Browsing by Author "Keet, Maria"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    Open Access
    Data to information to text summaries of financial data
    (2018) Kyle, Cameron; Keet, Maria
    The field of auditing is becoming increasingly dependent on information technology as auditors are forced to follow the increasingly complex information processing of their clients. There exists a need for a system that can convert vast quantities of data generated by existing systems and data analytics techniques, into usable information and then into a format that is easy for someone not trained in data analytics to understand. This is possible through Natural Language Generation (NLG). The field of auditing has not previously been applied to this pipeline. This research looks at the auditing of Investment Fund Management, of which a specific procedure is the comparison of two time series (one of the fund being tested and another of the benchmark it is supposed to follow) to identify potential misstatements in the investment fund. We solve this problem through a combination of incremental innovations on existing techniques in the text planning stage as well as pre-NLG processing steps, with effective leveraging of accepted sentence planning and realisation techniques. Additionally, fuzzy logic is used to provide a more human decision system. This allows the system to transform data into information and then into text. This has been evaluated by experts and achieved positive results with regard to audit impact, readability and understandability, while falling slight short of the stated accuracy targets. These preliminary results are positive in general and are therefore encouraging for further development.
  • Thumbnail Image
    Item
    Open Access
    Design and optimisation of a low cost Cognitive Mesh Network
    (2016) Nleya, Sindiso M; Keet, Maria; Bagula, Antoine
    Wireless Mesh Networks (WMNs) have been touted as the most promising wireless technology in providing high-bandwidth Internet access to rural, remote and under-served areas, with relatively lower investment cost as compared to traditional access networks. WMNs structurally comprise of mesh routers and mesh clients. Furthermore, WMNs have an envisaged ability to provide a heterogeneous network system that integrates wireless technologies such as IEEE 802.22 WRAN, IEEE 802.16 WiMAX, IEEE 802.11 Wi-Fi, Blue-tooth etc. The recent proliferation of new devices on the market such as smart phones and, tablets, and the growing number of resource hungry applications has placed a serious strain on spectrum availability which gives rise to the spectrum scarcity problem. The spectrum scarcity problem essentially results in increased spectrum prices that hamper the growth and efficient performance of WMNs as well as subsequent transformation of WMN into the envisaged next generation networks. Recent developments in TV white space communications technology and the emergence of Cognitive radio devices that facilitate Dynamic Spectrum Access (DSA) have provided an opportunity to mitigate the spectrum scarcity problem. To solve the scarcity problem, this thesis reconsiders the classical Network Engineering (NE) and Traffic Engineering (TE) problems to objectively design a low cost Cognitive Mesh network that promotes efficient resources utilization and thereby achieve better Quality of Service (QoS) levels.
  • Thumbnail Image
    Item
    Open Access
    A foundation for ontology modularisation
    (2017) Dawood, Zubeida C; Keet, Maria
    There has been great interest in realising the Semantic Web. Ontologies are used to define Semantic Web applications. Ontologies have grown to be large and complex to the point where it causes cognitive overload for humans, in understanding and maintaining, and for machines, in processing and reasoning. Furthermore, building ontologies from scratch is time-consuming and not always necessary. Prospective ontology developers could consider using existing ontologies that are of good quality. However, an entire large ontology is not always required for a particular application, but a subset of the knowledge may be relevant. Modularity deals with simplifying an ontology for a particular context or by structure into smaller ontologies, thereby preserving the contextual knowledge. There are a number of benefits in modularising an ontology including simplified maintenance and machine processing, as well as collaborative efforts whereby work can be shared among experts. Modularity has been successfully applied to a number of different ontologies to improve usability and assist with complexity. However, problems exist for modularity that have not been satisfactorily addressed. Currently, modularity tools generate large modules that do not exclusively represent the context. Partitioning tools, which ought to generate disjoint modules, sometimes create overlapping modules. These problems arise from a number of issues: different module types have not been clearly characterised, it is unclear what the properties of a 'good' module are, and it is unclear which evaluation criteria applies to specific module types. In order to successfully solve the problem, a number of theoretical aspects have to be investigated. It is important to determine which ontology module types are the most widely-used and to characterise each such type by distinguishing properties. One must identify properties that a 'good' or 'usable' module meets. In this thesis, we investigate these problems with modularity systematically. We begin by identifying dimensions for modularity to define its foundation: use-case, technique, type, property, and evaluation metric. Each dimension is populated with sub-dimensions as fine-grained values. The dimensions are used to create an empirically-based framework for modularity by classifying a set of ontologies with them, which results in dependencies among the dimensions. The formal framework can be used to guide the user in modularising an ontology and as a starting point in the modularisation process. To solve the problem with module quality, new and existing metrics were implemented into a novel tool TOMM, and an experimental evaluation with a set of modules was performed resulting in dependencies between the metrics and module types. These dependencies can be used to determine whether a module is of good quality. For the issue with existing modularity techniques, we created five new algorithms to improve the current tools and techniques and experimentally evaluate them. The algorithms of the tool, NOMSA, performs as well as other tools for most performance criteria. For NOMSA's generated modules, two of its algorithms' generated modules are good quality when compared to the expected dependencies of the framework. The remaining three algorithms' modules correspond to some of the expected values for the metrics for the ontology set in question. The success of solving the problems with modularity resulted in a formal foundation for modularity which comprises: an exhaustive set of modularity dimensions with dependencies between them, a framework for guiding the modularisation process and annotating module, a way to measure the quality of modules using the novel TOMM tool which has new and existing evaluation metrics, the SUGOI tool for module management that has been investigated for module interchangeability, and an implementation of new algorithms to fill in the gaps of insufficient tools and techniques.
  • Thumbnail Image
    Item
    Open Access
    Natural language interface to relational database: a simplified customization approach
    (2016) Mvumbi, Tresor; Keet, Maria; Bagula, Antoine
    Natural language interfaces to databases (NLIDB) allow end-users with no knowledge of a formal language like SQL to query databases. One of the main open problems currently investigated is the development of NLIDB systems that are easily portable across several domains. The present study focuses on the development and evaluation of methods allowing to simplify customization of NLIDB targeting relational databases without sacrificing coverage and accuracy. This goal is approached by the introduction of two authoring frameworks that aim to reduce the workload required to port a NLIDB to a new domain. The first authoring approach is called top-down; it assumes the existence of a corpus of unannotated natural language sample questions used to pre-harvest key lexical terms to simplify customization. The top-down approach further reduces the configuration workload by autoincluding the semantics for negative form of verbs, comparative and superlative forms of adjectives in the configuration model. The second authoring approach introduced is bottom-up; it explores the possibility of building a configuration model with no manual customization using the information from the database schema and an off-the-shelf dictionary. The evaluation of the prototype system with geo-query, a benchmark query corpus, has shown that the top-down approach significantly reduces the customization workload: 93% of the entries defining the meaning of verbs and adjectives which represents the hard work has been automatically generated by the system; only 26 straightforward mappings and 3 manual definitions of meaning were required for customization. The top-down approach answered correctly 74.5 % of the questions. The bottom-up approach, however, has correctly answered only 1/3 of the questions due to insufficient lexicon and missing semantics. The use of an external lexicon did not improve the system's accuracy. The bottom-up model has nevertheless correctly answered 3/4 of the 105 simple retrieval questions in the query corpus not requiring nesting. Therefore, the bottom-up approach can be useful to build an initial lightweight configuration model that can be incrementally refined by using the failed queries to train a topdown model for example. The experimental results for top-down suggest that it is indeed possible to construct a portable NLIDB that reduces the configuration effort while maintaining a decent coverage and accuracy.