SepLL: Separating Latent Class Labels from Weak Supervision Noise

TL;DR

SepLL is a novel deep learning approach that separates task-related and labeling function-specific information in the latent space, improving weak supervision learning without pre-processing.

Contribution

It introduces an end-to-end transformer-based model that disentangles labeling function noise from true labels in the latent space, enhancing weak supervision learning.

Findings

Achieves state-of-the-art performance on Wrench text classification.

Effectively separates label noise from true signal in latent space.

No pre-processing or label correction needed.

Abstract

In the weakly supervised learning paradigm, labeling functions automatically assign heuristic, often noisy, labels to data samples. In this work, we provide a method for learning from weak labels by separating two types of complementary information associated with the labeling functions: information related to the target label and information specific to one labeling function only. Both types of information are reflected to different degrees by all labeled instances. In contrast to previous works that aimed at correcting or removing wrongly labeled instances, we learn a branched deep model that uses all data as-is, but splits the labeling function information in the latent space. Specifically, we propose the end-to-end model SepLL which extends a transformer classifier by introducing a latent space for labeling function specific and task-specific information. The learning signal is only given by the labeling functions matches, no pre-processing or label model is required for our method. Notably, the task prediction is made from the latent layer without any direct task signal. Experiments on Wrench text classification tasks show that our model is competitive with the state-of-the-art, and yields a new best average performance.