Feature Extraction Functions for Neural Logic Rule Learning

TL;DR

This paper introduces feature extraction functions that incorporate human logic rules into neural networks, enhancing interpretability and flexibility without requiring special mathematical encodings.

Contribution

The proposed method uses programmatic feature functions to integrate logic rules into neural networks, offering a general and flexible approach compared to existing neural logic methods.

Findings

Improved sentiment classification performance

No need for special mathematical encodings

Flexible integration of human knowledge

Abstract

Combining symbolic human knowledge with neural networks provides a rule-based ante-hoc explanation of the output. In this paper, we propose feature extracting functions for integrating human knowledge abstracted as logic rules into the predictive behavior of a neural network. These functions are embodied as programming functions, which represent the applicable domain knowledge as a set of logical instructions and provide a modified distribution of independent features on input data. Unlike other existing neural logic approaches, the programmatic nature of these functions implies that they do not require any kind of special mathematical encoding, which makes our method very general and flexible in nature. We illustrate the performance of our approach for sentiment classification and compare our results to those obtained using two baselines.