IWE-Net: Instance Weight Network for Locating Negative Comments and its application to improve Traffic User Experience

Abstract

Weakly supervised learning aims at coping with scarce labeled data. Previous weakly supervised studies typically assume that there is only one kind of weak supervision in data. In many applications, however, raw data usually contains more than one kind of weak supervision at the same time. For example, in user experience enhancement from Didi, one of the largest online ride-sharing platforms, the ride comment data contains severe label noise (due to the subjective factors of passengers) and severe label distribution bias (due to the sampling bias). We call such a problem as ‘compound weakly supervised learning’. In this paper, we propose the CWSL method to address this problem based on Didi ride-sharing comment data. Specifically, an instance reweighting strategy is employed to cope with severe label noise in comment data, where the weights for harmful noisy instances are small. Robust criteria like AUC rather than accuracy and the validation performance are optimized for the correction of biased data label. Alternating optimization and stochastic gradient methods accelerate the optimization on large-scale data. Experiments on Didi ride-sharing comment data clearly validate the effectiveness. We hope this work may shed some light on applying weakly supervised learning to complex real situations.

Robust Deep Semi-Supervised Learning: A Brief Introduction

“…SSL typically assumes that there is only incomplete supervision. However, in real practice, these weak supervisions often occur simultaneously [134], i.e., compound weakly supervised learning. For example, the compound with incomplete and inaccurate supervised data, i.e., the labels in the small labeled data set could contain label noise [118], [135], the compound with incomplete and inexact supervised data, such as multi-instance SSL [136], [137].…”

“…For example, the compound with incomplete and inaccurate supervised data, i.e., the labels in the small labeled data set could contain label noise [118], [135], the compound with incomplete and inexact supervised data, such as multi-instance SSL [136], [137]. Compound weakly supervised data is commonly occurred in real-world applications as reported in [118], [134], [135]. Robust SSL methods to the compound weakly supervised data are still under-studied.…”