From 000b1273c30e72f43f57344668026eda46793c5e Mon Sep 17 00:00:00 2001 From: dukesun99 Date: Mon, 20 Feb 2023 22:22:55 +0800 Subject: [PATCH 1/2] Fix a bug to cause top k results having the same title --- src/refined/inference/processor.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/refined/inference/processor.py b/src/refined/inference/processor.py index 667dc68..a126e28 100644 --- a/src/refined/inference/processor.py +++ b/src/refined/inference/processor.py @@ -157,6 +157,7 @@ def process_text( :return: a list of spans with predictions attached to each span (sorted by start indices) or a list of mention dictionary (in distant supervision dataset format) when ds_format is True. """ + print("Processing text: ", text) all_spans = [] if spans is not None: doc = Doc.from_text_with_spans( @@ -374,7 +375,7 @@ def process_tensors(self, batch: BatchedElementsTns, ner_threshold: float = 0.5, span.entity_linking_model_confidence_score = predicted_entity_confidence[span_idx] span.top_k_predicted_entities = [ (Entity(wikidata_entity_id=f'Q{entity_id}', - wikipedia_entity_title=self.preprocessor.qcode_to_wiki.get(wikidata_id) + wikipedia_entity_title=self.preprocessor.qcode_to_wiki.get(f'Q{entity_id}') if self.preprocessor.qcode_to_wiki is not None else None ), round(score, 4)) From 3f1d23ed591e1817db594a0290af3cbed3bf6c33 Mon Sep 17 00:00:00 2001 From: dukesun99 Date: Mon, 20 Feb 2023 22:27:21 +0800 Subject: [PATCH 2/2] Remove debugging info --- src/refined/inference/processor.py | 1 - 1 file changed, 1 deletion(-) diff --git a/src/refined/inference/processor.py b/src/refined/inference/processor.py index a126e28..b324327 100644 --- a/src/refined/inference/processor.py +++ b/src/refined/inference/processor.py @@ -157,7 +157,6 @@ def process_text( :return: a list of spans with predictions attached to each span (sorted by start indices) or a list of mention dictionary (in distant supervision dataset format) when ds_format is True. """ - print("Processing text: ", text) all_spans = [] if spans is not None: doc = Doc.from_text_with_spans(