najwaa/absa-combined-p2-aspect

SetFit Aspect Model with sentence-transformers/all-MiniLM-L6-v2

This is a SetFit model that can be used for Aspect Based Sentiment Analysis (ABSA). This SetFit model uses sentence-transformers/all-MiniLM-L6-v2 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification. In particular, this model is in charge of filtering aspect span candidates.

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

Model Details

Model Description

• Model Type: SetFit
• Sentence Transformer body: sentence-transformers/all-MiniLM-L6-v2
• Classification head: a LogisticRegression instance
• spaCy Model: en_core_web_sm
• SetFitABSA Aspect Model: najwaa/absa-combined-p2-aspect
• SetFitABSA Polarity Model: najwaa/absa-combined-p2-polarity
• Maximum Sequence Length: 256 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
aspect	'light weight:this computer is so light weight and easy to carry.' 'carry:this computer is so light weight and easy to carry.' 'lightweight:very lightweight.'
no aspect	'computer:this computer is so light weight and easy to carry.' "premium:The build quality feels premium but it's surprisingly heavy for daily commuting." "commuting:The build quality feels premium but it's surprisingly heavy for daily commuting."

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(
    "najwaa/absa-combined-p2-aspect",
    "najwaa/absa-combined-p2-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	2	17.9094	52

Label	Training Sample Count
no aspect	524
aspect	480

Training Hyperparameters

• batch_size: (32, 32)
• 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
• l2_weight: 0.01
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: True

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0001	1	0.3707	-
0.0006	10	-	0.3251
0.0013	20	-	0.3248
0.0019	30	-	0.3243
0.0025	40	-	0.3235
0.0032	50	0.3447	0.3225
0.0038	60	-	0.3213
0.0044	70	-	0.3199
0.0051	80	-	0.3182
0.0057	90	-	0.3164
0.0063	100	0.3396	0.3144
0.0070	110	-	0.3122
0.0076	120	-	0.3098
0.0082	130	-	0.3074
0.0089	140	-	0.3049
0.0095	150	0.3198	0.3022
0.0101	160	-	0.2991
0.0108	170	-	0.2960
0.0114	180	-	0.2928
0.0120	190	-	0.2894
0.0126	200	0.3344	0.2860
0.0133	210	-	0.2826
0.0139	220	-	0.2797
0.0145	230	-	0.2767
0.0152	240	-	0.2738
0.0158	250	0.2961	0.2712
0.0164	260	-	0.2696
0.0171	270	-	0.2679
0.0177	280	-	0.2661
0.0183	290	-	0.2642
0.0190	300	0.2741	0.2625
0.0196	310	-	0.2609
0.0202	320	-	0.2598
0.0209	330	-	0.2592
0.0215	340	-	0.2587
0.0221	350	0.2744	0.2584
0.0228	360	-	0.2582
0.0234	370	-	0.2582
0.0240	380	-	0.2584
0.0247	390	-	0.2583
0.0253	400	0.2679	0.2583
0.0259	410	-	0.2584
0.0266	420	-	0.2584

Framework Versions

• Python: 3.11.12
• SetFit: 1.1.2
• Sentence Transformers: 4.1.0
• spaCy: 3.7.5
• Transformers: 4.51.3
• PyTorch: 2.6.0+cu124
• Datasets: 3.6.0
• Tokenizers: 0.21.1

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}
}