BiLSTM　has　been　widely　used　in　the　field　of　text　classification,　but　the　model　still　cannot　accurately　measure　the　importance　of　each　word　and　cannot　extract　text　features　more　effectively.　In　order　to　solve　this　problem,　this　paper　proposes　a　BiLSTM-WSAttention　model.　The　neural　network　model　of　BiLSTM-WSAttention　is　used　to　classify　text.　The　BiLSTM-WSAttention　neural　network　model　combines　the　context　of　words　and　sentences,　and　extracts　contextual　semantic　information　from　two　perspectives:　front　to　back　and　back　to　front.　At　the　same　time,　this　article　introduces　an　attention　mechanism.　Since　text　is　composed　of　sentences,　sentences　are　composed　of　words,　and　the　importance　of　words　and　sentences　depends　on　context　information,　so　this　article　includes　word-level　attention　mechanisms　and　sentence-level.　The　attention　mechanism　assigns　different　weight　values　to　different　words　and　sentences.　Finally,　the　method　proposed　in　this　article　is　compared　with　the　classification　methods　of　Naive-Bayes,　CNN,　RNN,　and　BLSTM　on　the　same　data　set.　The　experimental　results　show　that:　Compared　with　other　classification　methods,　the　neural　network　model　BiLSTM-WSAttention　proposed　in　this　article　is　effective　on　this　data　set.　©　2021　IEEE.