Fintan Costello

Efficient Creativity: Constraint-Guided Conceptual Combination

Abstract This paper describes a theory that explains both the creativity and the efficiency of people’s conceptual combination. In the constraint theory, conceptual combination is controlled by three constraints of diagnosticity, plausibility, and informativeness. The constraints derive from the pragmatics of communication as applied to compound phrases. The creativity of combination arises because the constraints can be satisfied in many different ways. The constraint theory yields an algorithmic model of the efficiency of combination. The C 3 model admits the full creativity of combination and yet efficiently settles on the best interpretation for a given phrase. The constraint theory explains many empirical regularities in conceptual combination, and makes various empirically verified predictions. In computer simulations of compound phrase interpretation, the C 3 model has produced results in general agreement with people’s responses to the same phrases.

Applying computational models of spatial prepositions to visually situated dialog

This article describes the application of computational models of spatial prepositions to visually situated dialog systems. In these dialogs, spatial prepositions are important because people often use them to refer to entities in the visual context of a dialog. We first describe a generic architecture for a visually situated dialog system and highlight the interactions between the spatial cognition module, which provides the interface to the models of prepositional semantics, and the other components in the architecture. Following this, we present two new computational models of topological and projective spatial prepositions. The main novelty within these models is the fact that they account for the contextual effect which other distractor objects in a visual scene can have on the region described by a given preposition. We next present psycholinguistic tests evaluating our approach to distractor interference on prepositional semantics, and illustrate how these models are used for both interpretation and generation of prepositional expressions.

Surprisingly Rational: Probability Theory Plus Noise Explains Biases in Judgment

The systematic biases seen in peoples probability judgments are typically taken as evidence that people do not use the rules of probability theory when reasoning about probability but instead use heuristics, which sometimes yield reasonable judgments and sometimes yield systematic biases. This view has had a major impact in economics, law, medicine, and other fields; indeed, the idea that people cannot reason with probabilities has become a truism. We present a simple alternative to this view, where people reason about probability according to probability theory but are subject to random variation or noise in the reasoning process. In this account the effect of noise is canceled for some probabilistic expressions. Analyzing data from 2 experiments, we find that, for these expressions, peoples probability judgments are strikingly close to those required by probability theory. For other expressions, this account produces systematic deviations in probability estimates. These deviations explain 4 reliable biases in human probabilistic reasoning (conservatism, subadditivity, conjunction, and disjunction fallacies). These results suggest that peoples probability judgments embody the rules of probability theory and that biases in those judgments are due to the effects of random noise.

Investigating the Relations used in Conceptual Combination

How do people understand noun–noun compounds such as volcano science and pear bowl? In this paper, we present evidence against one approach to noun–noun compounds, namely that of arranging the meanings of compounds into a small, finite taxonomy of general semantic relations. Using a typical relation taxonomy, we conducted an experiment examining how people classify compounds into the taxonomy’s relation categories. We found that people often select not one but several relations for each compound; for example, people classify coffee stain as coffee MAKES stain, stain MADE OF coffee, coffee CAUSES stain and stain DERIVED FROM coffee. A natural metric for relational similarity follows from our experimental data; we found that using cluster analysis to group compounds’ interpretations with respect to this metric produced groupings that were different from the original taxonomic categories, suggesting that there is more than one way to classify the meanings of compounds. We also found that compounds which had similar constituent concepts tended to be interpreted with similar relations, indicating that the intrinsic properties of a compound’s constituent concepts help determine how that compound is interpreted. Such findings are problematic for taxonomic theories of conceptual combination

UCD-FC: Deducing semantic relations using WordNet senses that occur frequently in a database of noun-noun compounds

This paper describes a system for classifying semantic relations among nominals, as in SemEval task 4. This system uses a corpus of 2,500 compounds annotated with WordNet senses and covering 139 different semantic relations. Given a set of nominal pairs for training, as provided in the SemEval task 4 training data, this system constructs for each training pair a set of features made up of relations and WordNet sense pairs which occurred with those nominals in the corpus. A Naive Bayes learning algorithm learns associations between these features and relation membership categories. The identification of relations among nominals in test items takes place on the basis of these associations.

Conceptual Combination: A Theoretical Review

In the last decade, research on categorisation has been extended to deal with concept combination; the formation of complex concepts by combining two separate concepts (e.g., ‘red apple’, ‘striped orange’, ‘fake gun’). This research is important for two main reasons; first, it tests the adequacy of theories of categorisation, and second, it looks at how new concepts come into being. This paper provides a critical review of the main theories of conceptual combination. It shows how these theories have emerged from traditional theories on categorisation (e.g., classical and prototype theories) and assesses some of the trends that have emerged in the theoretical development of the area.

Computational modelling of switching behaviour in repeated gambles

We present a computational model which predicts people’s switching behaviour in repeated gambling scenarios such as the Iowa Gambling Task. This Utility-Caution model suggests that people’s tendency to switch away from an option is due to a utility factor which reflects the probability and the amount of losses experienced compared to gains, and a caution factor which describes the number of choices made consecutively in that option. Using a novel next-choice-prediction method, the Utility-Caution model was tested using two sets of data on the performance of participants in the Iowa Gambling Task. The model produced significantly more accurate predictions of people’s choices than the previous Bayesian expected-utility model and expectancy-valence model.

General and specific paraphrases of semantic relations between nouns

Many English noun pairs suggest an almost limitless array of semantic interpretation. A fruit bowl might be described as a bowl for fruit, a bowl that contains fruit, a bowl for holding fruit, or even (perhaps in a modern sculpture class), a bowl made out of fruit. These interpretations vary in syntax, semantic denotation, plausibility, and level of semantic detail. For example, a headache pill is usually a pill for preventing headaches, but might, perhaps in the context of a list of side effects, be a pill that can cause headaches (Levi, J. N. 1978. The Syntax and Semantics of Complex Nominals. New York: Academic Press.). In addition to lexical ambiguity, both relational ambiguity and relational vagueness make automatic semantic interpretation of these combinations difficult. While humans parse these possibilities with ease, computational systems are only recently gaining the ability to deal with the complexity of lexical expressions of semantic relations. In this paper, we describe techniques for paraphrasing the semantic relations that can hold between nouns in a noun compound, using a semi-supervised probabilistic method to rank candidate paraphrases of semantic relations, and describing a new method for selecting plausible relational paraphrases at arbitrary levels of semantic specification. These methods are motivated by the observation that existing semantic relation classification schemes often exhibit a highly skewed class distribution, and that lexical paraphrases of semantic relations vary widely in semantic precision.

A Model-Based Theory of Conceptual Combination

Traditionally, the meaning of a combined concept like “wooden spoon” has been treated compositionally; that is, the meaning of the combined concept is seen as being solely determined by the meaning of its constituents. However, in the last decade, research has shown that other knowledge — background, world knowledge — also plays a role in comprehending combined concepts. However, few models specify how this knowledge comes into play. The present paper proposes a theory of conceptual combination which demonstrates how world knowledge is accessed.

A comparison of word similarity measures for noun compound disambiguation

Noun compounds occur frequently in many languages, and the problem of semantic disambiguation of these phrases has many potential applications in natural language processing and other areas. One very common approach to this problem is to define a set of semantic relations which capture the interaction between the modifier and the head noun, and then attempt to assign one of these semantic relations to each compound. For example, the compound phrase flu virus could be assigned the semantic relation causal (the virus causes the flu); the relation for desert wind could be location (the wind is located in the desert). In this paper we investigate methods for learning the correct semantic relation for a given noun compound by comparing the new compound to a training set of hand-tagged instances, using the similarity of the words in each compound. The main contribution of this paper is to directly compare distributional and knowledge-based word similarity measures for this task, using various datasets and corpora. We find that the knowledge based system provides a much better performance when adequate training data is available.