Ranking Convolutional Recurrent Neural Networks for Purchase Stage Identification on Unbalanced Twitter Data

摘要

Users often use social media to share their interest in products. We propose to iden tify purchase stages from Twitter data fol lowing the AIDA model (Awareness, In terest, Desire, Action). In particular, we define the task of classifying the purchase stage of each tweet in a user's tweet se quence. We introduce RCRNN, a Ranking Convolutional Recurrent Neural Network which computes tweet representations us ing convolution over word embeddings and models a tweet sequence with gated recurrent units. Also, we consider various methods to cope with the unbalanced la bel distribution in our data and show that a ranking layer outperforms class weights.