nazneen commited on
Commit
13a6f0e
1 Parent(s): 78ce8a3

model documentation

Browse files
Files changed (1) hide show
  1. README.md +201 -0
README.md ADDED
@@ -0,0 +1,201 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - pegasus
4
+
5
+ ---
6
+ # Model Card for brio-xsum-cased
7
+
8
+
9
+ # Model Details
10
+
11
+ ## Model Description
12
+
13
+ BRIO: Bringing Order to Abstractive Summarization
14
+
15
+ - **Developed by:** Yale LILY Lab
16
+ - **Shared by [Optional]:** Hugging Face
17
+ - **Model type:** PEGASUS
18
+ - **Language(s) (NLP):** Text2Text Generation
19
+ - **License:** More information needed
20
+ - **Related Models:**
21
+ - **Parent Model:** PEGASUS
22
+ - **Resources for more information:**
23
+ - [Github Repo](https://github.com/Yale-LILY/BRIO)
24
+ - [Associated Paper](https://arxiv.org/abs/2203.16804)
25
+ - [Associated Space](https://huggingface.co/spaces/darveen/text_summarizer)
26
+
27
+
28
+ # Uses
29
+
30
+ ## Direct Use
31
+ This model can be used for the task of Text2Text Generation
32
+
33
+ ## Downstream Use [Optional]
34
+
35
+ The model creators note in the [associated paper](https://arxiv.org/abs/2203.16804)
36
+ > It is possible to apply our method in a reinforcement learning setting, where the candidate summaries are dynamically generated.
37
+
38
+ ## Out-of-Scope Use
39
+
40
+
41
+ The model should not be used to intentionally create hostile or alienating environments for people.
42
+
43
+ # Bias, Risks, and Limitations
44
+
45
+
46
+ Significant research has explored bias and fairness issues with language models (see, e.g., [Sheng et al. (2021)](https://aclanthology.org/2021.acl-long.330.pdf) and [Bender et al. (2021)](https://dl.acm.org/doi/pdf/10.1145/3442188.3445922)). Predictions generated by the model may include disturbing and harmful stereotypes across protected classes; identity characteristics; and sensitive, social, and occupational groups.
47
+
48
+
49
+ ## Recommendations
50
+
51
+
52
+ Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
53
+
54
+
55
+ # Training Details
56
+
57
+ ## Training Data
58
+ The model creators note in the [associated paper](https://arxiv.org/abs/2203.16804)
59
+ > CNNDM4: is a large scale news dataset.
60
+ Nallapati et al: we treat the news articles as the source documents and the associated highlights as the summaries.
61
+ XSum5: is a highly abstractive dataset of articles from the British Broadcasting Corporation (BBC). NYT6: contains articles from the New York Times and the associated summaries
62
+
63
+ ## Training Procedure
64
+
65
+
66
+ ### Preprocessing
67
+ The model creators note in the [associated paper](https://arxiv.org/abs/2203.16804)
68
+ > We follow Kedzie et al. (2018) for data preprocessing and splitting, and use the associated archival abstracts as the summaries
69
+
70
+ ### Speeds, Sizes, Times
71
+
72
+ More information needed
73
+
74
+ # Evaluation
75
+
76
+
77
+ ## Testing Data, Factors & Metrics
78
+
79
+ ### Testing Data
80
+
81
+ More information needed
82
+
83
+ ### Factors
84
+
85
+ More information needed
86
+
87
+ ### Metrics
88
+
89
+ More information needed
90
+
91
+ ## Results
92
+
93
+
94
+
95
+ ### CNNDM
96
+ | | ROUGE-1 | ROUGE-2 | ROUGE-L |
97
+ |----------|---------|---------|---------|
98
+ | BART | 44.16 | 21.28 | 40.90 |
99
+ | Ours | 47.78 | 23.55 | 44.57 |
100
+
101
+
102
+ ### XSum
103
+ | | ROUGE-1 | ROUGE-2 | ROUGE-L |
104
+ |----------|---------|---------|---------|
105
+ | Pegasus | 47.21 | 24.56 | 39.25 |
106
+ | Ours | 49.07 | 25.59 | 40.40 |
107
+
108
+
109
+ ### NYT
110
+ | | ROUGE-1 | ROUGE-2 | ROUGE-L |
111
+ |----------|---------|---------|---------|
112
+ | BART | 55.78 | 36.61 | 52.60 |
113
+ | Ours | 57.75 | 38.64 | 54.54 |
114
+
115
+
116
+
117
+ # Model Examination
118
+ The model creators note in the [associated paper](https://arxiv.org/abs/2203.16804)
119
+ We attribute BRIO-Ctr’s superior performance to its use of the same model architecture (BART) for both candidate generation and scoring, while SimCLS uses RoBERTa as the evaluation model. As a result, BRIO-Ctr maximizes the parameter sharing between the two stages, and preserves the power of the Seq2Seq model pre-trained on the same dataset.
120
+
121
+ # Environmental Impact
122
+
123
+
124
+ Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
125
+
126
+ - **Hardware Type:** More information needed
127
+ - **Hours used:** More information needed
128
+ - **Cloud Provider:** More information needed
129
+ - **Compute Region:** More information needed
130
+ - **Carbon Emitted:** More information needed
131
+
132
+ # Technical Specifications [optional]
133
+
134
+ ## Model Architecture and Objective
135
+ The model creators note in the [associated paper](https://arxiv.org/abs/2203.16804)
136
+
137
+ > Formulate summarization as a sequence-to-sequence (Seq2Seq) problem
138
+
139
+ ## Compute Infrastructure
140
+
141
+ More information needed
142
+
143
+ ### Hardware
144
+
145
+ More information needed
146
+
147
+ ### Software
148
+
149
+ More information needed
150
+
151
+ # Citation
152
+
153
+
154
+ **BibTeX:**
155
+ ```
156
+ @misc{https://doi.org/10.48550/arxiv.2203.16804,
157
+ doi = {10.48550/ARXIV.2203.16804},
158
+
159
+ url = {https://arxiv.org/abs/2203.16804},
160
+
161
+ author = {Liu, Yixin and Liu, Pengfei and Radev, Dragomir and Neubig, Graham},
162
+
163
+ keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
164
+
165
+ title = {BRIO: Bringing Order to Abstractive Summarization},
166
+ ```
167
+
168
+
169
+
170
+ # Glossary [optional]
171
+
172
+ More information needed
173
+
174
+ # More Information [optional]
175
+
176
+ More information needed
177
+
178
+ # Model Card Authors [optional]
179
+
180
+ Yale LILY Lab in collaboration with Ezi Ozoani and the Hugging Face team
181
+
182
+ # Model Card Contact
183
+
184
+ More information needed
185
+
186
+ # How to Get Started with the Model
187
+
188
+ Use the code below to get started with the model.
189
+
190
+ <details>
191
+ <summary> Click to expand </summary>
192
+
193
+ ```python
194
+ from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
195
+
196
+ tokenizer = AutoTokenizer.from_pretrained("Yale-LILY/brio-xsum-cased")
197
+
198
+ model = AutoModelForSeq2SeqLM.from_pretrained("Yale-LILY/brio-xsum-cased")
199
+
200
+ ```
201
+ </details>