Witschel, Hans Friedrich
E-Mail-Adresse
Geburtsdatum
Suchergebnisse
A new Retrieval Function for Ontology-Based Complex Case Descriptions
This work focuses on case-based reasoning in domains where cases have complex structures with relationships to an arbitrary number of other (potentially complex and structured) entities and where case characterisations (queries) are potentially incomplete. We summarise the requirements for such domains in terms of case representation and retrieval functions. We then analyse properties of existing similarity measures used in CBR { above all symmetry { and argue that some of these properties are not desirable. By exploiting analogies with retrieval functions in the area of information retrieval { where similar functions have been replaced by new ones not exhibiting the aforementioned undesired properties { we derive a new asymmetric ranking function for case retrieval. On a generated test-bed, we show that indeed the new function results in di erent ranking of cases { and use testbed examples to illustrate why this is desirable from a user's perspective.
Breaking free from your information prison - A recommender based on semantically enriched context descriptions
Information repositories, implemented as Enterprise Portals (EP) on the intranet, are increasingly popular in companies of all sizes. Enterprise Portals allow for structuring information in a way that resembles the organization of paper copies, i.e. simulating folders and registries and furthermore, provide simple routines for publishing and collaborating. Hence, in general, such kind of information management is not much different from paper management: electronic documents must be uploaded into the Enterprise Portal manually, filed into folders (which have to be created manually, too), tagged and related to other information objects if need be. With this approach information structuring remains subject to the individual user leading to the well-known problems of multiple filing, overlooking relevant information and incomprehensible Folder structure. The SEEK!sem project aims at improving such kind of information system by automatically identifying and recommending related information resources to be added to a folder. The recommendations are based on rules, exploiting content and context similarity of information resources. Rules can be created upfront, based on explicitly defined Relations between information objects. They can also be machine learned, i.e. the recommender exploits the existing linkage between documents, folders and other objects to learn “relatedness rules”. In either case, potential new connections are inferred by applying the rules in a reasoning step. Recommended new connections are ranked by the sum of the scores of all applied rules – the rule scores, again, can either be provided by experts or machinelearned. The applied rules can serve as an explanation of a recommendation, i.e. they can assist users in understanding why a particular connection is suggested.
Where Did I(t) put it? A holistic solution to the automatic construction of topic trees for navigation
Managing information based on hierarchical structures is prevailing, be it by storing documents physically in a file structure like MS explorer or virtually in topic trees as in many web applications. The problem is that the structure evolves over time, created individually and hence reflecting individual opinions of how information objects should be grouped. This leads to time consuming searches and error prone retrieval results since relevant documents might be stored elsewhere. Our approach aims at solving the problem by replacing or complementing the manually created navigation structures by automatically created ones. We consider existing approaches for clustering and labelling and focus on yet unrewarding aspects like having information objects in inner nodes (as it is common in folder hierarchies) and cognitively adequate labelling for textual and non-textual resources. Evaluation was done by knowledge experts based on a comparison of retrieval time for finding given documents in manually and automatic generated information structures and showed the advantage of automatically created topic trees.