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:

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 this SetFit model to filter these possible aspect span candidates.
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: id_core_news_trf
SetFitABSA Aspect Model: Funnyworld1412/ABSA_mpnet_MiniLM-L6-aspect
SetFitABSA Polarity Model: Funnyworld1412/ABSA_mpnet_MiniLM-L6-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

'pencarian lawan:kapada supercell game nya bagus seru tolong diperbaiki pencarian lawan bermain ketemu player trophy mahkotanya jaraknya dapet berpengaruh peleton akun perbedaan level'
'game:kapada supercell game nya bagus seru tolong diperbaiki pencarian lawan bermain ketemu player trophy mahkotanya jaraknya dapet berpengaruh peleton akun perbedaan level'
'bugnya:bugnya nakal banget y coc cr aja sukanya ngebug pas match suka hitam match relog kalo udah relog lawan udah 1 2 mahkota kecewa sih bintang nya 1 aja bug nya diurus bintang lawannya kadang g setara levelnya dahlah gk suka banget kalo main 2 vs 2 temen suka banget afk coba fitur report'

no aspect

'player trophy mahkotanya jaraknya:kapada supercell game nya bagus seru tolong diperbaiki pencarian lawan bermain ketemu player trophy mahkotanya jaraknya dapet berpengaruh peleton akun perbedaan level'
'peleton akun perbedaan level:kapada supercell game nya bagus seru tolong diperbaiki pencarian lawan bermain ketemu player trophy mahkotanya jaraknya dapet berpengaruh peleton akun perbedaan level'
'y coc cr:bugnya nakal banget y coc cr aja sukanya ngebug pas match suka hitam match relog kalo udah relog lawan udah 1 2 mahkota kecewa sih bintang nya 1 aja bug nya diurus bintang lawannya kadang g setara levelnya dahlah gk suka banget kalo main 2 vs 2 temen suka banget afk coba fitur report'

Evaluation

Metrics

Label	Accuracy
all	0.8317

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(
    "Funnyworld1412/ABSA_mpnet_MiniLM-L6-aspect",
    "Funnyworld1412/ABSA_mpnet_MiniLM-L6-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	29.9357	80

Label	Training Sample Count
no aspect	3834
aspect	1266

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
0.0001	1	0.2801	-
0.0039	50	0.2365	-
0.0078	100	0.1068	-
0.0118	150	0.3401	-
0.0157	200	0.2112	-
0.0196	250	0.3529	-
0.0235	300	0.2338	-
0.0275	350	0.2039	-
0.0314	400	0.2006	-
0.0353	450	0.2939	-
0.0392	500	0.2053	-
0.0431	550	0.2036	-
0.0471	600	0.2229	-
0.0510	650	0.105	-
0.0549	700	0.2222	-
0.0588	750	0.1815	-
0.0627	800	0.2915	-
0.0667	850	0.276	-
0.0706	900	0.1682	-
0.0745	950	0.2328	-
0.0784	1000	0.2422	-
0.0824	1050	0.2753	-
0.0863	1100	0.2292	-
0.0902	1150	0.0791	-
0.0941	1200	0.3849	-
0.0980	1250	0.0964	-
0.1020	1300	0.1612	-
0.1059	1350	0.2755	-
0.1098	1400	0.1133	-
0.1137	1450	0.038	-
0.1176	1500	0.3195	-
0.1216	1550	0.0091	-
0.1255	1600	0.3148	-
0.1294	1650	0.1693	-
0.1333	1700	0.2411	-
0.1373	1750	0.2463	-
0.1412	1800	0.2807	-
0.1451	1850	0.112	-
0.1490	1900	0.2623	-
0.1529	1950	0.2465	-
0.1569	2000	0.4591	-
0.1608	2050	0.0556	-
0.1647	2100	0.0962	-
0.1686	2150	0.4525	-
0.1725	2200	0.2674	-
0.1765	2250	0.1513	-
0.1804	2300	0.3457	-
0.1843	2350	0.1415	-
0.1882	2400	0.0454	-
0.1922	2450	0.0156	-
0.1961	2500	0.2741	-
0.2	2550	0.1334	-
0.2039	2600	0.1838	-
0.2078	2650	0.1346	-
0.2118	2700	0.1022	-
0.2157	2750	0.3999	-
0.2196	2800	0.0953	-
0.2235	2850	0.1201	-
0.2275	2900	0.111	-
0.2314	2950	0.1081	-
0.2353	3000	0.1926	-
0.2392	3050	0.1047	-
0.2431	3100	0.2367	-
0.2471	3150	0.2034	-
0.2510	3200	0.0824	-
0.2549	3250	0.0338	-
0.2588	3300	0.2468	-
0.2627	3350	0.0082	-
0.2667	3400	0.0023	-
0.2706	3450	0.1106	-
0.2745	3500	0.1315	-
0.2784	3550	0.004	-
0.2824	3600	0.0836	-
0.2863	3650	0.2716	-
0.2902	3700	0.1873	-
0.2941	3750	0.4066	-
0.2980	3800	0.1448	-
0.3020	3850	0.0137	-
0.3059	3900	0.3471	-
0.3098	3950	0.1144	-
0.3137	4000	0.0596	-
0.3176	4050	0.0377	-
0.3216	4100	0.3316	-
0.3255	4150	0.0709	-
0.3294	4200	0.0515	-
0.3333	4250	0.2029	-
0.3373	4300	0.1191	-
0.3412	4350	0.2397	-
0.3451	4400	0.492	-
0.3490	4450	0.1178	-
0.3529	4500	0.3647	-
0.3569	4550	0.0098	-
0.3608	4600	0.2114	-
0.3647	4650	0.2392	-
0.3686	4700	0.2194	-
0.3725	4750	0.0578	-
0.3765	4800	0.0771	-
0.3804	4850	0.1582	-
0.3843	4900	0.0643	-
0.3882	4950	0.1372	-
0.3922	5000	0.0308	-
0.3961	5050	0.1247	-
0.4	5100	0.3076	-
0.4039	5150	0.1152	-
0.4078	5200	0.2112	-
0.4118	5250	0.0042	-
0.4157	5300	0.0869	-
0.4196	5350	0.0196	-
0.4235	5400	0.2406	-
0.4275	5450	0.3306	-
0.4314	5500	0.2328	-
0.4353	5550	0.008	-
0.4392	5600	0.0388	-
0.4431	5650	0.3812	-
0.4471	5700	0.6268	-
0.4510	5750	0.4426	-
0.4549	5800	0.1407	-
0.4588	5850	0.297	-
0.4627	5900	0.2657	-
0.4667	5950	0.1767	-
0.4706	6000	0.0152	-
0.4745	6050	0.2344	-
0.4784	6100	0.0447	-
0.4824	6150	0.0675	-
0.4863	6200	0.3086	-
0.4902	6250	0.5258	-
0.4941	6300	0.0826	-
0.4980	6350	0.0079	-
0.5020	6400	0.1817	-
0.5059	6450	0.0767	-
0.5098	6500	0.0221	-
0.5137	6550	0.0419	-
0.5176	6600	0.2452	-
0.5216	6650	0.0232	-
0.5255	6700	0.0804	-
0.5294	6750	0.1752	-
0.5333	6800	0.0127	-
0.5373	6850	0.0454	-
0.5412	6900	0.1759	-
0.5451	6950	0.0435	-
0.5490	7000	0.0109	-
0.5529	7050	0.0162	-
0.5569	7100	0.0133	-
0.5608	7150	0.2363	-
0.5647	7200	0.4987	-
0.5686	7250	0.1149	-
0.5725	7300	0.4613	-
0.5765	7350	0.3837	-
0.5804	7400	0.2439	-
0.5843	7450	0.0014	-
0.5882	7500	0.0177	-
0.5922	7550	0.0051	-
0.5961	7600	0.0418	-
0.6	7650	0.0061	-
0.6039	7700	0.2205	-
0.6078	7750	0.1769	-
0.6118	7800	0.0071	-
0.6157	7850	0.2271	-
0.6196	7900	0.3049	-
0.6235	7950	0.0016	-
0.6275	8000	0.2263	-
0.6314	8050	0.0057	-
0.6353	8100	0.1408	-
0.6392	8150	0.0303	-
0.6431	8200	0.0026	-
0.6471	8250	0.1743	-
0.6510	8300	0.2078	-
0.6549	8350	0.1764	-
0.6588	8400	0.0127	-
0.6627	8450	0.2435	-
0.6667	8500	0.0527	-
0.6706	8550	0.247	-
0.6745	8600	0.002	-
0.6784	8650	0.0087	-
0.6824	8700	0.1866	-
0.6863	8750	0.0087	-
0.6902	8800	0.1589	-
0.6941	8850	0.1848	-
0.6980	8900	0.0298	-
0.7020	8950	0.0081	-
0.7059	9000	0.3057	-
0.7098	9050	0.2059	-
0.7137	9100	0.2154	-
0.7176	9150	0.0013	-
0.7216	9200	0.1961	-
0.7255	9250	0.0129	-
0.7294	9300	0.0021	-
0.7333	9350	0.2106	-
0.7373	9400	0.0008	-
0.7412	9450	0.1261	-
0.7451	9500	0.1948	-
0.7490	9550	0.013	-
0.7529	9600	0.208	-
0.7569	9650	0.2382	-
0.7608	9700	0.0054	-
0.7647	9750	0.1869	-
0.7686	9800	0.0334	-
0.7725	9850	0.0197	-
0.7765	9900	0.0057	-
0.7804	9950	0.0056	-
0.7843	10000	0.0043	-
0.7882	10050	0.0025	-
0.7922	10100	0.6808	-
0.7961	10150	0.043	-
0.8	10200	0.0536	-
0.8039	10250	0.2435	-
0.8078	10300	0.0051	-
0.8118	10350	0.0653	-
0.8157	10400	0.017	-
0.8196	10450	0.0036	-
0.8235	10500	0.1561	-
0.8275	10550	0.001	-
0.8314	10600	0.1975	-
0.8353	10650	0.2378	-
0.8392	10700	0.1276	-
0.8431	10750	0.0719	-
0.8471	10800	0.1951	-
0.8510	10850	0.0446	-
0.8549	10900	0.2045	-
0.8588	10950	0.0598	-
0.8627	11000	0.0094	-
0.8667	11050	0.1117	-
0.8706	11100	0.0528	-
0.8745	11150	0.0047	-
0.8784	11200	0.1492	-
0.8824	11250	0.2204	-
0.8863	11300	0.0089	-
0.8902	11350	0.0709	-
0.8941	11400	0.1111	-
0.8980	11450	0.0048	-
0.9020	11500	0.0173	-
0.9059	11550	0.2862	-
0.9098	11600	0.2745	-
0.9137	11650	0.0054	-
0.9176	11700	0.0074	-
0.9216	11750	0.0036	-
0.9255	11800	0.0869	-
0.9294	11850	0.2333	-
0.9333	11900	0.15	-
0.9373	11950	0.066	-
0.9412	12000	0.1742	-
0.9451	12050	0.0009	-
0.9490	12100	0.1246	-
0.9529	12150	0.1674	-
0.9569	12200	0.1937	-
0.9608	12250	0.0724	-
0.9647	12300	0.0044	-
0.9686	12350	0.0013	-
0.9725	12400	0.0313	-
0.9765	12450	0.0925	-
0.9804	12500	0.1742	-
0.9843	12550	0.2294	-
0.9882	12600	0.1073	-
0.9922	12650	0.038	-
0.9961	12700	0.1866	-
1.0	12750	0.0141	0.2274

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

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}
}