Finnish-actions/SetFit-FinBERT1-A3-denial

Detect Denials in Asynchronous Conversation Comments

SetFit with TurkuNLP/bert-base-finnish-cased-v1

This is a SetFit model that can be used for Text Classification of actions in asynchronous conversation. This particular model detects if a comment includes a denial or not. The configuration of the model is that the model is based on only one annotator's annotations (annotator A3). Metric evaluations are based on conservative ground truth (see paper).

This SetFit model uses TurkuNLP/bert-base-finnish-cased-v1 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification.

Model Details

Model Description

• Model Type: SetFit
• Sentence Transformer body: TurkuNLP/bert-base-finnish-cased-v1
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 2 classes
• Language: Finnish

Model Sources

• Repository: https://github.com/henniina/Detecting-paired-actions
• Paper: Paakki, H., Toivanen, P. and Kajava K. (forthcoming). Implicit and Indirect: Detecting Face-threatening and Paired Actions in Asynchronous Online Conversations. Northern European Journal of Language Technology (NEJLT), 0(0).

Model Labels

Label	Examples
0	'Etunimi Sukunimi jep näin tulee käymään.' 'Joskus sitä lähdettiin Saksan kanssa idän maille ihmeisiin.Nyt ainakin Venäjä on niin heikko että se helähtää vain kun ollaan Uralilla yhdessä USAn kanssa.Venäjä on nujerettava.Kaverina USA sillä on kokemusta valtioiden valloituksista hallinon vaihtanisesta ja Putini kans vaihdetaan ja tähtilippu Kremlin torniin mostetaan!' 'No jos haluaa lapsensa kotiin jättää, niin luultavasti se lupa joko lomaan tai kotiopetukseen tulee koululta. Oman lapsen kohdalla riitti rehtorille ilmoitus ja kaikki oli kunnossa ettei nuori tule kouluun. Tehtävät tulee kotiin, niin kuin esim. lomalle jäädessä.'
1	'Etunimi ei, kyllä ne rahat löytää ennemmin Afrikan savanneille... EI Suomen vähäosaisten hyväksi... ikävää mutta totta... Suomen kansan on tingittävä menoistaan 🤣' 'Etunimi Sukunimi ei kun pysyy omalla puolella mitä venäjä ei osaa. Loput on saanut kyllä lukea lehdistä tässä jos semmoisia luet.' 'Ei tullut vielä kuopattua, vastaus oli muihin kommentteihin lähinnä sillä tasolla, että odotetaan mitä tapahtuu, vaikka parantuisi. Täällä on kommentteja kiukkunaamoista lähtien, eli herätys👍'

Evaluation

Metrics

Label	Metric
all	0.9056

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

from setfit import SetFitModel

# Download from the 🤗 Hub
model = SetFitModel.from_pretrained("Finnish-actions/SetFit-FinBERT1-A3-denial")
# Run inference
preds = model("Mitä yrität sanoa?")

Downstream Use

NB. This model has been trained on data coming from Finnish language asynchronous conversations under crisis related news on Facebook. This specific model has been trained to detect whether a comment includes a question or not. It reflects only one of our annotators' label interpretations, so the best use of our models (see our paper) would be to combine a set of models we provide on our Huggingface (Finnish-actions), and use a model ensemble to provide label predictions. It needs to be noted also that the model may not be well applicable outside of its empirical context, so in downstream applications, one should always conduct an evaluation of the model applicability using manually annotated data from that specific context (see our paper for annotation instructions).

Out-of-Scope Use

Please use this model only for action detection and analysis. Uses of this model and the involved data for generative purposes (e.g. NLG) is prohibited.

Bias, Risks and Limitations

Note that the model may produce errors. Due to the size of the training dataset, model may not generalize very well even for other novel topics within the same context. Note that model predictions should not be regarded as final judgments e.g. for online moderation purposes, but each case should also be regarded individually if using model predictions to support moderation. Also, the annotations only reflect three (though experienced) annotators' interpretations, so there might be perspectives on data intepretation that have not been taken into account here.

If model is used to support moderation on social media, we recommend that final judgments should always be left for human moderators.

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	1	19.9605	213

Label	Training Sample Count
0	890
1	73

Framework Versions

• Python: 3.11.9
• SetFit: 1.1.3
• Sentence Transformers: 3.2.0
• Transformers: 4.44.0
• PyTorch: 2.4.0+cu124
• Datasets: 2.21.0
• Tokenizers: 0.19.1

Citation

If you use this model, please cite the following work:

BibTeX

@article{paakki-implicit-indirect,
    doi = {forthcoming},
    url = {forthcoming},
    author = {Paakki, Henna and Toivanen, Pihla and Kajava, Kaisla},
    title = {Implicit and Indirect: Detecting Face-threatening and Paired Actions in Asynchronous Online Conversations},
    publisher = {Northern European Journal of Language Technology},
    year = {2025}
}