Publication

Abstract : In health community forums, patients seek information, acquire knowledge and support each other. They find help by asking questions about their illness. Finding similar people with similar issues to obtain correct information is often a great challenge for them. Hence, this paper proposes a query similarity model to retrieve the similar query that is already discussed. The model is implemented with a Siamese neural network to assess the semantic similarity between the queries. The presence of several medical terms imposed great challenge to the model to generalize the query well. Hence, the model is fine-tuned with a transfer learning technique on a medical question–answer dataset. Several variants of the model are compared, and the best model could achieve an F1-score of 82.4%. The result showed that the proposed method is very effective for natural language understanding in medical domain.