Symbolic Knowledge Extraction and Injection with Sub-symbolic Predictors: a Systematic Literature Review

Giovanni Ciatto, Federico Sabbatini, Andrea Agiollo, Matteo Magnini, Andrea Omicini

ACM Computing Surveys

In this paper we focus on the issue of opacity of sub-symbolic machine-learning predictors by promoting two complementary activities—namely, symbolic knowledge extraction (SKE) and injection (SKI) from and into sub-symbolic predictors. There, we consider as symbolic any language that is intelligible and interpretable for both human beings and computers. Along this line, in this work we propose a general meta-model for both SKE and SKI, along with two taxonomies for the classification of SKE and SKI methods. By adopting a XAI perspective, we highlight how both SKE and SKI methods can be exploited to either mitigate or overcome the aforementioned opacity issue. Notably, our taxonomies are attained by surveying and classifying the existing SKE/SKI methods from the literature, following a systematic approach, and by generalising the results of previous surveys targeting specific sub-topics of either SKE or SKI alone. More precisely, we analyse nearly 100 methods for SKE and approximately 80 methods for SKI, and categorise them according to their purpose, operation, expected input/output data and predictor types. Our taxonomies may be of interest for data scientists aiming at selecting the most adequate SKE/SKI method for their needs, and also work as suggestions for researchers interested in filling the gaps of the current state of the art, as well as for developers willing to implement SKE/SKI-based technologies.

(keywords) Logic ; Machine learning theory ; Hybrid symbolic-numeric methods ; Knowledge representation and reasoning 

Partita IVA: 01131710376 — Copyright © 2008–2023 APICe@DISI – PRIVACY