Dedicated Farm-Haystack Question Answering System for Pregnant Women and Neonates Using Corona Virus Literature

The global pandemic, COVID-19 has made it more important to quickly and precisely retrieve critical information for effective use by specialists in a wide range of fields. Domain question answering system will work or produce good results to certain extent but still favour more to the pretrained dataset. In this work we target developing a customized question answering framework that can assist the medical network with retrieval of answers to important logical questions like risk factors, effective modes of communication, various treatment options for target high-risk populaces like pregnant women and neonates.The proposed framework uses a customized Farm-Haystack question answering system and introduces a novel pipeline architecture using latent dirichlet allocation and bidirectional encoder representation from transformers for embedding the information. The system is modeled to produce the best and reliable answers for the delicate population, which requires more efficient answers rather than generic population, which can be answered using pretrained systems. In this context, the system has showed the accurate and compact answers for different inquiries related to the sensitive population.