Mutual Implication as a Measure of Textual Equivalence


 Semantic Textual Similarity (STS) and paraphrase de- tection are two NLP tasks that have a high focus on the meaning of sentences, current research in both re- lies heavily comparing fragments text. Little to no work has been done studying inference-centric ap- proaches solve these tasks. We study relation be- tween existing what we call mutual implica- tion (MI), binary relationship between sentences holds when they textually entail each other. MI thus shifts STS detection un- derstanding sentence terms its in- ferential properties. comparison MI, paraphrasing, work. then argue should be considered complementary evaluation met- ric for advancing areas as diverse machine translation, natural language inference, etc. Finally, limitations discuss possibilities overcoming them.