SetFit Polarity Model

This is a SetFit model that can be used for Aspect Based Sentiment Analysis (ABSA). 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
Classification head: a LogisticRegression instance
spaCy Model: id_core_news_trf
SetFitABSA Aspect Model: jetri20/ABSA_review_game_geometry-aspect
SetFitABSA Polarity Model: jetri20/ABSA_review_game_geometry-polarity
Maximum Sequence Length: 8192 tokens
Number of Classes: 2 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
negatif	'geometri dash kebanyakan level nya sulit mah:sih game level nya sulit banget level clutter funk susah nya ampun level 11 nya level sulit nya level 10 sulit menyerah sabar banting hp saking sulit nya ya geometri dash kebanyakan level nya sulit mah geometri dash game stress saking susah nya' 'emang sih level iklan muncul masuk level:iklan emang sih level iklan muncul masuk level mending kayak mah' 'game apasih sampe strees:game apasih sampe strees gitu final boss level the tower susahnya ampun kenapasih kalo hijau pas udah nya jatuh mati sih cube nya cuman 1 hp ya ngeselin sih'
positif	'game bagus tolong kasi:game bagus tolong kasi iklan' 'game bagus lemot level:game bagus lemot level dikit kedepan nya perbaiki kalo gk ku hapus karna lemot' 'gamenya seru susah banget:gamenya seru susah banget tower'

Label

Examples

negatif

'geometri dash kebanyakan level nya sulit mah:sih game level nya sulit banget level clutter funk susah nya ampun level 11 nya level sulit nya level 10 sulit menyerah sabar banting hp saking sulit nya ya geometri dash kebanyakan level nya sulit mah geometri dash game stress saking susah nya'
'emang sih level iklan muncul masuk level:iklan emang sih level iklan muncul masuk level mending kayak mah'
'game apasih sampe strees:game apasih sampe strees gitu final boss level the tower susahnya ampun kenapasih kalo hijau pas udah nya jatuh mati sih cube nya cuman 1 hp ya ngeselin sih'

positif

'game bagus tolong kasi:game bagus tolong kasi iklan'
'game bagus lemot level:game bagus lemot level dikit kedepan nya perbaiki kalo gk ku hapus karna lemot'
'gamenya seru susah banget:gamenya seru susah banget tower'

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(
    "jetri20/ABSA_review_game_geometry-aspect",
    "jetri20/ABSA_review_game_geometry-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	3	20.6854	70

Label	Training Sample Count
konflik	0
negatif	173
netral	0
positif	148

Training Hyperparameters

batch_size: (4, 4)
num_epochs: (1, 1)
max_steps: -1
sampling_strategy: oversampling
num_iterations: 5
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: False
warmup_proportion: 0.1
seed: 42
eval_max_steps: -1
load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0012	1	0.2078	-
0.0623	50	0.4009	-
0.1245	100	0.0204	-
0.1868	150	0.0249	-
0.2491	200	0.0238	-
0.3113	250	0.016	-
0.3736	300	0.0114	-
0.4359	350	0.2153	-
0.4981	400	0.0032	-
0.5604	450	0.004	-
0.6227	500	0.0022	-
0.6849	550	0.2173	-
0.7472	600	0.0019	-
0.8095	650	0.0007	-
0.8717	700	0.0014	-
0.9340	750	0.0007	-
0.9963	800	0.0012	-
1.0	803	-	0.338

Framework Versions

Python: 3.10.13
SetFit: 1.0.3
Sentence Transformers: 3.0.1
spaCy: 3.7.5
Transformers: 4.36.2
PyTorch: 2.1.2
Datasets: 2.19.2
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}
}