Multimodal AutoML on Tables with Text Fields

Xingjian Shi, Jonas Mueller, Nick Erickson, Mu Li, Alex Smola

Keywords: Multimodal AutoML, Text Data, Tabular Data, Natural Language Processing, Supervised Learning

TL;DR: We evaluate multimodal ML strategies over a new benchmark for classification/regression with data tables that jointly contain numeric, categorical, and text fields.

Abstract: We consider the design of automated supervised learning systems for data tables that not only contain numeric/categorical columns, but text fields as well. Here we assemble 15 multimodal data tables that each contain some text fields and stem from a real business application. Over this benchmark, we evaluate numerous multimodal AutoML strategies, including standard two-stage approaches where NLP is used to featurize the text such that AutoML for tabular data can then be applied. We identify practically superior strategies based on multimodal adaptations of Transformer networks and stack ensembling of these networks with classical tabular models. Compared with human data science teams, the best fully automated methodology discovered through our benchmark manages to rank 1st place when fit to the raw text/tabular data in two MachineHack prediction competitions and 2nd place (out of 2380 teams) in Kaggle's Mercari Price Suggestion Challenge.

URL: See Dataset Benchmark URL comment