SetFit Polarity Model with BAAI/bge-small-en-v1.5

This is a SetFit model that can be used for Aspect Based Sentiment Analysis (ABSA). This SetFit model uses BAAI/bge-small-en-v1.5 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification. In particular, this model is in charge of classifying aspect polarities.

The model has been trained using an efficient few-shot learning technique that involves:

Fine-tuning a Sentence Transformer with contrastive learning.
Training a classification head with features from the fine-tuned Sentence Transformer.

This model was trained within the context of a larger system for ABSA, which looks like so:

Use a spaCy model to select possible aspect span candidates.
Use a SetFit model to filter these possible aspect span candidates.
Use this SetFit model to classify the filtered aspect span candidates.

Model Details

Model Description

Model Type: SetFit
Sentence Transformer body: BAAI/bge-small-en-v1.5
Classification head: a LogisticRegression instance
spaCy Model: en_core_web_lg
SetFitABSA Aspect Model: asadnaqvi/setfitabsa-aspect
SetFitABSA Polarity Model: asadnaqvi/setfitabsa-polarity
Maximum Sequence Length: 512 tokens
Number of Classes: 4 classes

Model Sources

Repository: SetFit on GitHub
Paper: Efficient Few-Shot Learning Without Prompts
Blogpost: SetFit: Efficient Few-Shot Learning Without Prompts

Model Labels

Label	Examples
Informative	"The upcoming visit of Saudi Arabia:The upcoming visit of Saudi Arabia's crown prince Mohammed bin Salman (MBS) to India is not a routine affair." "'s crown prince Mohammed bin Salman (MBS):The upcoming visit of Saudi Arabia's crown prince Mohammed bin Salman (MBS) to India is not a routine affair." 'to burnish his legitimacy after the international:The trip to India is evidently timed to burnish his legitimacy after the international opprobrium that followed the murder of The Washington Post journalist Jamal Khashoggi.'
Negative	'that followed the murder of The Washington:The trip to India is evidently timed to burnish his legitimacy after the international opprobrium that followed the murder of The Washington Post journalist Jamal Khashoggi.' "Arabia's disastrous military intervention in Yemen or:India for its part has refrained from even a hint of disapproval of Saudi Arabia's disastrous military intervention in Yemen or its misguided attempts to isolate Qatar, never mind the brutal assassination of Khashoggi." 'condemn the Soviet invasion but privately urged:India sought to adopt a more nuanced stance; it did not openly condemn the Soviet invasion but privately urged Moscow to pull back.'
Positive	"in fostering stronger relations with countries in:Prime Minister Narendra Modi's government has invested considerable time and energy in fostering stronger relations with countries in West Asia." "has invested considerable time and energy in fostering stronger:Prime Minister Narendra Modi's government has invested considerable time and energy in fostering stronger relations with countries in West Asia." "security and economic ties with Saudi Arabia:Modi's visit to Riyadh in 2016 gave a fillip to security and economic ties with Saudi Arabia."
Ambivalent	"a hint of disapproval of Saudi Arabia:India for its part has refrained from even a hint of disapproval of Saudi Arabia's disastrous military intervention in Yemen or its misguided attempts to isolate Qatar, never mind the brutal assassination of Khashoggi."

Evaluation

Metrics

Label	Accuracy
all	0.7065

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 AbsaModel

# Download from the 🤗 Hub
model = AbsaModel.from_pretrained(
    "asadnaqvi/setfitabsa-aspect",
    "asadnaqvi/setfitabsa-polarity",
)
# Run inference
preds = model("The food was great, but the venue is just way too busy.")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	11	27.7071	45

Label	Training Sample Count
Ambivalent	1
Informative	73
Negative	20
Positive	5

Training Hyperparameters

batch_size: (128, 128)
num_epochs: (5, 5)
max_steps: -1
sampling_strategy: oversampling
body_learning_rate: (2e-05, 1e-05)
head_learning_rate: 0.01
loss: CosineSimilarityLoss
distance_metric: cosine_distance
margin: 0.25
end_to_end: False
use_amp: True
warmup_proportion: 0.1
seed: 42
eval_max_steps: -1
load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0217	1	0.2599	-
1.0870	50	0.0608	0.3526
2.1739	100	0.0253	0.4091
3.2609	150	0.0159	0.4497
4.3478	200	0.0035	0.4437

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.12
SetFit: 1.0.3
Sentence Transformers: 2.7.0
spaCy: 3.7.4
Transformers: 4.40.1
PyTorch: 2.2.1+cu121
Datasets: 2.19.0
Tokenizers: 0.19.1

Citation

BibTeX

@article{https://doi.org/10.48550/arxiv.2209.11055,
    doi = {10.48550/ARXIV.2209.11055},
    url = {https://arxiv.org/abs/2209.11055},
    author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren},
    keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
    title = {Efficient Few-Shot Learning Without Prompts},
    publisher = {arXiv},
    year = {2022},
    copyright = {Creative Commons Attribution 4.0 International}
}

asadnaqvi
/

setfitabsa-polarity