Hugging Face's logo

cccornflake
/
modu_absa-aspect

Text Classification
setfit
Safetensors
sentence-transformers
bert
absa
generated_from_setfit_trainer
Model card Files Community
Edit model card

SetFit Aspect Model with snunlp/KR-SBERT-V40K-klueNLI-augSTS

This is a SetFit model that can be used for Aspect Based Sentiment Analysis (ABSA). This SetFit model uses snunlp/KR-SBERT-V40K-klueNLI-augSTS 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 Sources

Model Labels

Label Examples
no aspect
  • '둘쨋날은 미친듯이 밟아봤더니 기어가 헛돌면서:둘쨋날은 미친듯이 밟아봤더니 기어가 헛돌면서 틱틱 소리가 나서 경악.'
  • '소리가 나서 경악:둘쨋날은 미친듯이 밟아봤더니 기어가 헛돌면서 틱틱 소리가 나서 경악.'
  • '삐꾸를:이거 뭐 삐꾸를 준 거 아냐 불안하고, 거금 투자한 게 왜 이래.. 싶어서 정이 확 떨어졌는데 산 곳 가져가서 확인하니 기어 텐션 문제라고 고장 아니래.'
aspect
  • '기어 텐션:이거 뭐 삐꾸를 준 거 아냐 불안하고, 거금 투자한 게 왜 이래.. 싶어서 정이 확 떨어졌는데 산 곳 가져가서 확인하니 기어 텐션 문제라고 고장 아니래.'
  • '동영상 재생하면서 자막 중 모르는 내용 있으면 터치해서 바로 검색하는 기능:동영상 재생하면서 자막 중 모르는 내용 있으면 터치해서 바로 검색하는 기능 때문에 산 건데 이게 에러다..'
  • '118g:스펙상으로는 116g인데 집에서 저울로 재어보니 118g, 실제로 들어보니 돌덩이 같다.'

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(
    "cccornflake/modu_absa-aspect",
    "cccornflake/modu_absa-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 1 12.3710 45
Label Training Sample Count
no aspect 2534
aspect 806

Training Hyperparameters

  • batch_size: (4, 4)
  • num_epochs: (1, 1)
  • max_steps: -1
  • sampling_strategy: oversampling
  • num_iterations: 20
  • 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
0.0029 1 0.144 -
0.1471 50 0.1638 -
0.2941 100 0.0111 -
0.4412 150 0.0003 -
0.5882 200 0.0001 -
0.7353 250 0.0002 -
0.8824 300 0.0001 -
0.0017 1 0.1847 -
0.0833 50 0.0009 -
0.1667 100 0.0001 -
0.25 150 0.0001 -
0.3333 200 0.0002 -
0.4167 250 0.0 -
0.5 300 0.0 -
0.5833 350 0.0 -
0.6667 400 0.0 -
0.75 450 0.0 -
0.8333 500 0.0 -
0.9167 550 0.0001 -
1.0 600 0.0 -
0.0000 1 0.0112 -
0.0015 50 0.4302 -
0.0030 100 0.0006 -
0.0045 150 0.2262 -
0.0060 200 0.3926 0.1537
0.0075 250 0.1391 -
0.0090 300 0.0051 -
0.0105 350 0.018 -
0.0120 400 0.2427 0.1252
0.0135 450 0.0004 -
0.0150 500 0.0006 -
0.0165 550 0.0027 -
0.0180 600 0.021 0.1186
0.0195 650 0.0092 -
0.0210 700 0.2152 -
0.0225 750 0.2208 -
0.0240 800 0.1252 0.0971
0.0254 850 0.0891 -
0.0269 900 0.0007 -
0.0284 950 0.2289 -
0.0299 1000 0.2675 0.1191
0.0314 1050 0.0022 -
0.0329 1100 0.0027 -
0.0344 1150 0.0004 -
0.0359 1200 0.0323 0.1083
0.0374 1250 0.2118 -
0.0389 1300 0.0018 -
0.0404 1350 0.0001 -
0.0419 1400 0.2381 0.0795
0.0434 1450 0.0004 -
0.0449 1500 0.0 -
0.0464 1550 0.0001 -
0.0479 1600 0.0003 0.0807
0.0494 1650 0.0017 -
0.0509 1700 0.0018 -
0.0524 1750 0.0017 -
0.0539 1800 0.0003 0.0962
0.0554 1850 0.3088 -
0.0569 1900 0.234 -
0.0584 1950 0.0015 -
0.0599 2000 0.0015 0.0872
0.0614 2050 0.0009 -
0.0629 2100 0.0003 -
0.0644 2150 0.0002 -
0.0659 2200 0.0058 0.0811
0.0674 2250 0.0001 -
0.0689 2300 0.122 -
0.0704 2350 0.2157 -
0.0719 2400 0.0001 0.0944
0.0734 2450 0.0 -
0.0749 2500 0.0002 -
0.0763 2550 0.0007 -
0.0778 2600 0.0002 0.0965
0.0793 2650 0.0001 -
0.0808 2700 0.0002 -
0.0823 2750 0.0008 -
0.0838 2800 0.0 0.0954
0.0853 2850 0.0126 -
0.0868 2900 0.0541 -
0.0883 2950 0.0002 -
0.0898 3000 0.0005 0.0841
0.0913 3050 0.0001 -
0.0928 3100 0.1914 -
0.0943 3150 0.0001 -
0.0958 3200 0.0001 0.0815
0.0973 3250 0.0051 -
0.0988 3300 0.0002 -
0.1003 3350 0.0 -
0.1018 3400 0.0001 0.0876
0.1033 3450 0.0 -
0.1048 3500 0.0 -
0.1063 3550 0.2448 -
0.1078 3600 0.0 0.0893
0.1093 3650 0.0 -
0.1108 3700 0.0001 -
0.1123 3750 0.0001 -
0.1138 3800 0.0 0.0883
0.1153 3850 0.0149 -
0.1168 3900 0.0006 -
0.1183 3950 0.0009 -
0.1198 4000 0.0021 0.0963
0.1213 4050 0.0002 -
0.1228 4100 0.0 -
0.1243 4150 0.0 -
0.1257 4200 0.0013 0.1065
0.1272 4250 0.0026 -
0.1287 4300 0.0001 -
0.1302 4350 0.0001 -
0.1317 4400 0.0 0.0845
0.1332 4450 0.0 -
0.1347 4500 0.0008 -
0.1362 4550 0.0003 -
0.1377 4600 0.0006 0.0934
0.1392 4650 0.0005 -
0.1407 4700 0.0006 -
0.1422 4750 0.0001 -
0.1437 4800 0.0 0.1055
0.1452 4850 0.0 -
0.1467 4900 0.0 -
0.1482 4950 0.0 -
0.1497 5000 0.0001 0.097
0.1512 5050 0.0 -
0.1527 5100 0.0001 -
0.1542 5150 0.0 -
0.1557 5200 0.0 0.0961
0.1572 5250 0.2286 -
0.1587 5300 0.0 -
0.1602 5350 0.0008 -
0.1617 5400 0.0 0.0946
0.1632 5450 0.0012 -
0.1647 5500 0.0 -
0.1662 5550 0.0 -
0.1677 5600 0.0 0.0942
0.1692 5650 0.0 -
0.1707 5700 0.0001 -
0.1722 5750 0.0 -
0.1737 5800 0.0 0.0962
0.1751 5850 0.0 -
0.1766 5900 0.0 -
0.1781 5950 0.0 -
0.1796 6000 0.0 0.1079
0.1811 6050 0.0 -
0.1826 6100 0.0011 -
0.1841 6150 0.0 -
0.1856 6200 0.0 0.0939
0.1871 6250 0.0 -
0.1886 6300 0.0 -
0.1901 6350 0.0 -
0.1916 6400 0.0 0.1091
0.1931 6450 0.0017 -
0.1946 6500 0.001 -
0.1961 6550 0.0 -
0.1976 6600 0.0 0.0893
0.1991 6650 0.001 -
0.2006 6700 0.0 -
0.2021 6750 0.0 -
0.2036 6800 0.0004 0.0909
0.2051 6850 0.0019 -
0.2066 6900 0.0006 -
0.2081 6950 0.0 -
0.2096 7000 0.0 0.091
0.2111 7050 0.0 -
0.2126 7100 0.0 -
0.2141 7150 0.0001 -
0.2156 7200 0.0 0.1216
0.2171 7250 0.0 -
0.2186 7300 0.0 -
0.2201 7350 0.0 -
0.2216 7400 0.0 0.0885
0.2231 7450 0.0 -
0.2246 7500 0.0 -
0.2260 7550 0.0 -
0.2275 7600 0.0 0.0921
0.2290 7650 0.0 -
0.2305 7700 0.0 -
0.2320 7750 0.0005 -
0.2335 7800 0.0 0.0933
0.2350 7850 0.0 -
0.2365 7900 0.0 -
0.2380 7950 0.0 -
0.2395 8000 0.0 0.0908
0.2410 8050 0.0 -
0.2425 8100 0.0 -
0.2440 8150 0.0006 -
0.2455 8200 0.0 0.0891
0.2470 8250 0.0 -
0.2485 8300 0.0 -
0.25 8350 0.0 -
0.2515 8400 0.0001 0.0918
0.2530 8450 0.0 -
0.2545 8500 0.0 -
0.2560 8550 0.0 -
0.2575 8600 0.0 0.0797
0.2590 8650 0.0 -
0.2605 8700 0.0 -
0.2620 8750 0.0 -
0.2635 8800 0.0 0.0843
0.2650 8850 0.0 -
0.2665 8900 0.0 -
0.2680 8950 0.0 -
0.2695 9000 0.0 0.0811
0.2710 9050 0.0 -
0.2725 9100 0.0 -
0.2740 9150 0.0 -
0.2754 9200 0.091 0.0858
0.2769 9250 0.0 -
0.2784 9300 0.0 -
0.2799 9350 0.0 -
0.2814 9400 0.0 0.0886
0.2829 9450 0.0 -
0.2844 9500 0.0 -
0.2859 9550 0.0 -
0.2874 9600 0.0 0.09
0.2889 9650 0.0 -
0.2904 9700 0.0 -
0.2919 9750 0.0 -
0.2934 9800 0.0 0.0887
0.2949 9850 0.0 -
0.2964 9900 0.0001 -
0.2979 9950 0.0 -
0.2994 10000 0.0 0.103
0.3009 10050 0.0 -
0.3024 10100 0.0 -
0.3039 10150 0.0 -
0.3054 10200 0.0 0.106
0.3069 10250 0.0 -
0.3084 10300 0.0 -
0.3099 10350 0.0 -
0.3114 10400 0.0 0.0861
0.3129 10450 0.0 -
0.3144 10500 0.0 -
0.3159 10550 0.2242 -
0.3174 10600 0.0 0.0972
0.3189 10650 0.0 -
0.3204 10700 0.0 -
0.3219 10750 0.0 -
0.3234 10800 0.0008 0.0966
0.3249 10850 0.0 -
0.3263 10900 0.0 -
0.3278 10950 0.0 -
0.3293 11000 0.0 0.0924
0.3308 11050 0.0 -
0.3323 11100 0.0 -
0.3338 11150 0.0 -
0.3353 11200 0.0 0.0988
0.3368 11250 0.0 -
0.3383 11300 0.0 -
0.3398 11350 0.0 -
0.3413 11400 0.0 0.096
0.3428 11450 0.0 -
0.3443 11500 0.0 -
0.3458 11550 0.0 -
0.3473 11600 0.0 0.1044
0.3488 11650 0.0 -
0.3503 11700 0.0 -
0.3518 11750 0.0 -
0.3533 11800 0.0 0.093
0.3548 11850 0.0 -
0.3563 11900 0.0 -
0.3578 11950 0.0 -
0.3593 12000 0.0 0.0926
0.3608 12050 0.0 -
0.3623 12100 0.0 -
0.3638 12150 0.0 -
0.3653 12200 0.0 0.092
0.3668 12250 0.0 -
0.3683 12300 0.0 -
0.3698 12350 0.0 -
0.3713 12400 0.0 0.0913
0.3728 12450 0.0 -
0.3743 12500 0.0 -
0.3757 12550 0.0 -
0.3772 12600 0.0 0.092
0.3787 12650 0.0 -
0.3802 12700 0.0 -
0.3817 12750 0.0 -
0.3832 12800 0.0 0.0932
0.3847 12850 0.0 -
0.3862 12900 0.0004 -
0.3877 12950 0.0 -
0.3892 13000 0.0 0.0946
0.3907 13050 0.0 -
0.3922 13100 0.0 -
0.3937 13150 0.0 -
0.3952 13200 0.0 0.0934
0.3967 13250 0.0 -
0.3982 13300 0.0 -
0.3997 13350 0.0 -
0.4012 13400 0.0 0.0926
0.4027 13450 0.0 -
0.4042 13500 0.0 -
0.4057 13550 0.0 -
0.4072 13600 0.0 0.0928
0.4087 13650 0.0 -
0.4102 13700 0.0 -
0.4117 13750 0.0 -
0.4132 13800 0.0 0.0932
0.4147 13850 0.0 -
0.4162 13900 0.0 -
0.4177 13950 0.0 -
0.4192 14000 0.0 0.0929
0.4207 14050 0.0 -
0.4222 14100 0.0 -
0.4237 14150 0.0 -
0.4251 14200 0.0 0.09
0.4266 14250 0.0 -
0.4281 14300 0.0 -
0.4296 14350 0.0 -
0.4311 14400 0.0 0.0891
0.4326 14450 0.0 -
0.4341 14500 0.0 -
0.4356 14550 0.0 -
0.4371 14600 0.0 0.0933
0.4386 14650 0.0 -
0.4401 14700 0.0 -
0.4416 14750 0.0 -
0.4431 14800 0.0 0.0923
0.4446 14850 0.0 -
0.4461 14900 0.0 -
0.4476 14950 0.0 -
0.4491 15000 0.0 0.0929
0.4506 15050 0.0 -
0.4521 15100 0.0012 -
0.4536 15150 0.0 -
0.4551 15200 0.0 0.0913
0.4566 15250 0.0 -
0.4581 15300 0.0 -
0.4596 15350 0.0 -
0.4611 15400 0.0 0.1148
0.4626 15450 0.0 -
0.4641 15500 0.0 -
0.4656 15550 0.0 -
0.4671 15600 0.0 0.1061
0.4686 15650 0.0 -
0.4701 15700 0.0 -
0.4716 15750 0.0 -
0.4731 15800 0.0 0.1166
0.4746 15850 0.0 -
0.4760 15900 0.0 -
0.4775 15950 0.0 -
0.4790 16000 0.0 0.094
0.4805 16050 0.0 -
0.4820 16100 0.0 -
0.4835 16150 0.0 -
0.4850 16200 0.0 0.0963
0.4865 16250 0.0027 -
0.4880 16300 0.0 -
0.4895 16350 0.0 -
0.4910 16400 0.0002 0.0946
0.4925 16450 0.0 -
0.4940 16500 0.0 -
0.4955 16550 0.0 -
0.4970 16600 0.0 0.0896
0.4985 16650 0.0 -
0.5 16700 0.0 -
0.5015 16750 0.0 -
0.5030 16800 0.0 0.0884
0.5045 16850 0.0 -
0.5060 16900 0.0 -
0.5075 16950 0.0 -
0.5090 17000 0.0 -

Framework Versions

  • Python: 3.9.18
  • SetFit: 1.1.0.dev0
  • Sentence Transformers: 2.7.0
  • spaCy: 3.7.4
  • Transformers: 4.40.0
  • PyTorch: 1.12.0
  • Datasets: 2.19.1
  • 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}
}
Downloads last month
4
Safetensors
Model size
117M params
Tensor type
F32
·
Inference Examples
Text Classification
Inference API (serverless) has been turned off for this model.

Finetuned from