Sentiment　analysis,　which　includes　aspect　sentiment　classification,　is　an　important　and　basic　task　in　natural　language　processing　(NLP).　Aspect　sentiment　analysis　can　provide　complete　and　in-depth　results　with　increased　attention　on　the　aspect　level.　Different　context　words　influence　the　sentiment　polarity　of　a　sentence　variably,　and　sentiment　polarity　varies　based　on　the　different　aspects　in　a　sentence.　Take　the　following　sentence　as　an　example:　‘The　voice　quality　of　this　phone　is　not　good,　but　the　battery　life　is　long’.　If　the　aspect　is　voice　quality,　then　the　sentiment　polarity　is　‘negative’;　if　the　battery　life　aspect　is　considered,　the　sentiment　polarity　should　be　‘positive’.　Therefore,　aspect　is　crucial　when　we　explore　aspect　sentiment　in　the　sentence.　Recurrent　neural　network　(RNN),　a　suitable　model　for　dealing　with　NLP,　demonstrates　good　performance　in　aspect　sentiment　classification.　The　literature　on　sentiment　classification　by　utilising　convolutional　neural　network　(CNN)　is　extensive.　In　this　study,　we　utilise　a　convolutional　long　short-term　memory　(C-LSTM)　model　for　sentence　representation　to　deal　with　aspect　sentiment.　C-LSTM　extracts　a　sequence　of　high-level　phrase　representations　by　using　CNN,　and　representations　obtained　from　CNN　are　fed　to　an　LSTM　to　obtain　sentence　representation　[1].　C-LSTM　can　capture　local　features　of　phrases　and　global　and　temporal　sentence　semantics.　We　incorporate　the　attention　mechanism　into　C-LSTM　to　obtain　a　sentence　representation　that　selectively　concentrates　on　several　aspects　of　information　and　thus　achieve　improved　performance.　We　classify　aspect　sentiment　without　using　a　syntactic　parser　or　other　language　features　in　a　benchmark　dataset　from　Twitter.　©　2018　Indian　Pulp　and　Paper　Technical　Association.　All　Rights　Reserved.