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 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: sentence-transformers/all-mpnet-base-v2
Classification head: a LogisticRegression instance
spaCy Model: en_core_web_sm
SetFitABSA Aspect Model: joshuasundance/setfit-absa-all-MiniLM-L6-v2-laptops-aspect
SetFitABSA Polarity Model: joshuasundance/setfit-absa-all-mpnet-base-v2-laptops-polarity
Maximum Sequence Length: 384 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
neutral	'skip taking the cord with me because:I charge it at night and skip taking the cord with me because of the good battery life.' 'The tech guy then said the:The tech guy then said the service center does not do 1-to-1 exchange and I have to direct my concern to the "sales" team, which is the retail shop which I bought my netbook from.' 'all dark, power light steady, hard:\xa0One night I turned the freaking thing off after using it, the next day I turn it on, no GUI, screen all dark, power light steady, hard drive light steady and not flashing as it usually does.'
positive	'of the good battery life.:I charge it at night and skip taking the cord with me because of the good battery life.' 'is of high quality, has a:it is of high quality, has a killer GUI, is extremely stable, is highly expandable, is bundled with lots of very good applications, is easy to use, and is absolutely gorgeous.' 'has a killer GUI, is extremely:it is of high quality, has a killer GUI, is extremely stable, is highly expandable, is bundled with lots of very good applications, is easy to use, and is absolutely gorgeous.'
negative	'then said the service center does not do:The tech guy then said the service center does not do 1-to-1 exchange and I have to direct my concern to the "sales" team, which is the retail shop which I bought my netbook from.' 'concern to the "sales" team, which is:The tech guy then said the service center does not do 1-to-1 exchange and I have to direct my concern to the "sales" team, which is the retail shop which I bought my netbook from.' 'on, no GUI, screen all:\xa0One night I turned the freaking thing off after using it, the next day I turn it on, no GUI, screen all dark, power light steady, hard drive light steady and not flashing as it usually does.'
conflict	'-No backlit keyboard, but not:-No backlit keyboard, but not an issue for me.' "to replace the battery once, but:I did have to replace the battery once, but that was only a couple months ago and it's been working perfect ever since."

Evaluation

Metrics

Label	Accuracy
all	0.7008

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

python

from setfit import AbsaModel

# Download from the 🤗 Hub
model = AbsaModel.from_pretrained(
    "joshuasundance/setfit-absa-all-MiniLM-L6-v2-laptops-aspect",
    "joshuasundance/setfit-absa-all-mpnet-base-v2-laptops-polarity",
    spacy_model="en_core_web_sm",
)
# Run inference
preds = model("This laptop meets every expectation and Windows 7 is great!")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	3	25.5873	48

Label	Training Sample Count
conflict	2
negative	45
neutral	30
positive	49

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
0.0120	1	0.2721	-
0.6024	50	0.0894	0.2059
1.2048	100	0.0014	0.2309
1.8072	150	0.0006	0.2359
2.4096	200	0.0005	0.2373
3.0120	250	0.0004	0.2364
3.6145	300	0.0003	0.2371

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.11.7
SetFit: 1.0.3
Sentence Transformers: 2.3.0
spaCy: 3.7.2
Transformers: 4.37.2
PyTorch: 2.1.2+cu118
Datasets: 2.16.1
Tokenizers: 0.15.1

Citation

BibTeX

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}
}