Pathway Computation in Models Derived from Bio-Science Text Sources

Abstract

This paper outlines a system, OntoScape, serving to accomplish complex inference tasks on knowledge bases and bio-models derived from life-science text corpora. The system applies so-called natural logic, a form of logic which is readable for humans. This logic affords ontological representations of complex terms appearing in the text sources. Along with logical propositions, the system applies a semantic graph representation facilitating calculation of bio-pathways. More generally, the system affords means of query answering appealing to general and domain specific inference rules.

Deductive Querying of Natural Logic Bases

“…It appears only in deduced sentences by way of the closed world assumption. In [Andreasen et al, 2018] we discuss alternative quantifiers to the given defaults (every|some|no) Cnoun V erb (some|every) Cnoun…”

“…The original individual propositions remain unaffected by this compilation. Inference rules for dynamically creating these new concepts and fitting them into the ontology are described in our [Andreasen et al, 2018].…”

“…[Andreasen and Nilsson, 2004]. Furthermore, NaturaLog bears some affinity to description logic as discussed in [Andreasen et al, 2018]. However, as one noteworthy difference, NaturaLog abandons forced use of copula sentences in favor of sentences with free choice of transitive verbs from the considered knowledge base domain.…”

“…The present paper focuses on query functionalities for NaturaLog knowledge bases and follows up on our former publications on the application of natural logic in [Andreasen et al, 2017a], [Andreasen et al, 2015], [Andreasen et al, 2017b], . The details of the syntax and the formal semantics of the Nat-uraLog logic are elaborated in the coming paper [Andreasen et al, 2018] that also specifies NaturaLog in terms of predicate logic and introduces the concomitant graph conception of NaturaLog knowledge bases.…”

On the Relationship between a Computational Natural Logic and Natural Language

“…Individual entities may, however, if necessary, be dealt with as stipulated singleton classes (having no subclasses in so far as the empty class is left out in our setup). This basis, while being application-neutral at the outset, appears to be particularly useful for applications within the bio-sciences as discussed in (Smith et al, 2005;Schulz and Hahn, 2004) as well as in our (Andreasen et al, 2014a;Andreasen et al, 2014b;Andreasen and Nilsson, 2014;Nilsson, 2015;Andreasen et al, 2015). General, natural language descriptions in natural sciences abound with classes, let us just mention the Linnean and chemical and medical taxonomies.…”