Efficient Cost-Aware Cascade Ranking in Multi-Stage Retrieval

By Ruey-Cheng Chen, Luke Gallagher, Roi Blanco, J. Shane Culpepper. 2017

87 citations recorded by the source corpus.

Complex machine learning models are now an integral part of modern, large-scale retrieval systems. However, collection size growth continues to outpace advances in efficiency improvements in the learning models which achieve the highest effectiveness. In this paper, we re-examine the importance of tightly integrating feature costs into multi-stage learning-to-rank (LTR) IR systems. We present a novel approach to optimizing cascaded ranking models which can directly leverage a variety of different state-of-the-art LTR rankers such as LambdaMART and Gradient Boosted Decision Trees. Using our cascade model, we conclusively show that feature costs and the number of documents being re-ranked in each stage of the cascade can be balanced to maximize both efficiency and effectiveness. Finally, we also demonstrate that our cascade model can easily be deployed on commonly used collections to achieve state-of-the-art effectiveness results while only using a subset of the features required by the full model.

Subfields: Computer Vision and Pattern Recognition, Information Systems

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Jimmy J. Lin
2. Rodrigo Nogueira
3. Wei Yang
4. Kyunghyun Cho
5. Hui Fang
6. Peilin Yang
7. Zhuyun Dai
8. Jamie Callan

In Information Systems

1. Jimmy J. Lin
2. Rodrigo Nogueira
3. Wei Yang
4. Kyunghyun Cho
5. Peilin Yang
6. Hui Fang
7. Zeynep Akkalyoncu Yilmaz
8. Sebastian Hofstätter

Citations in the corpus

• Multi-Stage Document Ranking with BERT (2019, background): “This basic approach has received much interest in academia (Matveeva et al., 2006; Wang et al., 2011; Asadi and Lin, 2013; Chen et al., 2017; Mackenzie et al., 2018) as well as industry.”
• Anserini (2018, background): “What about reproducibility beyond simple bag-of-words ranking models? Although we have not explicitly addressed this (next) challenge, we believe that baseline bag-of-words ranking models also have an important role to play in modern multi-stage ranking architectures, which have been explored by res”