Datasets:

Modalities:
Image
Languages:
English
ArXiv:
Libraries:
Datasets
License:
Psychias commited on
Commit
e879b8f
·
1 Parent(s): 171565f

Update dataset card documentation

Browse files
Files changed (1) hide show
  1. README.md +41 -0
README.md CHANGED
@@ -68,12 +68,53 @@ python scripts/check_keyframes.py # "OK: all keyframe references resolve."
68
  Frames are added only, never deleted, by `scripts/push_keyframes.py`, which uploads the lecture
69
  folders the released rows reference and this repo lacks.
70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
71
  ## Citation
72
 
 
 
73
  ```bibtex
74
  @misc{ad4edu2026,
75
  title = {AD4Edu: Benchmarking Audio Description Quality for Slide-Based Lecture Videos},
76
  author = {Psychias, Stelios},
77
  year = {2026}
78
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
79
  ```
 
68
  Frames are added only, never deleted, by `scripts/push_keyframes.py`, which uploads the lecture
69
  folders the released rows reference and this repo lacks.
70
 
71
+ ## ADEduEval
72
+
73
+ ADEduEval is a diagnostic benchmark for evaluating audio descriptions of lecture content. It
74
+ operationalizes a 45-rule rubric across six dimensions—style, terminology, length, deixis,
75
+ faithfulness, and non-redundancy—and returns aggregate, category-level, and rule-level results.
76
+ Rather than providing only a single quality score, ADEduEval identifies the specific criteria that
77
+ each candidate description satisfies or violates. Its three evaluation tiers are available through
78
+ a single entry point: `scripts/eval_adedueval.py` (Python module:
79
+ `ad4edu.validation.adedueval_eval`).
80
+
81
+ ## Licensing
82
+
83
+ The synthetic ADs and the rule standard used with these frames are released under **CC BY-NC-SA
84
+ 4.0**, consistent with both source datasets — **LPM** and **M3AV** are each CC BY-NC-SA 4.0 (LPM's
85
+ videos are additionally under the standard YouTube License). The keyframes in this repository are
86
+ source-derived lecture stills and are not covered by the license for the synthetic ADs or rule
87
+ standard. Source videos and full per-lecture transcripts/OCR are not redistributed. **Confirm
88
+ per-source terms before any public release** and prefer reconstructing source-derived material from
89
+ the original corpora over redistributing it.
90
+
91
  ## Citation
92
 
93
+ Cite AD4Edu **and** the two source datasets it is derived from.
94
+
95
  ```bibtex
96
  @misc{ad4edu2026,
97
  title = {AD4Edu: Benchmarking Audio Description Quality for Slide-Based Lecture Videos},
98
  author = {Psychias, Stelios},
99
  year = {2026}
100
  }
101
+
102
+ @inproceedings{lee2023lpm,
103
+ title = {Lecture Presentations Multimodal Dataset: Towards Understanding Multimodality in Educational Videos},
104
+ author = {Lee, Dong Won and Ahuja, Chaitanya and Liang, Paul Pu and Natu, Sanika and Morency, Louis-Philippe},
105
+ booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
106
+ year = {2023},
107
+ note = {arXiv:2208.08080}
108
+ }
109
+
110
+ @inproceedings{chen2024m3av,
111
+ title = {{M3AV}: A Multimodal, Multigenre, and Multipurpose Audio-Visual Academic Lecture Dataset},
112
+ author = {Chen, Zhe and Liu, Heyang and Yu, Wenyi and Sun, Guangzhi and Liu, Hongcheng and Wu, Ji and Zhang, Chao and Wang, Yu and Wang, Yanfeng},
113
+ booktitle = {Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
114
+ pages = {9041--9060},
115
+ year = {2024},
116
+ publisher = {Association for Computational Linguistics},
117
+ doi = {10.18653/v1/2024.acl-long.489},
118
+ note = {arXiv:2403.14168}
119
+ }
120
  ```