SHAP for sentiment analysis

An example of SHAP on sentiment analysis. The SHAP explainer for NLP tasks only supports TextClassificationPipeline in the transformer library. If using this explainer, please cite the original work: https://github.com/slundberg/shap.

[1]:

import transformers
from omnixai.data.text import Text
from omnixai.explainers.nlp import ShapText

[2]:

# A transformer model for sentiment analysis
model = transformers.pipeline(
    'sentiment-analysis',
    model='distilbert-base-uncased-finetuned-sst-2-english',
    return_all_scores=True
)

[3]:

x = Text([
    "What a great movie! if you have no taste.",
    "it was a fantastic performance!",
    "best film ever",
    "such a great show!",
    "it was a horrible movie",
    "i've never watched something as bad"
])
explainer = ShapText(model=model)
explanations = explainer.explain(x)
explanations.ipython_plot()

Partition explainer: 7it [00:17,  3.51s/it]

Instance 0: Class POSITIVE

What a great movie ! if you have no taste .

Instance 1: Class POSITIVE

it was a fantastic performance !

Instance 2: Class POSITIVE

best film ever

Instance 3: Class POSITIVE

such a great show !

Instance 4: Class NEGATIVE

it was a horrible movie

Instance 5: Class POSITIVE

i ' ve never watched something as bad