Meaning- and ontology-based technologies for high-precision language an information-processing computational systems
Introduction
The most deadly air disaster in history is the notorious collision of KLM 4805 and Pan Am 1736 on Tenerife in 1977 ALPA n.d.. The probable cause was of a linguistic, semantic nature: the ambiguity that led to a discrepancy in grasping the state of the world between the crew of the KLM plane and air traffic control (ATC). Most egregiously, while taxiing for take-off in thick fog, a KLM crew member used the non-standard phrase, We are at take off, in communication with ATC, who responded, OK. ATC took the phrase to mean that the KLM plane had taxied to the departure point and was merely informing the tower of this fact without intending to do anything further without an ATC command. The captain of the KLM plane took the ok as his clearance to take off, which he had been waiting for impatiently, or at least not as the explicit withholding of the command to take off. The KLM plane proceeded to take off while the Pan Am plane was still taxiing on the same runway, invisible to the KLM crew because of the fog. The planes collided with only a tenth of the roughly 600 people involved surviving, all on the Pan Am plane.
Detecting, and possibly avoiding, miscommunication, in particular in potentially fatal results as in aviation radio traffic, is a motivation for the approach to automatic comprehension and disambiguation of meaning that we propose in the bulk of the paper. It is well-known that non-native speakers of English in the aviation profession, mainly air traffic control and flight crews, are a potential source of error, often fatal ones [31]. But even when English is the native language of both speakers, or one speaker when he is interacting with an instrument or computer, and even when less is at stake than in air traffic communication, differing underlying world models or differing situational instantiations of such models can result in miscommunication. Regardless of English proficiency of the human partners in the interaction, they often do not detect that a statement is compatible with two models, i.e., ambiguous. But one, sometimes both, of the models is not compatible with the actual state of the world, and if acted upon leads to deadly results. Much worse situations occur when one of the communication partners is the computer that cannot understand what a natural language text means at all. Yet, most NLP research today is not meaning-based, and neither is CAE which deals with information in natural language or, as a matter of fact, in any other form. This is precisely what the paper addresses and proposes to change, suggesting a semi-mature and partially implemented technology as a means to achieve the goal of adding the real-semantic dimension to CAE work—not to replace the current state of affairs bur rather to complement and enrich the field.
Section snippets
Meaning in computer aided engineering applications
The ATC application is just an example of a computer-aided engineering application. Our understanding of CAE is that of any real-life application that is currently executed wholly or partially by humans and that needs to be automated to expedite, avoid errors, or optimize in any other way. Obviously, it includes any real-life applications that will be developed in the future. The main point of the paper is that any, repeat ANY CAE application stands to gain significantly from actual
Form and content
The main thrust of this paper is to demonstrate the essential importance of natural language understanding in computational systems for virtually any existing and certainly future applications. Be it machine translation, historically the earliest natural language processing application, now almost defunct, or text mining, or summarization, or more recently, e-discovery, which is the forensics of e-mail and other online documents as the highly regulated stage of litigation, it seems obvious that
Ontological semantic technology
We utilize a computational semantic approach called Ontological Semantic Technology (OST). The approach, historically based on the Ontological Semantics of the 1990s [23] and references there) and considerably revised since (see [15], [28], [33], [39], [37], [38], [36]), is a combination of static and dynamic resources as shown in Fig. 2. The centerpiece of OST is the language-independent semi-automatically constructed ontology consisting of concepts and relationships among them. Each supported
Implicit vs. explicit Information
We have addressed in Raskin et al. [29] and Taylor et al. [39], the ability to determine information about the speaker based on what they do not make explicit. Similarly, based on our world knowledge, we can guess the meaning of unfamiliar word or phrase [38], [36]. What we want to focus on here is that it is possible for the same information to be interpreted in multiple ways, depending on the various states of background knowledge that we have, or based on priming. We have to deal with the
Grain size and script verification
An ontology can be looked at as a 3-D representation of the world, where each concept is represented as a 2-D plane (see Fig. 3). The black lines represent is-a links, blue lines within the planes represent properties of a given concept. For example, concept C1 has 4 properties. Each child concept inherits properties of the parent, e.g. C11, C12 and C13 inherit the 4 properties of C1. In addition to the parent properties, each child has its own properties that are not shared with the parent.
Misunderstanding leading to disaster
Let us now return to aviation radio traffic with two fatal examples of miscommunication that was compatible with two interpretations of the same situation. For the example in the introduction as well the disaster described below, the lack of overlap in understanding of the state of the situation could have been detected by an automated system, monitoring the radio traffic and processing the meaning of the statements of the crews and ATC.
In the case of the Tenerife disaster, the ontological
Summary and conclusion
In this paper, we have tried to make a case for the necessity and feasibility of an ontology- and meaning-based approach for natural language understanding (NLU) computational systems as well as for information processing systems of non-linguistic kind. We have shown an example of a disaster that may happen when meaning is not accessed or accessed inadequately, for instance, without disambiguation. We discussed the very nature of meaning which tends to be taken for granted, misunderstood, and
References (42)
- AHFD, The American Heritage First Dictionary, Houghton Mifflin, Boston-New York,...
- ALPA (n.d.), Airline Pilots Association. n.d. Aircraft Accident Report, Pan American World Airways Boeing 747, N 737...
Linguistic Theories of Humor
(1994)- et al.
Script theory revis(it)ed: joke similarity and joke representation model
Humor
(1991) Indexical expressions
Mind
(1954)- et al.
Situations and Attitudes
(1983) - Tim Berners-Lee, Panel presentation, LinkedDataPlanet, Lotico Meet-Up Semantic Web, New York, NY, June 17,...
- T. Berners-Lee, J. Hendler, O. Lassila, The Semantic Web, Scientific American Magazine,...
- et al.
Readings in Knowledge Representation
(1985) - et al.
Knowledge Representation and Reasoning
(2004)
Meaning and Grammar: An Introduction to Semantics
Linguistic Semantics
Semantics in Generative Grammar
Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition
Effability and translation
Foundations of Statistical Natural Language Processing
Programs with common sense
Machine Learning
Cited by (10)
Towards machine translation of chinese complex structures
2020, Advances in Intelligent Systems and ComputingMethodology of designing computer ontology of subject discipline by future teachers-engineers
2019, CEUR Workshop ProceedingsCognitive imaging: Using knowledge representation for segmentation of MRA data
2018, International Journal of Cognitive Informatics and Natural IntelligenceCognitive imaging: Using knowledge representation for reliable segmentation of MR angiography data
2017, Proceedings of 2017 IEEE 16th International Conference on Cognitive Informatics and Cognitive Computing, ICCI*CC 2017Big data & analytics: An approach using audit experts' interviews
2017, Iberian Conference on Information Systems and Technologies, CISTI