Computational Semantics in Discourse: Underspecification, Resolution, and Inference

In this paper I introduce a formalism for natural language understandingbased on a computational implementation of Discourse RepresentationTheory. The formalism covers a wide variety of semantic phenomena(including scope and lexical ambiguities, anaphora and presupposition),is computationally attractive, and has a genuine inference component. Itcombines a well-established linguistic formalism (DRT) with advancedtechniques to deal with ambiguity (underspecification), and isinnovative in the use of first-order theorem proving techniques.The architecture of the formalism for natural language understandingthat I advocate consists of three levels of processing:underspecification, resolution, andinference. Each of these levels has a distinct function andtherefore employs a different kind of semantic representation. Themappings between these different representations define the interfacesbetween the levels.I show how underspecified semantic representations can be built in acompositional way (for a fragment of English Grammar) using standardtechniques borrowed from the -calculus, how inferences can becarried out on discourse representations using a translation tofirst-order logic, and how existing research prototypes (discourseprocessing and spoken-dialogue systems) implement the formalism.     

Focus interpretation in Thetic statements: Alternative Semantics and Optimality Theory Pragmatics

Broad focus (or informational integration or nonautonomy) is lexically and contextually constrained, but these constraints are not well understood. On a standard theory of focus interpretation, the presupposition of a broad focus is verified whenever those of two narrow foci are. I argue that to account for cases where two narrow foci are preferred, it is necessary to assume that broad focus competes with two narrow foci and implicates the opposite of what they presuppose. Central constraints on thetic statements are thus accounted for in an Optimality Theory (OT) enriched Alternative Semantics.     

Logical reasoning in natural language: It is all about knowledge

A formal, computational, semantically clean representation of natural language is presented. This representation captures the fact that logical inferences in natural language crucially depend on the semantic relation of entailment between sentential constituents such as determiner, noun, adjective, adverb, preposition, and verb phrases.The representation parallels natural language in that it accounts for human intuition about entailment of sentences, it preserves its structure, it reflects the semantics of different syntactic categories, it simulates conjunction, disjunction, and negation in natural language by computable operations with provable mathematical properties, and it allows one to represent coordination on different syntactic levels.The representation demonstrates that Boolean semantics of natural language can be successfully modeled in terms of representation and inference by knowledge representation formalisms with Boolean semantics. A novel approach to the problem of automatic inferencing in natural language is addressed. The algorithm for updating a computer knowledge base and reasoning with explicit negative, disjunctive, and conjunctive information based on computing subsumption relation between the representations of the appropriate sentential constituents is discussed with examples.     

Addressed Term Rewriting Systems: Syntax, Semantics, and Pragmatics: Extended Abstract

We present a formalism called Addressed Term Rewriting Systems, which can be used to define the operational semantics of programming languages, especially those involving sharing, recursive computations and cyclic data structures. Addressed Term Rewriting Systems are therefore well suited for describing object-based languages, as for instance the family of languages called , involving both functional and object-based features.     

Evaluation of Localized Semantics: Data, Methodology, and Experiments

We present a new data set of 1014 images with manual segmentations and semantic labels for each segment, together with a methodology for using this kind of data for recognition evaluation. The images and segmentations are from the UCB segmentation benchmark database (Martin et al., in International conference on computer vision, vol. II, pp. 416–421, 2001). The database is extended by manually labeling each segment with its most specific semantic concept in WordNet (Miller et al., in Int. J. Lexicogr. 3(4):235–244, 1990). The evaluation methodology establishes protocols for mapping algorithm specific localization (e.g., segmentations) to our data, handling synonyms, scoring matches at different levels of specificity, dealing with vocabularies with sense ambiguity (the usual case), and handling ground truth regions with multiple labels. Given these protocols, we develop two evaluation approaches. The first measures the range of semantics that an algorithm can recognize, and the second measures the frequency that an algorithm recognizes semantics correctly. The data, the image labeling tool, and programs implementing our evaluation strategy are all available on-line (kobus.ca//research/data/IJCV_2007). We apply this infrastructure to evaluate four algorithms which learn to label image regions from weakly labeled data. The algorithms tested include two variants of multiple instance learning (MIL), and two generative multi-modal mixture models. These experiments are on a significantly larger scale than previously reported, especially in the case of MIL methods. More specifically, we used training data sets up to 37,000 images and training vocabularies of up to 650 words. We found that one of the mixture models performed best on image annotation and the frequency correct measure, and that variants of MIL gave the best semantic range performance. We were able to substantively improve the performance of MIL methods on the other tasks (image annotation and frequency correct region labeling) by providing an appropriate prior.     

虚拟环境中装配设计语义的表达、传递与转化研究

研究了虚拟环境产品装配建模过程中装配语义的表达、装配语义与装配约束的转化技术,提出了采用语义－约束图对装配设计语义与约束进行维护。通过语义－约束图可以从语义层次和约束层次对产品装配模型进行编辑和维护。提出了采用模糊参量表达和处理产品装配设计中的模糊语义。基于模糊参量的语义表达不仅为设计信息输入提供了更大的自由度,而且扩展了计算机对模糊信息的处理能力,有利于产品设计意图的维护。通过产品装配信息从抽象到具体、从模糊到精确的转化,实现了虚拟装配设计系统对抽象、模糊设计信息的表达、传递与处理。     

不确定信息的认知结构表示、推理和学习

提出了一个对不确定信息进行表示、推理和学习的统一框架.通过引入4-值认知结构刻画不确定性,提供了更强的度量能力和更高的推理效率,同时支持相关不确定信息的有效获取,从而更加接近实际应用的需要.     

iFind: 一个结合语义和视觉特征的图像相关反馈检索系统

给出了一个结合语义与视觉特征信息的图像相关反馈检索系统－－iFind。系统通过图像的标注信息构造语义网络，并在相关反馈中与图像的视觉特征相结合，有效地实现了在两个层次上的相关反馈，在基于内容的图像检索中取得了较为理想的效果，具有一定的应用价值。     

Income, Time Effects and Direct Preferences in a Multimodal Choice Context: Application of Mixed RP/SP Models with Non-Linear Utilities

Transport problems typically involve at least two types of constraints, on income and on time. Therefore, the indirect utility function depends either on the income available after having subtracted the cost of the discrete alternative and on the free time left after having worked and travelled by each competing option. In the typical linear-in-the-attributes and in-the-parameters specification, that represents the first grade approximation of the indirect utility function, the effect of income and time constraints cancel out and only the cost and time of the alternatives matter in the comparison between them. From a microeconomic point of view this is equivalent to assume that income and time effects could be disregarded; which is not always the case. To account for these effects the utility function should include second order attributes; however, in non-linear utility functions it may not be easy to distinguish among several effects that could be relevant: direct preferences for good and leisure, and simple interactions between attributes other than income and time effects. This paper analyses these effects from a theoretical point of view focusing on the possible confounding problem in detecting income and time effects. We use a dataset collected for a modal choice context and containing both revealed and stated preference data, and estimate several NL models examining the effect of the different second-order terms on detecting income and time effects. We compared specifications including square cost and time attributes, interactions between time and cost, cost divided by the income available to be spent on free time, and time multiplied by free time. Our results confirm the strong effect of direct preferences for goods and leisure time on choice, and the potential confounding effect between quadratic attributes and other non-linear omitted terms. Finally, we also found that care should be taken in highlighting income and time effects using mixed data sources, since confounding effects can occur when non-linearities are accounted for in both data sets.     

Integrating scientific modeling and supporting dynamic hazard management with a GeoAgent-based representation of human–environment interactions: A drought example in Central Pennsylvania, USA

Recent natural disasters indicate that modern technologies for environmental monitoring, modeling, and forecasting are not well integrated with cross-level social responses in many hazard-management systems. This research addresses this problem through a Java-based multi-agent prototype system, GeoAgent-based Knowledge System (GeoAgentKS). This system allows: (1) computer representation of institutional regulations and behavioral rules used by multiple social institutions and individuals in cross-level human–environment interactions, (2) integration of this representation with scientific modeling of dynamic hazard development, and (3) application of automated reasoning that suggests to users the appropriate actions for supporting cooperative social responses. This paper demonstrates the software architecture of GeoAgentKS and presents such an integrated approach by modeling the drought management processes in Central Pennsylvania, USA. The results show that it is possible to use GeoAgentKS to represent multilevel human–environment interactions and to use those interactions as input to decision making in hazard management.