MattiaTintori/ABSA_Polarity_EN

SetFit Polarity Model with sentence-transformers/all-mpnet-base-v2

This is a SetFit model that can be used for Aspect Based Sentiment Analysis (ABSA). This SetFit model uses sentence-transformers/all-mpnet-base-v2 as the Sentence Transformer embedding model. A SetFitHead 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:

1. Fine-tuning a Sentence Transformer with contrastive learning.
2. 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:

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

Model Details

Model Description

• Model Type: SetFit
• Sentence Transformer body: sentence-transformers/all-mpnet-base-v2
• Classification head: a SetFitHead instance
• spaCy Model: en_core_web_trf
• SetFitABSA Aspect Model: setfit-absa-aspect
• SetFitABSA Polarity Model: MattiaTintori/Final_polarity_Colab
• Maximum Sequence Length: 384 tokens
• Number of Classes: 3 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
1	'waiter) We got no cheese offered for the pasta,:(food was delivered by a busboy, not waiter) We got no cheese offered for the pasta, our water and wine glasses remained EMPTY our entire meal, when we would have easily spent another $20 on wine.' 'by a busboy, not waiter) We got no cheese:(food was delivered by a busboy, not waiter) We got no cheese offered for the pasta, our water and wine glasses remained EMPTY our entire meal, when we would have easily spent another $20 on wine.' 'for the pasta, our water and wine glasses remained EMPTY our entire meal:(food was delivered by a busboy, not waiter) We got no cheese offered for the pasta, our water and wine glasses remained EMPTY our entire meal, when we would have easily spent another $20 on wine.'
2	'(food was delivered by a busboy:(food was delivered by a busboy, not waiter) We got no cheese offered for the pasta, our water and wine glasses remained EMPTY our entire meal, when we would have easily spent another $20 on wine.' 'glasses remained EMPTY our entire meal, when we would have:(food was delivered by a busboy, not waiter) We got no cheese offered for the pasta, our water and wine glasses remained EMPTY our entire meal, when we would have easily spent another $20 on wine.' 'spent another $20 on wine.:(food was delivered by a busboy, not waiter) We got no cheese offered for the pasta, our water and wine glasses remained EMPTY our entire meal, when we would have easily spent another $20 on wine.'
0	'few cocktails and enjoy our surroundings and each other.:20 minutes for our reservation but it gave us time to have a few cocktails and enjoy our surroundings and each other.' 'Barbecued codfish was gorgeously moist - as:Barbecued codfish was gorgeously moist - as if poached - yet the fabulous texture was let down by curiously bland seasoning - a spice rub might have overwhelmed, however herb mix or other sauce would have done much to enhance.' 'Even though its good seafood, the prices are too:Even though its good seafood, the prices are too high.'

Evaluation

Metrics

Label	F1
all	0.8170

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(
    "setfit-absa-aspect",
    "MattiaTintori/Final_polarity_Colab",
)
# 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	1	25.0463	79

Label	Training Sample Count
0	1148
1	607
2	489

Training Hyperparameters

• batch_size: (64, 4)
• num_epochs: (5, 32)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 10
• body_learning_rate: (5e-05, 5e-05)
• head_learning_rate: 0.04
• loss: CosineSimilarityLoss
• distance_metric: cosine_distance
• margin: 0.25
• end_to_end: False
• use_amp: True
• warmup_proportion: 0.1
• l2_weight: 0.01
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0014	1	0.3084	-
0.0285	20	0.2735	0.2591
0.0570	40	0.2228	0.2351
0.0855	60	0.2071	0.1993
0.1140	80	0.1522	0.1696
0.1425	100	0.1441	0.1671
0.1709	120	0.1632	0.161
0.1994	140	0.0966	0.1575
0.2279	160	0.1737	0.1504
0.2564	180	0.1092	0.1671
0.2849	200	0.1314	0.1459
0.3134	220	0.0972	0.1483
0.3419	240	0.1014	0.1537
0.3704	260	0.0506	0.1514
0.3989	280	0.0817	0.143
0.4274	300	0.0592	0.1526
0.4558	320	0.0311	0.1562
0.4843	340	0.038	0.1546
0.5128	360	0.0852	0.1497
0.5413	380	0.0359	0.144
0.5698	400	0.0449	0.1639
0.5983	420	0.0314	0.1517

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.12
• SetFit: 1.0.3
• Sentence Transformers: 3.0.1
• spaCy: 3.7.6
• Transformers: 4.39.0
• PyTorch: 2.4.0+cu121
• Datasets: 2.21.0
• Tokenizers: 0.15.2

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}
}