Update README.md
Browse files
README.md
CHANGED
|
@@ -1,19 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
# News Relevancy Classifiers
|
| 2 |
|
| 3 |
## FinBERT-ft-v3
|
| 4 |
|
| 5 |
-
 manually labeled into:
|
| 12 |
- 0 — Not Relevant
|
| 13 |
- 1 — Least Relevant
|
| 14 |
- 2 — Highly Relevant
|
| 15 |
- 3 — Most Relevant
|
| 16 |
-
- **HF Repo**: [`cloud0day3/finbert-ft-v3`](https://huggingface.co/cloud0day3/finbert-ft-v3) (latest v4 checkpoint)
|
| 17 |
- **Date Trained**: 2025-06-06
|
| 18 |
|
| 19 |
#### Model Inputs
|
|
@@ -31,3 +48,24 @@
|
|
| 31 |
2: "Highly Relevant",
|
| 32 |
3: "Most Relevant"
|
| 33 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language:
|
| 4 |
+
- fi
|
| 5 |
+
metrics:
|
| 6 |
+
- f1
|
| 7 |
+
- precision
|
| 8 |
+
- recall
|
| 9 |
+
- accuracy
|
| 10 |
+
base_model:
|
| 11 |
+
- google-bert/bert-base-uncased
|
| 12 |
+
- TurkuNLP/bert-base-finnish-cased-v1
|
| 13 |
+
pipeline_tag: text-classification
|
| 14 |
+
tags:
|
| 15 |
+
- classification
|
| 16 |
+
- news
|
| 17 |
+
---
|
| 18 |
# News Relevancy Classifiers
|
| 19 |
|
| 20 |
## FinBERT-ft-v3
|
| 21 |
|
| 22 |
+
|
| 23 |
|
| 24 |
### Model Description
|
| 25 |
+
- **Purpose**: This model is trained for a specific task in research, it is not a commmercial product and should not be used in for-profit.
|
| 26 |
- **Architecture**: `bert-base-finnish-cased-v1`
|
| 27 |
+
- **Fine-tuning task**: Four-class Finnish news-headline relevancy classification
|
| 28 |
- **Dataset**: ~225 Finnish headlines (2024–2025) manually labeled into:
|
| 29 |
- 0 — Not Relevant
|
| 30 |
- 1 — Least Relevant
|
| 31 |
- 2 — Highly Relevant
|
| 32 |
- 3 — Most Relevant
|
| 33 |
+
- **HF Repo**: [`cloud0day3/finbert-ft-v3`](https://huggingface.co/cloud0day3/finbert-ft-v3) (latest v4 checkpoint, 6 June 2025)
|
| 34 |
- **Date Trained**: 2025-06-06
|
| 35 |
|
| 36 |
#### Model Inputs
|
|
|
|
| 48 |
2: "Highly Relevant",
|
| 49 |
3: "Most Relevant"
|
| 50 |
}
|
| 51 |
+
|
| 52 |
+
|
| 53 |
+
#### Intended Use
|
| 54 |
+
- **Primary**: Automatically assign a relevancy score to Finnish news headlines so that downstream pipelines (e.g., filtering, ranking) can operate without manual triage.
|
| 55 |
+
|
| 56 |
+
#### Examples of use:
|
| 57 |
+
|
| 58 |
+
- Pre-filtering a news aggregation feed.
|
| 59 |
+
|
| 60 |
+
- Prioritizing headlines for editorial review.
|
| 61 |
+
|
| 62 |
+
- Input to summarization/retrieval pipelines.
|
| 63 |
+
|
| 64 |
+
#### Out-of-Scope Uses
|
| 65 |
+
- Any non-Finnish text (e.g., English, Swedish).
|
| 66 |
+
|
| 67 |
+
- Multi-sentence inputs or full articles (this model is tuned on single-sentence headlines).
|
| 68 |
+
|
| 69 |
+
- Tasks other than relevancy (e.g., sentiment analysis, topic modeling).
|
| 70 |
+
|
| 71 |
+
- High-risk decision making without human oversight (e.g., emergency alerts).
|