Achim G. Hoffmann

Incremental acquisition of search knowledge

The development of highly effective heuristics for search problems is a difficult and time-consuming task. We present a knowledge acquisition approach to incrementally model expert search processes. Though, experts do not normally have complete introspective access to that knowledge, their explanations of actual search considerations seem very valuable in constructing a knowledge-level model of their search processes.Furthermore, for the basis of our knowledge acquisition approach, we substantially extend the work done on Ripple-down rules which allows knowledge acquisition and maintenance without analysis or a knowledge engineer. This extension allows the expert to enter his domain terms during the KA process; thus the expert provides a knowledge-level model of his search process. We call this framework nested ripple-down rules.Our approach targets the implicit representation of the less clearly definable quality criteria by allowing the expert to limit his input to the system to explanations of the steps in the expert search process. These explanations are expressed in our search knowledge interactive language. These explanations are used to construct a knowledge base representing search control knowledge. We are acquiring the knowledge in the context of its use, which substantially supports the knowledge acquisition process. Thus, in this paper, we will show that it is possible to build effective search heuristics efficiently at the knowledge level. We will discuss how our system SmS1.3 (SmS for Smart Searcher) operates at the knowledge level as originally described by Newell. We complement our discussion by employing SmS for the acquisition of expert chess knowledge for performing a highly pruned tree search. These experimental results in the chess domain are evidence for the practicality of our approach.

BioOntoVerb: A top level ontology based framework to populate biomedical ontologies from texts

The Semantic Web can be conceived as an extension of the current Web where information is given well-defined meaning. In this scenario ontologies are crucial since they provide meaning and facilitate the search for contents and information. Ontology population is a knowledge acquisition activity used to transform data sources into instance data. The instantiation of ontologies with new knowledge is an important step towards the provision of valuable ontology-based services. In this paper, we present a methodology to be used for ontology population. For it, top level ontologies that define the basic semantic relations in biomedical domains are mapped onto semantic role labelling resources, where every semantic role defines the role of a verbal argument in the event expressed by the verb. The modular architecture employed in our work gives the system a high versatility, as resources have been developed separately and they can be easily adapted to most biomedical domain ontologies.

NRDR for the Acquisition of Search Knowledge

The contribution of this paper is three-fold: It substantially extends Ripple Down Rules, a proven effective method for building large knowledge bases without a knowledge engineer. Furthermore, we propose to develop highly effective heuristics searchers for combinatorial problems by a knowledge acquisition approach to acquire human search knowledge. Finally, our initial experimental results suggest, that this approach may allow experts to stepwise articulate their introspectively inaccessible knowledge.

Theoretical basis for hierarchical incremental knowledge acquisition

Human experts tend to introduce intermediate terms in giving their explanations. The experts explanation of such terms is operational for the context that triggered the explanation; however, term definitions remain often incomplete. Further, the experts (re) use of these terms is hierarchical (similar to natural language). In this paper, we argue that a hierarchical incremental knowledge acquisition (KA) process that captures the expert terms and operationalizes them while incompletely defined makes the KA task more effective. Towards this we present our knowledge representation formalism Nested Ripple Down Rules (NRDR) that is a substantial extension to the (Multiple Classification) Ripple Down Rule (RDR) KA framework. The incremental KA process with NRDR as the underlying knowledge representation has confirmation holistic features. This allows simultaneous incremental modelling and KA and eases the knowledge base (KB) development process.Our NRDR formalism preserves the strength of incremental refinement methods, that is the ease of maintenance of the KB. It also addresses some of their shortcomings: repetition, lack of explicit modelling and readability. KBs developed with NRDR describe an explicit model of the domain. This greatly enhances the reuseability of the acquired knowledge.This paper also presents a theoretical framework for analysing the structure of RDR in general and NRDR in particular. Using this framework, we analyse the conditions under which RDR converges towards the target KB. We discuss the maintenance problems of NRDR as a function of this convergence. Further, we analyse the conditions under which NRDR offers an effective approach for domain modelling. We show that the maintenance of NRDR requires similar effort to maintaining RDR for most of the KB development cycle. We show that when an NRDR KB shows an increase in maintenance requirement in comparison with RDR during its development, this added requirement can be automatically handled using stored past seen cases.

A New Approach for Scientific Citation Classification Using Cue Phrases

This paper introduces a new method for the rapid development of complex rule bases involving cue phrases for the purpose of classifying text segments. The method is based on Ripple-Down Rules, a knowledge acquisition method that proved very successful in practice for building medical expert systems and does not require a knowledge engineer. We implemented our system KAFTAN and demonstrate the applicability of our method to the task of classifying scientific citations. Building cue phrase rules in KAFTAN is easy and efficient. We demonstrate the effectiveness of our approach by presenting experimental results where our resulting classifier clearly outperforms previously built classifiers in the recent literature.

Acquisition of Search Knowledge

The development of highly effective heuristics for search problems is a difficult and time-consuming task. We present a knowledge acquisition approach to incrementally model expert search processes. Though, experts do not normally have introspective access to that knowledge, their explanations of actual search considerations seems very valuable in constructing a knowledge level model of their search processes. The incremental method was inspired by the work on Ripple-Down Rules which allows knowledge acquisition and maintenance without analysis or a knowledge engineer. We substantially extend Ripple Down Rules to allow undefined terms in the conditions. These undefined terms in turn become defined by Ripple Down Rules. The resulting framework is called Nested Ripple Down Rules. Our system SmS1.2 (SmS for Smart Searcher), has been employed for the acquisition of expert chess knowledge for performing a highly pruned tree search. Our first experimental results in the chess domain are evidence for the validity of our approach, even though a number of the planned features are still under development.

The dynamic locking heuristic-a new graph partitioning algorithm

In layout design finding efficiently good solutions to the hypergraph bipartitioning problem is of great importance. This paper introduces a new algorithm, the dynamic locking algorithm, which has empirically shown a significant improvement in the partitioning result compared to known algorithms.<<ETX>>

Simultaneous Modelling and Knowledge Acquisition Using NRDR

Incremental refinement methods of knowledge bases ease maintenance but fail to uncover the underlying domain model used by the expert. In this paper, we propose a new knowledge representation formalism for incremental acquisition and refinement of knowledge. It guides the expert in expressing his model of the domain during the actual knowledge acquisition process. This knowledge representation scheme, Nested Ripple Down Rules, is a substantial extension to Ripple Down Rule (RDR) knowledge acquisition framework. This paper introduces a theoretical framework for analysing the structure of RDR in general and NRDR in particular. Using this framework we analyse the conditions under which RDR converges towards the target knowledge base. Further, we analyse the conditions under which NRDR offers an effective approach for domain modelling. We discuss the maintenance problems of NRDR as a function of this convergence. We show that the maintenance of NRDR requires similar effort to maintaining RDR for most of the knowledge base development cycle. We show that when an NRDR knowledge base shows an increase in maintenance requirement in comparison with RDR during its development. this added requirement can be automatically handled.

Citation based summarisation of legal texts

This paper presents an approach towards using both incoming and outgoing citation information for document summarisation. Our work aims at generating automatically catchphrases for legal case reports, using, beside the full text, also the text of cited cases and cases that cite the current case. We propose methods to use catchphrases and sentences of cited/citing cases to extract catchphrases from the text of the target case. We created a corpus of cases, catchphrases and citations, and performed a ROUGE based evaluation, which shows the superiority of our citation-based methods over full-text-only methods.

RDRCE: combining machine learning and knowledge acquisition

We present a new interactive workbench RDRCE (RDR Case Explorer) to facilitate the combination of Machine Learning and manual Knowledge Acquisition for Natural Language Processing problems. We show how to use Brills well regarded transformational learning approach and convert its results into an RDR tree. RDRCE then strongly guides the systematic inspection of the generated RDR tree in order to further refine and improve it by manually adding more rules. Furthermore, RDRCE also helps in quickly recognising potential noise in the training data and allows to deal with noise effectively. Finally, we present a first study using RDRCE to build a high-quality Part-of-Speech tagger for English. After some 60 hours of manual knowledge acquisition, we already exceed slightly the state-of-the art performance on unseen benchmark test data and the fruits of some 15 years of further research in learning methods for Part-of-Speech taggers.