65 lines
2.1 KiB
Python
65 lines
2.1 KiB
Python
from typing import Dict, Iterable, List
|
|
from great_ai import GreatAI, use_model, ClassificationOutput
|
|
from sklearn.pipeline import Pipeline
|
|
|
|
from helpers import lemmatize, preprocess
|
|
|
|
|
|
@GreatAI.deploy
|
|
@use_model("small-domain-prediction-v2", version="latest")
|
|
def predict_domain(text: str, model: Pipeline, target_confidence: float = 20) -> List[ClassificationOutput]:
|
|
"""
|
|
Predict the scientific domain of the input text.
|
|
Return labels until their sum likelihood is larger than target_confidence.
|
|
"""
|
|
assert 0 <= target_confidence <= 100, "invalid argument"
|
|
|
|
text = preprocess(text)
|
|
|
|
token_mapping = {lemmatize(original): original for original in text.split(" ")}
|
|
feature_names = [
|
|
token_mapping.get(name)
|
|
for name in model.named_steps["vectorizer"].get_feature_names_out()
|
|
]
|
|
|
|
features = model.named_steps["vectorizer"].transform(
|
|
[" ".join(token_mapping.keys())]
|
|
)
|
|
prediction = model.named_steps["classifier"].predict_proba(features)[0]
|
|
best_classes = sorted(enumerate(prediction), key=lambda v: v[1], reverse=True)
|
|
|
|
results: List[ClassificationOutput] = []
|
|
for class_index, probability in best_classes:
|
|
weights = model.named_steps["classifier"].feature_log_prob_[class_index]
|
|
domain = model.named_steps["classifier"].classes_[class_index]
|
|
|
|
results.append(
|
|
ClassificationOutput(
|
|
label=domain,
|
|
confidence=round(probability * 100),
|
|
explanation=_get_explanation(
|
|
weights=weights,
|
|
counts=features.A[0],
|
|
words=feature_names,
|
|
),
|
|
)
|
|
)
|
|
|
|
if sum(r.confidence for r in results) >= target_confidence:
|
|
break
|
|
|
|
return results
|
|
|
|
|
|
def _get_explanation(
|
|
weights: Iterable[float],
|
|
counts: Iterable[float],
|
|
words: Iterable[str],
|
|
) -> List[str]:
|
|
most_influential = sorted((
|
|
(weight, word)
|
|
for weight, count, word in zip(weights, counts, words)
|
|
if count > 0
|
|
), reverse=True)[:5]
|
|
|
|
return [word for _, word in most_influential]
|