Universal Model in Online Customer Service

arXiv:2402.15666v1 Announce Type: new
Abstract: Building machine learning models can be a time-consuming process that often takes several months to implement in typical business scenarios. To ensure consistent model performance and account for variations in data distribution, regular retraining is necessary. This paper introduces a solution for improving online customer service in e-commerce by presenting a universal model for predict-ing labels based on customer questions, without requiring training. Our novel approach involves using machine learning techniques to tag customer questions …

abstract arxiv building business commerce consistent cs.ai cs.ir cs.lg customer customer service data distribution e-commerce machine machine learning machine learning models paper performance process retraining service solution type universal universal model