Neural Feature Selection for Learning to Rank
This work addresses efficiency challenges for deploying neural LETOR models in real-world information retrieval systems, representing an incremental improvement.
The paper tackles the problem of large model size and complexity in neural Learning to Rank (LETOR) systems, which impacts latency and update time in large-scale search, by proposing an architecture-agnostic neural feature selection method that reduces input size by up to 60% and training/inference time by up to 50% without performance loss.
LEarning TO Rank (LETOR) is a research area in the field of Information Retrieval (IR) where machine learning models are employed to rank a set of items. In the past few years, neural LETOR approaches have become a competitive alternative to traditional ones like LambdaMART. However, neural architectures performance grew proportionally to their complexity and size. This can be an obstacle for their adoption in large-scale search systems where a model size impacts latency and update time. For this reason, we propose an architecture-agnostic approach based on a neural LETOR model to reduce the size of its input by up to 60% without affecting the system performance. This approach also allows to reduce a LETOR model complexity and, therefore, its training and inference time up to 50%.