CONNER: A Concurrent ILP Learner in Description Logic

Abstract

Machine Learning (ML) approaches can achieve impressive results, but many lack transparency or have difficulties handling data of high structural complexity. The class of ML known as Inductive Logic Programming (ILP) draws on the expressivity and rigour of subsets of First Order Logic to represent both data and models. When Description Logics (DL) are used, the approach can be applied directly to knowledge represented as ontologies. ILP output is a prime candidate for explainable artificial intelligence; the expense being computational complexity. We have recently demonstrated how a critical component of ILP learners in DL, namely, cover set testing, can be speeded up through the use of concurrent processing. Here we describe the first prototype of an ILP learner in DL that benefits from this use of concurrency. The result is a fast, scalable tool that can be applied directly to large ontologies.

Ontology Graph Embeddings and ILP for Financial Forecasting

“…We conjecture that the results will further improve as we increase the data set, something we have tried to limit here to keep the complexity of the learning task under control. We shall also consider transferring the learning task on to a concurrent learner that scales up better [1]. It is important to note the hybrid nature of the approach, in which an ontology is populated with data extracted from company reports through text processing, then the embeddings for the graph nodes are generated, and clustered, all in an unsupervised manner.…”

“…These combine to a create a directed graph with entities represented by Unique Resource Identifiers (URI). Machine learning algorithms specialised in the use of DL to represent data and models also exist [9,1].…”

“…Since that study, we have taken our work in two directions. We have been working on the use of an in-house learner from ontologies in description logic [1] to automate the search for suitable ontology-based hypotheses. However, this article focuses on another approach, namely, the use of graph embeddings as produced by the node2vec algorithm [4] to extract important properties of the graph nodes reflecting the content and topology of each local neighbourhood in an unsupervised way.…”

Pair Trading with an Ontology of SEC Financial Reports

“…From an abstract point of view, the content of an ontology is equivalent to a set of Description Logic (DL) axioms, which provides the semantics for query languages (such as SPARQL [8]) and reasoners [4]. Machine learning algorithms specialised in the use of DL to represent data and models also exist [6], [2]. In this article, the information extracted from financial reports is converted to an ontology, and hand-written queries are used to extract the data sets needed to test the two research hypotheses.…”

“…This can reduce the potential for profit as an increasing number of traders adopt the same strategy and crowd the same stock market positions. It is therefore tempting to consider the use of additional information in the process of selecting the pairs in order to focus on those that are more likely to (1) meet the necessary condition, and (2) retain that property over time.…”