SEQ2SEQ++: A Multitasking-Based Seq2seq Model to Generate Meaningful and Relevant Answers

Question-answering chatbots have tremendous potential to complement humans in various fields. They are implemented using either rule-based or machine learning-based systems. Unlike the former, more scalable. Sequence-to-sequence (Seq2Seq) learning is one of most popular approaches and has shown remarkable progress since its introduction 2014. However, based on Seq2Seq show a weakness that it tends generate answers can be generic inconsistent with questions, thereby becoming meaningless and, therefore, may lower chatbot adoption rate. This attributed three issues: question encoder overfit, answer generation language model influence. Several recent methods utilize multitask (MTL) address this weakness. existing MTL models very little improvement over single-task learning, wherein they still answers. paper presents novel approach for called SEQ2SEQ++, which comprises multifunctional encoder, an decoder, ternary classifier. Additionally, SEQ2SEQ++ utilizes dynamic tasks loss weight mechanism calculation attention comprehensive mechanism. Experiments NarrativeQA SQuAD datasets were conducted gauge performance proposed comparison two recently models. The experimental results yields noteworthy improvements bilingual evaluation understudy, word error rate, Distinct-2 metrics.