Article Preview
TopIntroduction
Because and why co-operate in verbal reasoning: if someone asserts X because Y, it is reasonable to ask why X and get because Y. As such, X because Y is an act of induction—it builds knowledge; whereas, why X invokes an act of deduction—it uses that knowledge. This has several interesting properties. Firstly, causes are transitive: if X because Y, and Y because Z, then it follows that X because Z. Further, X because Y is a complex utterance because it contains two complete utterances, each of which must also make sense and are asserted in their own right. This paper describes the structures which support these concepts, and how they are mechanised by the language engine, Enguage.
EnguageTM is a Java library which achieves natural language understanding simply by defining arbitrary mappings between natural numbers (Wheatman, 2009, 2018b), representing the utterances and replies of a conversation. It is purely software based; it is not tied into any hardware product. However, such machine understanding is often met with a sceptical: but does it understand. The resolution to this—which is used by Enguage—was determined as reaction to utterance (Austin, 1962): if the reaction is as intended, which for Enguage involves some unequivocal reply, then it can be deemed to be understood. Achieving the required reply—felicity—is the aim of using Enguage.
Classical reasoning, of induction and deduction, was defined by Aristotle. However, the reasoning addressed by this paper was devised in (Peirce, 1955, Ch. 11, pp 150-156) and is divided into:
- •
Deduction, where given one or more premises, an argument is followed to reach a conclusion
- •
Induction, where the argument is created from the conclusion and premises
- •
Abduction, where possible premises can be ascertained from the argument and the conclusion
This reasoning has an analogy with Informatics. For example, a computer program is somewhat like a deductive process: given some inputs and a set of machine instructions, an output can be reached. Similarly, software engineering can be seen as an inductive process, given some inputs and outputs a set of machine instructions can be created. Following this software analogy, abduction is somewhat akin to software testing, in that it amounts to finding the appropriate input, given a known output and a set of machine instructions. It is this last property which this paper hopes to further illustrate.
This paper starts by supporting the claim that Enguage is a novel approach. It follows with a brief look at how language is understood in relation to concepts. It then details the concepts of why and because and the repertoires devised to support them, and how this is achieved by Enguage. It describes some necessary developments in its representation of language to achieve such reasoning. The results presented are the output from the why-because portion of the full Enguage test suite.
TopEnguage: Neither Siri Nor Alexa
The case for voice-driven software is clear: anyone who has difficulty with screens and keyboards deserves a better interface. With 370M mono-lingual Hindi speakers alone—a population larger than the US—who don’t have a keyboard (Gibbs, 2018), there is the need to by-pass traditional human-computer interfaces and to address the human-human interface. This endeavour is commonly seen as Artificial Intelligence, but this phrase is difficult to define. AI touches on Science Fiction, on the popular imagination, as well as being a catch-all term for several areas academic study such as, amongst others, Natural Language Processing—a wide field in its own right. As such, it might be suggested that Enguage is AI, because it processes text as natural language, and so it is 'up against' the likes of SIRI, Alexa and Google Assistant. This section briefly sketches this arena, distancing Enguage from both the term AI and conventional NLP.