The　fine-grained　affective　analysis　of　product　reviews　is　a　fine-grained　mining　of　the　content　of　reviews,　which　has　important　research　significance.　The　explicit　attribute-viewpoint　extraction　is　one　of　its　key　research　issues.　Due　to　the　complex　structure　and　colloquial　features　of　online　product　reviews,　traditional　models　are　not　ideal　for　the　extraction　of　explicit　attribute-views　pairs.　To　solve　this　problem,　this　paper　proposes　an　extraction　method　based　on　deep　learning　and　integrating　positional　relationship　information.　By　using　the　Bi-directional　Long　Short-Term　Memory　Network　to　overcome　the　problem　of　long　distance　dependence,　the　author　makes　full　use　of　the　context,　combines　the　attention　mechanism　to　reduce　the　noise　weight　of　sentence-level　informal　text,　and　integrates　the　feature　of　position　relation　to　enrich　the　feature　information.　The　experimental　results　show　that　compared　with　other　models,　the　network　model　proposed　in　this　paper　has　improved　the　call　rate　and　F1　value.　©　2019　Published　under　licence　by　IOP　Publishing　Ltd.