Recently, the aspect-opinion term pairs (AOTP) extraction task has gained substantial importance in the domain of aspect-based sentiment analysis. It intends to extract the potential pair of each aspect term with its corresponding opinion term present in a user review. Some existing studies heavily relied on the annotated aspect terms and/or opinion terms, or adopted external knowledge/resources to figure out the task. Therefore, in this study, we propose a novel end-to-end solution, called an Interactive AOTP (IAOTP) model, for exploring AOTP. The IAOTP model first tracks the boundary of each token in given aspect-specific and opinion-specific representations through a span-based operation. Next, it generates the candidate AOTP by formulating the dyadic relations between tokens through the Biaffine transformation. Then, it computes the positioning information to capture the significant distance relationship that each candidate pair holds. And finally, it jointly models collaborative interactions and prediction of AOTP through a 2D self-attention. Besides the IAOTP model, this study also proposes an independent aspect/opinion encoding model (a RS model) that formulates relational semantics to obtain aspect-specific and opinion-specific representations that can effectively perform the extraction of aspect and opinion terms. Detailed experiments conducted on the publicly available benchmark datasets for AOTP, aspect terms, and opinion terms extraction tasks, clearly demonstrate the significantly improved performance of our models relative to other competitive state-of-the-art baselines.