Public articles linked to the same research event.
PLOS digital health This study proposes a method that builds a schizophrenia QA dataset from publicly accessible online health forums using Topic-guided Semantic Modeling (TGSM) and a two-stage Retriever-Reader pipeline, yielding 415,602 posts, 35 topics, and 1,050 QA pairs, and shows that BioBERT fine-tuned on this dataset outperforms its base version and lighter baselines such as DistilBERT on precision and exact match.
This study proposes a method that builds a schizophrenia QA dataset from publicly accessible online health forums using Topic-guided Semantic Modeling (TGSM) and a two-stage Retriever-Reader pipeline, yielding 415,602 posts, 35 topics, and 1,050 QA pairs, and shows that BioBERT fine-tuned on this dataset outperforms its base version and lighter baselines such as DistilBERT on precision and exact match.
This study proposes a method that builds a schizophrenia QA dataset from publicly accessible online health forums using Topic-guided Semantic Modeling (TGSM) and a two-stage Retriever-Reader pipeline, yielding 415,602 posts, 35 topics, and 1,050 QA pairs, and shows that BioBERT fine-tuned on this dataset outperforms its base version and lighter baselines such as DistilBERT on precision and exact match.
This study proposes a method that builds a schizophrenia QA dataset from publicly accessible online health forums using Topic-guided Semantic Modeling (TGSM) and a two-stage Retriever-Reader pipeline, yielding 415,602 posts, 35 topics, and 1,050 QA pairs, and shows that BioBERT fine-tuned on this dataset outperforms its base version and lighter baselines such as DistilBERT on precision and exact match.