Integration of World Knowledge for Natural Language Understanding

Abstract

Aims and scope of the seriesThis series publishes books resulting from theoretical research on and reproductions of general Artificial Intelligence (AI). The book series focuses on the establishment of new theories and paradigms in AI. At the same time, the series aims at exploring multiple scientific angles and methodologies, including results from research in cognitive science, neuroscience, theoretical and experimental AI, biology and from innovative interdisciplinary methodologies.All books in this series are co-published with Springer.For more information on this series and our other book series, please visit our website at:www.atlantis-press.comForeword Inference-based natural language understanding (NLU) was a thriving area of research in the 1970s and 1980s. It resulted in good theoretical work and in interesting small-scale systems. But in the early 1990s it foundered on three difficulties:• Parsers were not accurate enough to produce predicate-argument relations reliably, so that inference had no place to start.• Inference processes were not efficient enough nor accurate enough.• There was no large knowledge base designed for NLU applications.The first of these difficulties has been overcome by progress in statistical parsing. The second problem is one that many people, including Ekaterina Ovchinnikova, are working on now. The research described in this volume addresses the third difficulty, and indeed shows considerable promise in overcoming it. For this reason, I believe Dr. Ovchinnikova's work has a real potential to reignite interest in inference-based NLU in the computational linguistics community.A key notion in her work is that there already exists sufficient world knowledge in a variety of resources, at a level of precision that enables their translation into formal logic. To my mind, the most important of these are WordNet and FrameNet, especially the latter, and she describes the kind of information one can get out of these resources. She exploits in particular the hierarchical information and the glosses in WordNet, generating 600,000 axioms.She also describes how one can utilize FrameNet to generate about 50,000 axioms representing relations between words and frames, and about 5000 axioms representing frameframe relations. Her analysis of FrameNet is quite thorough, and I found this part of her work inspiring.She also critically discusses foundational ontologies such as DOLCE, SUMO. and Open-Cyc, and domain-specific ontologies of the sort being constructed for the Semantic Web. v vi Integration of World Knowledge for Natural Language UnderstandingShe examines the problems raised by semi-formal ontologies, like YAGO and ConceptNet, which have been gleaned from text or Netizens and which may be more difficult to translate into formal logic. She also shows how to use distributional data for a default mode of processing when the required knowledge is not available.Her use of knowledge from a variety of resources, combined into a single system, leads to the very hard problem of ensuring consisten...

Robot perception errors and human resolution strategies in situated human–robot dialogue

“…For example, integrating world-knowledge [32] and/or linguistic ontological knowledge [3]; integrating spatial semantics into a compositional/attentional accounts of reference [23,24,31]; learning spatial semantics directly from sensor data using machine learning techniques [12,34]; modelling the functional aspects of spatial semantics in terms of predicting the dynamics of objects in the scene [10,42]; capturing the vagueness and gradation of spatial semantics [17,22,43]; and leveraging analogical reasoning mechanisms to enable agents to apply spatial semantics to new environments [13].…”

Empirical methods for computing phrasal and sentential semantics in Vietnamese

“…• Clauses from (1) to (12) are DCG clauses to describe number words in quantity phrases. We can define many DCG clauses for representing the semantic form of the number word with the following form: "num_word(NW) [NW].…”

“…Each semantic expression represents an event of sentence [1] which is knowledge in DataKnowledge Base of the VietQAS. The semantic expressions can be called a logical form [10][11][12] of NLU represented as a semantic tree form (Figs. 1, 2).…”

Multi-purpose natural language understanding linked to sensorimotor experience in humanoid robots

“…Different syntactic realizations of each predicate for each verb (e.g., from X, in X, out of X) were derived from syntactic patterns specified in FrameNet that were linked to the corresponding FrameNet roles. See [18] for more details on the generation of lexical axioms. A simple spatial axiom was added to reason about locations, which states that if an object is located at a part of a location (corner, top, side, etc.…”