SetFit Polarity Model with sentence-transformers/paraphrase-mpnet-base-v2

This is a SetFit model trained on the omymble/setfit-books-categories dataset that can be used for Aspect Based Sentiment Analysis (ABSA). This SetFit model uses sentence-transformers/paraphrase-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:

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

Model Details

Model Description

Model Sources

Model Labels

Label Examples
BOOK#AUDIENCE
  • 'I recommend this for fans of fantasy, or other books by Garth Nix'
  • 'I first got this book when I was eight and I totally loved it! I have read it every year since then! It is about a pair of twins who are born one VERY good and one EXTREMELY bad'
  • 'However, I did feel one particular scene might be rather nightmare-inducing for the youngest readers - so recommend this for the ages of 12 and above'
BOOK#AUTHOR
  • 'Banks has writin better books than this book,'
  • "Now in is astonishing new novel, Michael Dobbs throws brilliant fresh light upon Churchill's relationship with the Soviet spy and the twenty months of conspiracy, chance and outright treachery that were to propel Churchill from outcast to messiah and change the course of history"
  • 'Paul focuses on the problems of an intimate relationship and the decisions the teens make at that moment'
BOOK#GENERAL
  • 'This is the first book in the Keys to the Kingdom series by Garth Nix'
  • 'The book is a great read right until the end, so rare in non-fiction'
  • 'Anne Kingston did a marvellous job on this book'
BOOK#TITLE
  • 'Personal I loved My Darling My Hamburger'
  • 'But THE INTRUDERS is pretty much a middling effort, at least when it comes to the plot'
  • 'After reading several pages I relented and purchased Mister Monday'
CONTENT#CHARACTERS
  • "She's not a great writer but she's a fabulous storyteller and her Tony Hill/Carol Jordan mysteries are the best of the bunch"
  • 'but before he can do that he has to dodge fechters, run from enemys like Noon and Dawn, run from dinosaurs, try not to get killed, and try to prevent himself from having a asthma atackk!! But, thankfully he has some help from a girl named suzy, a guy named Dusk, and a talking toad'
  • 'But when a fight emerges between the two figures - Mister Monday and Sneezer - they both disappear without any further regard to Arthur'
CONTENT#GENRE
  • 'I love fantasy and science fiction, but this storyteller forgot something very important'
  • 'At first I was amused an entertained by Angela and Diabola the novel by Lynne Reid Banks, but as it progressed and became exceedingly darker, I read the jacket to find that this book was recommended for ages 9-12'
  • "Here's a thriller that really thrills"

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(
    "setfit-absa-aspect",
    "omymble/books-categories",
)
# 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 21.0917 78
Label Training Sample Count
BOOK#AUDIENCE 20
BOOK#AUTHOR 20
BOOK#GENERAL 20
BOOK#TITLE 20
CONTENT#CHARACTERS 20
CONTENT#GENRE 20

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
0.0106 1 0.2623 -
0.5319 50 0.1293 -
1.0638 100 0.0132 -
1.5957 150 0.0022 -
2.1277 200 0.0027 -
2.6596 250 0.0013 -
3.1915 300 0.0017 -
3.7234 350 0.0015 -
4.2553 400 0.0029 -
4.7872 450 0.0015 -
0.0106 1 0.0115 -
0.5319 50 0.009 0.1324
1.0638 100 0.0094 0.1267
1.5957 150 0.0007 0.1194
2.1277 200 0.0017 0.1256
2.6596 250 0.0008 0.1293
3.1915 300 0.0007 0.1173
3.7234 350 0.0008 0.1231
4.2553 400 0.0023 0.1272
4.7872 450 0.0008 0.1241
  • The bold row denotes the saved checkpoint.

Framework Versions

  • Python: 3.10.12
  • SetFit: 1.0.3
  • Sentence Transformers: 3.1.0
  • spaCy: 3.7.4
  • Transformers: 4.39.0
  • PyTorch: 2.3.1+cu121
  • Datasets: 2.20.0
  • 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}
}
Downloads last month
15
Safetensors
Model size
109M params
Tensor type
F32
·
Inference Examples
Inference API (serverless) has been turned off for this model.

Model tree for omymble/books-categories

Finetuned
(256)
this model

Dataset used to train omymble/books-categories