Commit
•
02dea4d
1
Parent(s):
38c4d78
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -3,7 +3,7 @@
|
|
| 3 |
---
|
| 4 |
license: mit
|
| 5 |
language:
|
| 6 |
-
-
|
| 7 |
tags:
|
| 8 |
- zero-shot-classification
|
| 9 |
- text-classification
|
|
@@ -14,7 +14,7 @@ metrics:
|
|
| 14 |
---
|
| 15 |
# xlm-roberta-large-german-cap
|
| 16 |
## Model description
|
| 17 |
-
An `xlm-roberta-large` model finetuned on
|
| 18 |
|
| 19 |
## How to use the model
|
| 20 |
#### Loading and tokenizing input data
|
|
@@ -90,34 +90,33 @@ training_args = TrainingArguments(
|
|
| 90 |
We also incorporated an EarlyStoppingCallback in the process with a patience of 2 epochs.
|
| 91 |
|
| 92 |
## Model performance
|
| 93 |
-
The model was evaluated on a test set of
|
| 94 |
-
Model accuracy is **0.
|
| 95 |
| label | precision | recall | f1-score | support |
|
| 96 |
|:-------------|------------:|---------:|-----------:|----------:|
|
| 97 |
-
| 0 | 0.
|
| 98 |
-
| 1 | 0.
|
| 99 |
-
| 2 | 0.
|
| 100 |
-
| 3 | 0.
|
| 101 |
-
| 4 | 0.
|
| 102 |
-
| 5 | 0.
|
| 103 |
-
| 6 | 0.
|
| 104 |
-
| 7 | 0.
|
| 105 |
-
| 8 | 0.
|
| 106 |
-
| 9 | 0.
|
| 107 |
-
| 10 | 0.
|
| 108 |
-
| 11 | 0.
|
| 109 |
-
| 12 | 0.
|
| 110 |
-
| 13 | 0.
|
| 111 |
-
| 14 | 0.
|
| 112 |
-
| 15 | 0.
|
| 113 |
-
| 16 |
|
| 114 |
-
| 17 | 0.75 | 0.
|
| 115 |
-
| 18 | 0.
|
| 116 |
-
| 19 | 0
|
| 117 |
-
| 20 | 0
|
| 118 |
-
|
|
| 119 |
-
|
|
| 120 |
-
| weighted avg | 0.71 | 0.7 | 0.7 | 5227 |
|
| 121 |
|
| 122 |
## Inference platform
|
| 123 |
This model is used by the [CAP Babel Machine](https://babel.poltextlab.com), an open-source and free natural language processing tool, designed to simplify and speed up projects for comparative research.
|
|
|
|
| 3 |
---
|
| 4 |
license: mit
|
| 5 |
language:
|
| 6 |
+
- de
|
| 7 |
tags:
|
| 8 |
- zero-shot-classification
|
| 9 |
- text-classification
|
|
|
|
| 14 |
---
|
| 15 |
# xlm-roberta-large-german-cap
|
| 16 |
## Model description
|
| 17 |
+
An `xlm-roberta-large` model finetuned on german training data labelled with [major topic codes](https://www.comparativeagendas.net/pages/master-codebook) from the [Comparative Agendas Project](https://www.comparativeagendas.net/).
|
| 18 |
|
| 19 |
## How to use the model
|
| 20 |
#### Loading and tokenizing input data
|
|
|
|
| 90 |
We also incorporated an EarlyStoppingCallback in the process with a patience of 2 epochs.
|
| 91 |
|
| 92 |
## Model performance
|
| 93 |
+
The model was evaluated on a test set of 601 examples (10% of the available data).<br>
|
| 94 |
+
Model accuracy is **0.71**.
|
| 95 |
| label | precision | recall | f1-score | support |
|
| 96 |
|:-------------|------------:|---------:|-----------:|----------:|
|
| 97 |
+
| 0 | 0.92 | 0.65 | 0.76 | 17 |
|
| 98 |
+
| 1 | 0.77 | 0.84 | 0.8 | 61 |
|
| 99 |
+
| 2 | 0.76 | 0.76 | 0.76 | 21 |
|
| 100 |
+
| 3 | 0.82 | 0.82 | 0.82 | 17 |
|
| 101 |
+
| 4 | 0.61 | 0.56 | 0.58 | 25 |
|
| 102 |
+
| 5 | 0.5 | 0.56 | 0.53 | 9 |
|
| 103 |
+
| 6 | 0.71 | 0.76 | 0.74 | 46 |
|
| 104 |
+
| 7 | 0.62 | 0.83 | 0.71 | 24 |
|
| 105 |
+
| 8 | 0.57 | 0.77 | 0.65 | 22 |
|
| 106 |
+
| 9 | 0.85 | 0.85 | 0.85 | 47 |
|
| 107 |
+
| 10 | 0.71 | 0.61 | 0.66 | 36 |
|
| 108 |
+
| 11 | 0.36 | 0.36 | 0.36 | 14 |
|
| 109 |
+
| 12 | 0.73 | 0.89 | 0.8 | 9 |
|
| 110 |
+
| 13 | 0.59 | 0.62 | 0.6 | 21 |
|
| 111 |
+
| 14 | 0.76 | 0.83 | 0.8 | 66 |
|
| 112 |
+
| 15 | 0.56 | 0.56 | 0.56 | 16 |
|
| 113 |
+
| 16 | 1 | 0.14 | 0.25 | 7 |
|
| 114 |
+
| 17 | 0.75 | 0.72 | 0.73 | 61 |
|
| 115 |
+
| 18 | 0.72 | 0.64 | 0.68 | 77 |
|
| 116 |
+
| 19 | 0 | 0 | 0 | 3 |
|
| 117 |
+
| 20 | 0 | 0 | 0 | 2 |
|
| 118 |
+
| macro avg | 0.63 | 0.61 | 0.6 | 601 |
|
| 119 |
+
| weighted avg | 0.71 | 0.71 | 0.71 | 601 |
|
|
|
|
| 120 |
|
| 121 |
## Inference platform
|
| 122 |
This model is used by the [CAP Babel Machine](https://babel.poltextlab.com), an open-source and free natural language processing tool, designed to simplify and speed up projects for comparative research.
|