Convolutional Neural Networks for Soft-Matching N-Grams in Ad-hoc Search

This paper presents -KNRM, a Convolutional Kernel-based Neural Ranking Model that models n-gram soft matches for ad-hoc search. Instead of exact matching query and document n-grams, -KNRM uses Convolutional Neural Networks to represent n-grams of various lengths and soft matches them in a unified embedding space. The n-gram soft matches are then utilized by the kernel pooling and learning-to-rank layers to generate the final ranking score. -KNRM can be learned end-to-end and fully optimized from user feedback. The learned model»s generalizability is investigated by testing how well it performs in a related domain with small amounts of training data. Experiments on English search logs, Chinese search logs, and TREC Web track tasks demonstrated consistent advantages of -KNRM over prior neural IR methods and feature-based methods.