This thesis is an inquiry into the nature of the high-level, rhetorical structure of unrestricted natural language texts, computational means to enable its derivation, and two applications (in automatic summarization and natural language generation) that follow from the ability to build such structures automatically. The thesis proposes a first-order formalization of the high-level, rhetorical structure of text. The formalization assumes that text can be sequenced into elementary units; that discourse relations hold between textual units of various sizes; that some textual units are more important to the writer's purpose than others; and that trees are a good approximation of the abstract structure of text. The formalization also introduces a linguistically motivated compositionality criterion, which is shown to hold for the text structures that are valid. The thesis proposes, analyzes theoretically, and compares empirically four algorithms for determining the valid text structures of ...

Abstract not found

This paper presents an approach to learning from situated, interactive tutorial instruction within an ongoing agent. Tutorial instruction is a flexible (and thus powerful) paradigm for teaching tasks because it allows an instructor to communicate whatever types of knowledge an agent might need in whatever situations might arise. To support this flexibility, however, the agent must be able to learn multiple kinds of knowledge from a broad range of instructional interactions. Our approach, called situated explanation, achieves such learning through a combination of analytic and inductive techniques. It combines a form of explanation-based learning that is situated for each instruction with a full suite of contextually guided responses to incomplete explanations. The approach is implemented in an agent called Instructo-Soar that learns hierarchies of new tasks and other domain knowledge from interactive natural language instructions. Instructo-Soar meets three key requirements of flexible...

This paper shows how agents' choice in communicative action can be designed to mitigate the effect of their resource 1/mits in the context of particular features of a collaborative planning task. I first motivate a number of hypotheses about effective language behavior based on a statistical analysis of a corpus of natural collaborative planning dialogues. These hypotheses are then tested in a dialogue testbed whose design is motivated by the corpus analysis. Experiments in the testbed examine the interaction between (1) agents' resource 1/mits in attentional capacity and inferential capacity; (2) agents' choice in communication; and (3) features of communicative tasks that affect task difficulty such as inferential complexity, degree of belief coordination required, and tolerance for errors. The results show that good algorithms for communication must be defined relative to the agents' resource 1/mits and the features of the task. Algorithms that are inefficient for inferentially simple, low coordination or fault-tolerant tasks are effective when tasks require coordination or complex inferences, or are fault-intolerant. The results provide an explanation for the occurrence of utterances in human dialogues that, prima facie, appear inefficient, and provide the basis for the design of effective algorithms for communicative choice for resource limited agents.

In this paper we discuss a study of the expression of procedural relations in multilingual user instructions in particular the relations of Generation and Enablement. These procedural relations are defined in terms of a plan representation model, and applied in a corpus study of English, French, and Portuguese instructions. The results of our analysis indicate specific guidelines for the tactical realisation of expressions of these relations in multilingual instructional text.

In this paper, we present a reinforcement learning approach for mapping natural language instructions to sequences of executable actions. We assume access to a reward function that defines the quality of the executed actions. During training, the learner repeatedly constructs action sequences for a set of documents, executes those actions, and observes the resulting reward. We use a policy gradient algorithm to estimate the parameters of a log-linear model for action selection. We apply our method to interpret instructions in two domains — Windows troubleshooting guides and game tutorials. Our results demonstrate that this method can rival supervised learning techniques while requiring few or no annotated training examples. 1 1

In contrast to current intelligent systems, which must be laboriously programmed for each task they are meant to perform, instructable agents can be taught new tasks and associated knowledge. This thesis presents a general theory of learning from tutorial instruction and its use to produce an instructable agent. Tutorial instruction is a particularly powerful form of instruction, because it allows the instructor to communicate whatever kind of knowledge a student needs at whatever point it is needed. To exploit this broad flexibility, however, a tutorable agent must support a full range of interaction with its instructor to learn a full range of knowledge. Thus, unlike most machine learning tasks, which target deep learning of a single kind of knowledge from a single kind of input, tutorability requires a breadth of learning from a broad range of instructional interactions. The theory of learning from tutorial...

Following verbal route instructions requires knowledge of language, space, action and perception. We present MARCO, an agent that follows free-form, natural language route instructions by representing and executing a sequence of compound action specifications that model which actions to take under which conditions. MARCO infers implicit actions from knowledge of both linguistic conditional phrases and from spatial action and local configurations. Thus, MARCO performs explicit actions, implicit actions necessary to achieve the stated conditions, and exploratory actions to learn about the world. We gathered a corpus of 786 route instructions from six people in three large-scale virtual indoor environments. Thirtysix other people followed these instructions and rated them for quality. These human participants finished at the intended destination on 69 % of the trials. MARCO followed the same instructions in the same environments, with a success rate of 61%. We measured the efficacy of action inference with MARCO variants lacking action inference: executing only explicit actions, MARCO succeeded on just 28 % of the trials. For this task, inferring implicit actions is essential to follow poor instructions, but is also crucial for many highly-rated route instructions.

This paper addresses this issue in the context of the expression of procedural relations between actions in instructional text. It employs the following four step approach to achieve this goal: (1) Collect a corpus of the relevant text type; (2) Perform a detailed linguistic study of a portion of this corpus, called the training set, and reserving the remainder as a testing set; (3) Implement the results of this study in a text generation system; (4) Compare the output of the system with the text found in the entire corpus. This has resulted in the construction of IMAGENE, an instructional text generation system which embodies a model of the forms of expression consistently used by instructional text writers over a broad range of instruction types. The details of IMAGENE's

It has long been noted that Natural Language utterances can communicate more than their conventional meaning (Grice, 1975). It has also been noted that behaving appropriately in response to instructions given in Natural Language requires understanding more than their conventional meaning (Suppes and Crangle, 1988; Webber and Di Eugenio, 1990; Webber et al., 1992). This paper addresses one mechanism by which speakers convey, and hearers derive, such additional aspects of meaning -- a mechanism we call pragmatic overloading. In pragmatic overloading, a clause interpreted as conveying directly or indirectly the goal fi of an action ff which is described by some other clause, forms the basis of constrained inference that leads to additional information about the action ff. The paper demonstrates pragmatic overloading through a variety of clausal adjuncts. We then discuss a framework that supports many of the inferences that pragmatic overloading gives rise to. This framework integrates a ...