INNER CODE UNIT · Python

data

NorskRegnesentral/skweak · skweak/aggregation.py:280

        data = np.full((len(unique_spans), len(sources)),fill_value=-1, dtype=np.int16)

        observed_labels = self.observed_labels
        # Populating the array with the labels from each source
        label_indices = {l: i for i, l in enumerate(observed_labels)}
        
        for source_index, source in enumerate(sources):
            for span in doc.spans[source]:
                if span.label_ in observed_labels:
                    span_index = spans_indices[(span.start, span.end)]
                    data[span_index, source_index] = label_indices[span.label_]

        # We only consider spans with at least one concrete prediction
        masking = np.full(len(unique_spans), fill_value=True, dtype=bool)
        for i, row in enumerate(data):
            if row.max() < 0:
                masking[i] = False
        data = data[masking]

View source record →