Add professor definition and basic implementation

aspect_based_sentiment_analysis/professors.py

@@ -0,0 +1,75 @@
+from abc import ABC
+from dataclasses import dataclass
+from typing import List, Tuple
+
+import numpy as np
+import tensorflow as tf
+
+from .aux_models import ReferenceRecognizer
+from .aux_models import PatternRecognizer
+from .data_types import Iterable
+from .data_types import PredictedExample
+from .data_types import OutputBatch
+from .data_types import Review
+from .data_types import Sentiment
+from .data_types import Task
+
+
+@dataclass
+class _Professor(ABC):
+    """ """
+    ref_recognizer: ReferenceRecognizer = None
+    pattern_recognizer: PatternRecognizer = None
+
+    def make_decision(
+            self,
+            task: Task,
+            output_batch: OutputBatch
+    ) -> Iterable[PredictedExample]:
+        """ """
+
+
+@dataclass
+class Professor(_Professor):
+
+    def make_decision(
+            self,
+            task: Task,
+            output_batch: OutputBatch
+    ) -> Iterable[PredictedExample]:
+        reviews = self.review(task, output_batch)
+        originals = self.get_originals(output_batch.scores)
+        for example, original, review in zip(task, originals, reviews):
+
+            if review.is_reference is False:
+                sentiment = Sentiment.neutral
+                scores = [0, 0, 0]
+            else:
+                sentiment, scores = original
+
+            prediction = PredictedExample.from_example(
+                example, sentiment=sentiment, scores=scores, review=review)
+            yield prediction
+
+    def review(
+            self,
+            task: Task,
+            output_batch: OutputBatch
+    ) -> Iterable[Review]:
+        for example, args in zip(task, output_batch):
+            is_reference = self.ref_recognizer(example, *args) \
+                if self.ref_recognizer else None
+            patterns = self.pattern_recognizer(example, *args) \
+                if self.pattern_recognizer and is_reference is not False else\
+                None
+            review = Review(is_reference, patterns)
+            yield review
+
+    @staticmethod
+    def get_originals(batch_scores: tf.Tensor) -> Iterable[
+        Tuple[Sentiment, List[float]]]:
+        for scores in batch_scores:
+            sentiment_id = np.argmax(scores).astype(int)
+            sentiment = Sentiment(sentiment_id)
+            scores = list(scores)
+            yield sentiment, scores