FS-distilroberta-fine-tuned

FinScience/FS-distilroberta-fine-tuned

FS-distilroberta-fine-tuned is a text classification model from FinScience. Use it when you need a label for a piece of text. It is set up for transformers.

Downloads · 30 days

0% of all-time downloads

All-time downloads

18K

Public

Repo size

657 MB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.bin329 MB · 99%

At a glance

Task: Text Classification
Library: transformers
Model type: roberta
Access: Public
Created: Mar 7, 2022
Updated: Mar 7, 2022
SHA: 559b00d9

Try a prompt

Model card

From the Hugging Face model README

FS-distilroberta-fine-tuned

The model was obtained by fine-tuning "mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis" model for sentiment analysis on financial news gathered by FinScience software. It predicts the sentiment of news with one label ("negative", "neutral" or "positive"). At the moment, the models works only in English.

Training data

The training dataset consists of 2558 titles of news that were manually labelled by FinScience Team using doccano tool. A "neutral" label was assigned to those news for which an agreement was not reached. 70% (1790 news) of such dataset was employed as training set, while 15% (384) as validation set and the remaining 15% as test set. F1-score (macro) was selected as the evaluation metric.

Set	Number of news	Scope
Training	1790	Training the model
Validation	384	Selecting the hyper-parameters
Test	384	Evaluating the performance

Accuracy

The table below summarizes the performance of the models that were tested on the same test set, consisting of 384 held-out titles:

Language	Accuracy	F1-score (macro)
FS-distilroberta-fine-tuned	76%	76%

Task: Text Classification
Library: transformers
Type: roberta
Languages: en
Created: Mar 7, 2022
Updated: Mar 7, 2022