Natural language generation (NLG) systems are computer software systems that pro- duce texts in English and other human languages, often from non-linguistic input data.

We are developing technology for generating English textual summaries of time-series data, in three domains: weather forecasts, gas-turbine sensor readings, and hospital intensive care data. Our weather-forecast generator is currently operational and being used daily by a meteorological company. We generate summaries in three steps: (a) selecting the most important trends and patterns to communicate; (b) mapping these patterns onto words and phrases; and (c) generating actual texts based on these words and phrases. In this paper we focus on the first step, (a), selecting the information to communicate, and describe how we perform this using modified versions of standard data analysis algorithms such as segmentation. The modifications arose out of empirical work with users and domain experts, and in fact can all be regarded as applications of the Gricean maxims of Quality, Quantity, Relevance, and Manner, which describe how a cooperative speaker should behave in order to help a hearer correctly interpret a text. The Gricean maxims are perhaps a key element of adapting data analysis algorithms for effective communication of information to human users, and should be considered by other researchers interested in communicating data to human users.

We are investigating techniques for producing textual summaries of time series data. Deep reasoning techniques have proven impractical because we lack perfect knowledge about users and their tasks. Data analysis techniques such as segmentation are more attractive, but they have been developed for data mining, not for communication. We examine how segmentation should be modified to make it suitable for generating textual summaries. Our algorithm has been implemented in a weather forecast generation system.

The SUMTIME project aims to develop better techniques for producing natural language summaries of time-series data. The initial phase of the project has focused on understanding how human experts perform this task, via knowledge acquisition and corpus analysis exercises. This has led to a number of observations, of which the most important is that producing human like summaries involves as much processing of domain knowledge as that of communicative and linguistic knowledge. Summarisation is not simply a verbalisation of how a set of numbers changes over time. Instead it is the process in which the author adds value to the raw data using his expertise and presents exactly the information that is relevant to the end-user. 1.

A great part of the product knowledge in manufacturing enterprises is only available in the form of natural language documents. The know-how recorded in these documents is an essential resource for successful competition in the market. From the viewpoint of knowledge management, however, documents have a severe limitation: They do not capture the wealth of knowledge contained in these documents, since the entire knowledge is not spelled out on the linguistic surface. In order to overcome this limitation, the notion of a document as a particular kind of realization of (or view on) the underlying knowledge is introduced. The paper discusses the major steps in realizing this approach to documents: Knowledge acquisition, knowledge representation, and techniques to automatically generate multilingual documents from knowledge bases. Further, the paper describes how the required product knowledge can be represented in a sharable and reusable way.

Measured inputs in control domains are often continuous. A discretization function is needed to map continuous inputs into multiple intervals or ranges of input values, so that they can be used as symbolic inputs to a rule-based system. The discretization parameters used to determine each interval play a critical role in the overall effectiveness of a rule-based system. This paper introduces a Hybrid Intelligent Architecture (HIA) that exploits the complementary features of expert systems and connectionist architectures to revise the initial domain knowledge and enhance its input characterization. HIA has three building blocks: a KnowledgeBased module, a Statistical module and a Connectionist Architecture module. A well defined format is used to describe the initial knowledge acquired from the application domain in a rule-based format and to enable its mapping into a uniform, three layer network. The statistical module updates both the rule-based and the connectionist subsystems by obs...

. Interviewing experts in order to elicit the knowledge they use in solving problems is a very common task in the knowledge acquisition phase. In practice, interviewing experts is a very tedious and costly task. Surprisingly, up to now there is very little technological support for performing interviews widely available. Recent developments in new hardware and software now provides a significantly better technological base to fill this gap. The Knowledge Mining Center is an easy-to-use, yet very flexible tool for interviewing. It allows for better presentation of information during interviews by using multimedia technology and online access to information such as design documents or protocols of earlier interviews. Its main advantage, however, is its capability to semiautomatically capture, segment, and annotate information verbally communicated during interviews. The paper motivates the need for technological support for interviewing, discusses design considerations for interviewing t...

Abstract not found

I hereby declare that I am the sole author of this thesis. This is a true copy of the thesis, including any required final revisions, as accepted by my examiners. I understand that my thesis may be made electronically available to the public. ii An automatically trained, statistically based, fuzzy inference system that functions as a classifier is produced. The hybrid system is designed specifically to be used as a decision support system. This hybrid system has several features which are of direct and immediate utility in the field of decision support, in-cluding a mechanism for the discovery of domain knowledge in the form of explanatory rules through the examination of training data; the evaluation of such rules using a simple probabilistic weighting mech-anism; the incorporation of input uncertainty using the vagueness abstraction of fuzzy systems; and the provision of a strong confidence measure to predict the probability of system failure. Analysis of the hybrid fuzzy system and its constituent parts allows commentary on the weighting scheme and performance of the “Pattern Discovery ” system on which it is based. Comparisons against other well known classifiers provide a benchmark of the performance of the

........................................................................................................................................xi Abbreviations ................................................................................................................................xii 1 INTRODUCTION ..................................................................................................................1 1.1 Motivation for the Research ..........................................................................................1 1.2 Overview of the Research Problems ..............................................................................2 1.3 Thesis Objectives and Considerations ...........................................................................6 1.4 Thesis Organisation ......................................................................................................12 1.5 Summary ......................................................................................