Add more details
Browse files
README.md
CHANGED
@@ -13,12 +13,33 @@ tags:
|
|
13 |
- RoBERTa-base
|
14 |
---
|
15 |
|
|
|
|
|
16 |
A model for predicting a subset of MediaFrames given an argument (has not to be structured in premise/ conclusion or something else). To investigate the generic frame classes, have a look at [The Media Frames Corpus: Annotations of Frames Across Issues](https://aclanthology.org/P15-2072/)
|
17 |
|
18 |
Also, this model was fine-tuned on the data provided by [this paper](https://aclanthology.org/P15-2072/). To be precise, we did the following:
|
19 |
|
20 |
> To apply these frames to arguments from DDO, we fine-tune a range of classifiers on a comprehensive training dataset of more than 10,000 newspaper articles that discuss immigration, same-sex marriage, and marijuana, containing 146,001 labeled text spans labeled with a single MediaFrame-class per annotator. To apply this dataset to our argumentative domain, we broaden the annotated spans to sentence level (see [here](https://www.degruyter.com/document/doi/10.1515/itit-2020-0054/html)). Since an argument can address more than a single frame, we design the argument-frame classification task as a multi-label problem by combining all annotations for a sentence into a frame target set. In addition, to broaden the target frame sets, we create new instances merging two instances by combining their textual representation and unifying their target frame set.
|
21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
On the test split of this composed dataset, we measure the following performances:
|
23 |
|
24 |
````txt
|
|
|
13 |
- RoBERTa-base
|
14 |
---
|
15 |
|
16 |
+
# Model for predicting MediaFrames on arguments
|
17 |
+
|
18 |
A model for predicting a subset of MediaFrames given an argument (has not to be structured in premise/ conclusion or something else). To investigate the generic frame classes, have a look at [The Media Frames Corpus: Annotations of Frames Across Issues](https://aclanthology.org/P15-2072/)
|
19 |
|
20 |
Also, this model was fine-tuned on the data provided by [this paper](https://aclanthology.org/P15-2072/). To be precise, we did the following:
|
21 |
|
22 |
> To apply these frames to arguments from DDO, we fine-tune a range of classifiers on a comprehensive training dataset of more than 10,000 newspaper articles that discuss immigration, same-sex marriage, and marijuana, containing 146,001 labeled text spans labeled with a single MediaFrame-class per annotator. To apply this dataset to our argumentative domain, we broaden the annotated spans to sentence level (see [here](https://www.degruyter.com/document/doi/10.1515/itit-2020-0054/html)). Since an argument can address more than a single frame, we design the argument-frame classification task as a multi-label problem by combining all annotations for a sentence into a frame target set. In addition, to broaden the target frame sets, we create new instances merging two instances by combining their textual representation and unifying their target frame set.
|
23 |
|
24 |
+
## Used arguments for fine-tuning
|
25 |
+
|
26 |
+
````txt
|
27 |
+
per_device_train_batch_size=16,
|
28 |
+
per_device_eval_batch_size=64,
|
29 |
+
group_by_length=False,
|
30 |
+
evaluation_strategy="epoch",
|
31 |
+
num_train_epochs=5,
|
32 |
+
save_strategy="epoch",
|
33 |
+
load_best_model_at_end=True,
|
34 |
+
save_total_limit=3,
|
35 |
+
metric_for_best_model="eval_macro avg -> f1-score",
|
36 |
+
greater_is_better=True,
|
37 |
+
learning_rate=5e-5,
|
38 |
+
warmup_ratio=0.1
|
39 |
+
````
|
40 |
+
|
41 |
+
## Performance
|
42 |
+
|
43 |
On the test split of this composed dataset, we measure the following performances:
|
44 |
|
45 |
````txt
|