Adding all missing named entities back into simplified biomedical text yields the highest cosine similarity and ROUGE-1 to the original among five insertion strategies, but the evaluation is partly circular and lacks significance testing.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Automated Feedback Loops to Protect Text Simplification with Generative AI from Information Loss
Adding all missing named entities back into simplified biomedical text yields the highest cosine similarity and ROUGE-1 to the original among five insertion strategies, but the evaluation is partly circular and lacks significance testing.