NazmusAshrafi's picture
Add SetFit ABSA model
736d448 verified
metadata
library_name: setfit
tags:
  - setfit
  - absa
  - sentence-transformers
  - text-classification
  - generated_from_setfit_trainer
metrics:
  - accuracy
widget:
  - text: >-
      ; got my car in about a:Well.. I added a new 🐎 to the stable! Special
      thanks to Matt at the @Tesla Clarkston location who made my Model Y order
      & delivery incredibly smooth.🙏 I'm super lucky & got my car in
      about a week of deciding to go for it 😳 Video coming soon about that
      process & more! https://t.co/PrP91xMnKk
  - text: >-
      . But the price could be cheaper:C’mon @elonmusk! Australians are busting
      to buy EVs & the best one is @Tesla imho. But the price could be
      cheaper, if you built a #gigafactory in Australia. 70% of the lithium in
      the cars is #aussie so why not set up a #gigafactorydownunder? All the
      talent and minerals are here!
  - text: >-
      generate more net profit from legacy auto:As with previous quarters, $TSLA
      will generate more net profit from legacy auto regulatory credits sales
      this quarter than legacy auto will make in gross profit by selling EVs.


      This just keeps adding insult to injury.
  - text: >-
      on keeping this car for 10 years:@_brivnii @Tesla I plan on keeping this
      car for 10 years total (so 6 more years at least). I don't feel the need
      to upgrade to a newer model even if price is no issue. This one has been
      reliable, and I got a good battery (no signs of degradation so far)
  - text: >-
      The driver’s car was a @Tesla:I took an @Uber home from the airport and my
      bill had a fuel surcharge on it because of the current price of gasoline. 


      The driver’s car was a @Tesla… 🤷
pipeline_tag: text-classification
inference: false
base_model: sentence-transformers/paraphrase-mpnet-base-v2
model-index:
  - name: SetFit Polarity Model with sentence-transformers/paraphrase-mpnet-base-v2
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: Unknown
          type: unknown
          split: test
        metrics:
          - type: accuracy
            value: 0.44
            name: Accuracy

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

This is a SetFit model 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
neutral
  • "i.e. just more profit for $TSLA:I'm pretty sure, all an EV tax incentive will do, is raise the price of Teslas, at least for the next few years.\n\ni.e. just more profit for $TSLA\nAs if demand wasn't abundant enough already."
  • "increase also increase profit and Tesla the:C'mon @SaraEisen you know as well as I do that the price increases in $TSLA vehicles is not related to any weakness in the stock today. It's purely macro today. Plus, price increase also increase profit and Tesla the only Auto maker that is making high margins on sales."
  • "when choosing a car. They just:The key thing people get wrong when thinking EV competition is bad for Tesla:\n\nVast majority of consumers aren't prioritizing the environment when choosing a car. They just want the best product & technology, which Tesla offers and just happens to be an EV\n\n$TSLA https://t.co/w3cKqeJkQW"
negative
  • "is raise the price of Teslas,:I'm pretty sure, all an EV tax incentive will do, is raise the price of Teslas, at least for the next few years.\n\ni.e. just more profit for $TSLA\nAs if demand wasn't abundant enough already."
  • '"The price of batteries for:"The price of batteries for electric vehicles looks set to rise in 2022 after many years of sharp decline. The supplies of lithium and other raw materials fail to keep up with huge demand." $NIO $TSLA $XPEV $LI\n\nhttps://t.co/2CAJCxTC2C'
  • '. But the price could be cheaper:C’mon @elonmusk! Australians are busting to buy EVs & the best one is @Tesla imho. But the price could be cheaper, if you built a #gigafactory in Australia. 70% of the lithium in the cars is #aussie so why not set up a #gigafactorydownunder? All the talent and minerals are here!'
positive
  • 'a $30k car with $70k:John Hennessey gets a $TSLA Plaid. \nA retired OEM executive describes Tesla as a $30k car with $70k in batteries. \nThe perfect description of a Tesla https://t.co/m5J5m3AuMJ'
  • "want the best product &:The key thing people get wrong when thinking EV competition is bad for Tesla:\n\nVast majority of consumers aren't prioritizing the environment when choosing a car. They just want the best product & technology, which Tesla offers and just happens to be an EV\n\n$TSLA https://t.co/w3cKqeJkQW"
  • "the most important product on Earth;:Tesla's 4680 battery (and it's manufactuing process) will end up being the most important product on Earth; it's that important. It will enable massive scale & cost reductions over time. It will enable faster charging times & longer range EVs, both will drive adoption.\n\n$TSLA"

Evaluation

Metrics

Label Accuracy
all 0.44

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(
    "NazmusAshrafi/setfit-MiniLM-mpnet-absa-tesla-tweet-aspect",
    "NazmusAshrafi/setfit-MiniLM-mpnet-absa-tesla-tweet-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 26 46.2121 61
Label Training Sample Count
negative 11
neutral 12
positive 10

Training Hyperparameters

  • batch_size: (16, 2)
  • num_epochs: (1, 16)
  • 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: 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
0.0217 1 0.186 -

Framework Versions

  • Python: 3.10.12
  • SetFit: 1.0.3
  • Sentence Transformers: 2.2.2
  • spaCy: 3.6.1
  • Transformers: 4.35.2
  • PyTorch: 2.1.0+cu121
  • Datasets: 2.16.1
  • Tokenizers: 0.15.1

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}
}