:`torchaudio.prototype.models.squim_subjective_base`.
     The weights are under `Creative Commons Attribution Non Commercial 4.0 International