An Evaluation of the Kernel Based Neural Ranking Model in NTCIR-13 WWW

This paper describes CMUIR’s participation in the NTCIR13 We Want Web (WWW) task. In the context of the Chinese subtask, we experimented with a neural network approach using the kernel based neural ranking model (KNRM). The model learns a word embedding that encodes IRcustomized soft match patterns from a Chinese search log. The learned model is then directly applied to re-rank the baseline run result lists of the Chinese subtask. We extend K-NRM to incorporate multiple document fields for richer text presentation. We also experimented with different reranking cutoffs to reduce the effect of the gap between training and testing domains. Evaluation results confirmed the effectiveness of K-NRM.