Research Publications

2017

Rens G, Moodley D. A hybrid POMDP-BDI agent architecture with online stochastic planning and plan caching. Cognitive Systems Research. 2017;43. doi:http://dx.doi.org/10.1016/j.cogsys.2016.12.002.

This article presents an agent architecture for controlling an autonomous agent in stochastic, noisy environments. The architecture combines the partially observable Markov decision process (POMDP) model with the belief-desire-intention (BDI) framework. The Hybrid POMDP-BDI agent architecture takes the best features from the two approaches, that is, the online generation of reward-maximizing courses of action from POMDP theory, and sophisticated multiple goal management from BDI theory. We introduce the advances made since the introduction of the basic architecture, including (i) the ability to pursue and manage multiple goals simultaneously and (ii) a plan library for storing pre-written plans and for storing recently generated plans for future reuse. A version of the architecture is implemented and is evaluated in a simulated environment. The results of the experiments show that the improved hybrid architecture outperforms the standard POMDP architecture and the previous basic hybrid architecture for both processing speed and effectiveness of the agent in reaching its goals.

@article{147,
  author = {Gavin Rens and Deshen Moodley},
  title = {A hybrid POMDP-BDI agent architecture with online stochastic planning and plan caching},
  abstract = {This article presents an agent architecture for controlling an autonomous agent in stochastic, noisy environments. The architecture combines the partially observable Markov decision process (POMDP) model with the belief-desire-intention (BDI) framework. The Hybrid POMDP-BDI agent architecture takes the best features from the two approaches, that is, the online generation of reward-maximizing courses of action from POMDP theory, and sophisticated multiple goal management from BDI theory. We introduce the advances made since the introduction of the basic architecture, including (i) the ability to pursue and manage multiple goals simultaneously and (ii) a plan library for storing pre-written plans and for storing recently generated plans for future reuse. A version of the architecture is implemented and is evaluated in a simulated environment. The results of the experiments show that the improved hybrid architecture outperforms the standard POMDP architecture and the previous basic hybrid architecture for both processing speed and effectiveness of the agent in reaching its goals.},
  year = {2017},
  journal = {Cognitive Systems Research},
  volume = {43},
  pages = {1-20},
  publisher = {Elsevier B.V.},
  isbn = {1389-0417},
  doi = {http://dx.doi.org/10.1016/j.cogsys.2016.12.002},
}
Davel MH, Giwa O. The Effect of Language Identification Accuracy on Speech Recognition Accuracy of Proper Names. Proc. Annual Symp. Pattern Recognition Association of South Africa and Mechatronics International Conference. 2017. https://repository.nwu.ac.za/bitstream/handle/10394/26440/2017Effect.pdf?sequence=3&isAllowed=y.

This work focused on the implications of producing language-based pronunciation variants for proper name recognition, where different LID techniques are used to predict the most probable source language(s) of a word. Both G2P and ASR performance were analysed and compared. To understand the implications of creating LID-based dictionaries, we considered four dictionaries that were generated using a combination of LID and G2P prediction. The same language-specific G2P predictors were used for all dictionaries but different LID options were evaluated: when the true source language is known, when a single source language is predicted, when multiple source languages are predicted, and when it is simply assumed that all words may be from all relevant source languages. These dictionaries were evaluated against a reference dictionary (developed with each of the corpora and manually corrected/verified) to measure G2P accuracy. ASR performance was evaluated by developing a full-blown ASR system and evaluating performance with both a flat and trained LM. The extent to which LID accuracy influences ASR performance is most visible from the results in Table V. The effect of improved LID tags can be observed, with the Single approach performing the best of the predicted tags. While the G2P error accounts for the WER difference between Ref-LID and Manual results, the difference between the Single and the Ref-LID results provides a measure of the effect of remaining LID prediction error During G2P analysis, it became clear that the way in which variants are dealt with during accuracy calculation has a large effect on measured performance. Using existing G2P accuracy measures, variants in the hypothesised dictionary are not penalised sufficiently, an issue we aim to address in future work. Abstract—Utilising the known language of origin of a name can be useful when predicting the pronunciation of the name. When this language is not known, automatic language identification (LID) can be used to influence which language-specific grapheme-to-phoneme (G2P) predictor is triggered to produce a pronunciation for the name. We investigate the implications when both the LID system and the G2P system generate errors: what influence does this have on a resulting speech recognition system? We experiment with different approaches to LID-based dictionary creation and report on results in four South African languages: Afrikaans, English, Sesotho and isiZulu.

@proceedings{172,
  author = {Marelie Davel and Oluwapelumi Giwa},
  title = {The Effect of Language Identification Accuracy on Speech Recognition Accuracy of Proper Names},
  abstract = {This work focused on the implications of producing language-based pronunciation variants for proper name recognition, where different LID techniques are used to predict the most probable source language(s) of a word. Both G2P and ASR performance were analysed and compared. To understand the implications of creating LID-based dictionaries, we considered four dictionaries that were generated using a combination of LID and G2P prediction. The same language-specific G2P predictors were used for all dictionaries but different LID options were evaluated: when the true source language is known, when a single source language is predicted, when multiple source languages are predicted, and
when it is simply assumed that all words may be from all relevant source languages. These dictionaries were evaluated against a reference dictionary (developed with each of the corpora and manually corrected/verified) to measure G2P accuracy. ASR performance was evaluated by developing a full-blown ASR system and evaluating performance with both a flat and trained LM.
The extent to which LID accuracy influences ASR performance is most visible from the results in Table V. The effect of improved LID tags can be observed, with the Single approach performing the best of the predicted tags. While the G2P error accounts for the WER difference between Ref-LID and Manual results, the difference between the Single and the Ref-LID results provides a measure of the effect of remaining LID prediction error During G2P analysis, it became clear that the way in which variants are dealt with during accuracy calculation has a large effect on measured performance. Using existing G2P accuracy measures, variants in the hypothesised dictionary are not penalised sufficiently, an issue we aim to address in future work.
Abstract—Utilising the known language of origin of a name can be useful when predicting the pronunciation of the name. When this language is not known, automatic language identification (LID) can be used to influence which language-specific grapheme-to-phoneme (G2P) predictor is triggered to produce a pronunciation for the name. We investigate the implications when both the LID system and the G2P system generate errors: what influence does this have on a resulting speech recognition system? We experiment with different approaches to LID-based dictionary creation and report on results in four South African languages: Afrikaans, English, Sesotho and isiZulu.},
  year = {2017},
  journal = {Proc. Annual Symp. Pattern Recognition Association of South Africa and Mechatronics International Conference},
  pages = {187-192},
  month = {29/11-01/12},
  url = {https://repository.nwu.ac.za/bitstream/handle/10394/26440/2017Effect.pdf?sequence=3&isAllowed=y},
}
Davel MH, Giwa O. Bilateral G2P Accuracy: Measuring the effect of variants. Proc. Annual Symp. Pattern Recognition Association of South Africa and Mechatronics International Conference . 2017.

Many G2P techniques (such as JSMs) allow a varying number of variants to be generated. To optimize ASR results, systems must be trained and tested using the same dictionary, which is time-consuming and computationally expensive, while measuring G2P accuracy is computationally inexpensive. This paper focused on how variants are dealt with during accuracy calculation, which in turn has a significant effect on measured G2P performance. Based on the existing G2P accuracy measures, where variants in the hypothesised dictionary are not penalised sufficiently, we propose a straightforward technique to measure G2P accuracy when both the reference and hypothesised dictionaries contain variants. This technique automatically penalises a dictionary for over- or undergenerating variants. Using this new metric, if we order the dictionaries according to G2P performance (best to worst), we obtain an ordering that correlates with the actual ASR performance observed. Finally, this technique can be used to set variant thresholds for unseen words, based on accuracies observed on a small seen subset of the pronunciation dictionary. Abstract—Incorporating pronunciation variants in a dictionary is controversial, as this can be either advantageous or detrimental for a speech recognition system. Grapheme-tophoneme (G2P) accuracy can help guide this decision, but calculating the G2P accuracy of variant-based dictionaries is not fully straightforward. We propose a variant matching technique to measure G2P accuracy in a principled way, when both the reference and hypothesised dictionaries may include variants. We use the new measure to evaluate G2P accuracy and speech recognition performance of systems developed with an existing set of dictionaries, and observe a better correlation between G2P accuracy and speech recognition performance, than when utilising alternative metrics.

@proceedings{170,
  author = {Marelie Davel and O. Giwa},
  title = {Bilateral G2P Accuracy: Measuring the effect of variants},
  abstract = {Many G2P techniques (such as JSMs) allow a varying number of variants to be generated. To optimize ASR results, systems must be trained and tested using the same dictionary, which is time-consuming and computationally
expensive, while measuring G2P accuracy is computationally inexpensive. This paper focused on how variants are dealt with during accuracy calculation, which in turn has a significant effect on measured G2P performance. Based on the existing G2P accuracy
measures, where variants in the hypothesised dictionary are not penalised sufficiently, we propose a straightforward technique to measure G2P accuracy when both the reference and hypothesised dictionaries contain variants. This technique
automatically penalises a dictionary for over- or undergenerating variants. Using this new metric, if we order the dictionaries according to G2P performance (best to worst), we obtain an ordering that correlates with the actual ASR performance observed.
Finally, this technique can be used to set variant thresholds for unseen words, based on accuracies observed on a small seen subset of the pronunciation dictionary.
Abstract—Incorporating pronunciation variants in a dictionary is controversial, as this can be either advantageous or detrimental for a speech recognition system. Grapheme-tophoneme (G2P) accuracy can help guide this decision, but calculating the G2P accuracy of variant-based dictionaries is not fully straightforward. We propose a variant matching technique to measure G2P accuracy in a principled way, when both the reference and hypothesised dictionaries may include variants. We use the new measure to evaluate G2P accuracy
and speech recognition performance of systems developed with an existing set of dictionaries, and observe a better correlation between G2P accuracy and speech recognition performance, than when utilising alternative metrics.},
  year = {2017},
  journal = {Proc. Annual Symp. Pattern Recognition Association of South Africa and Mechatronics International Conference},
  pages = {2018-213},
  month = {29/11-01/12},
}
van Niekerk D. Evaluating acoustic modelling of lexical stress for Afrikaans speech synthesis. Proc. Annual Symp. Pattern Recognition Association of South Africa and Mechatronics International Conference. 2017.

In this work an evaluation of acoustic modelling of lexical stress was presented, with the following distinct contributions and outcomes:the potential to explicitly specify it as input in different contexts, the extent to which this can be done in practice with perceptually significant result needs to be investigated: • A more detailed analysis of the objective measurements may be done to better understand the results in Table V and some indications in Table II that having a secondary stress level may be useful. For these reasons the current annotation should be considered a work-in-progress. • Given a successful explicit lexical stress feature, it may now be possible to consider further work on prosodic modelling, in the first instance to improve the synthesis of compound words perhaps by generalising patterns from simplex words or by using more powerful modelling techniques (results in Table V may indicate that more work can be done in these sparse contexts), and secondly on the implementation of higher-level prosody (e.g. prosodic prominence) which may depend on the lexical stress pattern [9]. Lastly, the positive perceptual results obtained here on a small speech corpus suggests that more accurate descriptions of lexical pronunciation of other under-resourced languages, especially in South Africa [11], may also be worthwhile. Abstract—An explicit lexical stress feature is investigated for statistical parametric speech synthesis in Afrikaans: Firstly, objective measures are used to assess proposed annotation protocols and dictionaries compared to the baseline (implicit modelling) on the Lwazi 2 text-to-speech corpus. Secondly, the best candidates are evaluated on additional corpora. Finally, a comparative subjective evaluation is conducted to determine the perceptual impact on text-to-speech synthesis. The best candidate dictionary is associated with favourable objective results obtained on all corpora and was preferred in the subjective test. This suggests that it may form a basis for further refinement and work on improved prosodic models. Index Terms—pronunciation dictionary, under-resourced language, syllable-stress, lexical stress, Afrikaans, speech synthesis, text-to-speech.

@proceedings{171,
  author = {D. van Niekerk},
  title = {Evaluating acoustic modelling of lexical stress for Afrikaans speech synthesis},
  abstract = {In this work an evaluation of acoustic modelling of lexical stress was presented, with the following distinct contributions and outcomes:the potential to explicitly specify it as input in different contexts, the extent to which this can be done in
practice with perceptually significant result needs to be investigated:
• A more detailed analysis of the objective measurements may be done to better understand the results in Table V and some indications in Table II that having a secondary stress level may be useful. For these reasons the current
annotation should be considered a work-in-progress.
• Given a successful explicit lexical stress feature, it may now be possible to consider further work on prosodic modelling, in the first instance to improve the synthesis of compound words perhaps by generalising patterns from
simplex words or by using more powerful modelling techniques (results in Table V may indicate that more work can be done in these sparse contexts), and secondly on the implementation of higher-level prosody (e.g. prosodic prominence) which 
may depend on the lexical stress pattern [9]. 
Lastly, the positive perceptual results obtained here on a small speech corpus suggests that more accurate descriptions of lexical pronunciation of other under-resourced languages, especially in South Africa [11], may also be worthwhile.
Abstract—An explicit lexical stress feature is investigated for statistical parametric speech synthesis in Afrikaans: Firstly, objective measures are used to assess proposed annotation protocols
and dictionaries compared to the baseline (implicit modelling) on the Lwazi 2 text-to-speech corpus. Secondly, the best candidates are evaluated on additional corpora. Finally, a comparative
subjective evaluation is conducted to determine the perceptual impact on text-to-speech synthesis. The best candidate dictionary is associated with favourable objective results obtained on all
corpora and was preferred in the subjective test. This suggests that it may form a basis for further refinement and work on improved prosodic models.
Index Terms—pronunciation dictionary, under-resourced language, syllable-stress, lexical stress, Afrikaans, speech synthesis, text-to-speech.},
  year = {2017},
  journal = {Proc. Annual Symp. Pattern Recognition Association of South Africa and Mechatronics International Conference},
  month = {30/11-01/12},
}

2016

Ojeme B, Mbogho A, Meyer T. Probabilistic Expert Systems for Reasoning in Clinical Depressive Disorders. 15th IEEE International Conference on Machine Learning and Applications (ICMLA). 2016. doi:10.1109/ICMLA.2016.0105.

Like other real-world problems, reasoning in clinical depression presents cognitive challenges for clinicians. This is due to the presence of co-occuring diseases, incomplete data, uncertain knowledge, and the vast amount of data to be analysed. Current approaches rely heavily on the experience, knowledge, and subjective opinions of clinicians, creating scalability issues. Automating this process requires a good knowledge representation technique to capture the knowledge of the domain experts, and multidimensional inferential reasoning approaches that can utilise a few bits and pieces of information for efficient reasoning. This study presents knowledge-based system with variants of Bayesian network models for efficient inferential reasoning, translating from available fragmented depression data to the desired information in a visually interpretable and transparent manner. Mutual information, a Conditional independence test-based method was used to learn the classifiers.

@proceedings{361,
  author = {Blessing Ojeme and Audrey Mbogho and Thomas Meyer},
  title = {Probabilistic Expert Systems for Reasoning in Clinical Depressive Disorders},
  abstract = {Like other real-world problems, reasoning in clinical depression presents cognitive challenges for clinicians. This is due to the presence of co-occuring diseases, incomplete data, uncertain knowledge, and the vast amount of data to be analysed. Current approaches rely heavily on the experience, knowledge, and subjective opinions of clinicians, creating scalability issues. Automating this process requires a good knowledge representation technique to capture the knowledge of the domain experts, and multidimensional inferential reasoning approaches that can utilise a few bits and pieces of information for efficient reasoning. This study presents knowledge-based system with variants of Bayesian network models for efficient inferential reasoning, translating from available fragmented depression data to the desired information in a visually interpretable and transparent manner. Mutual information, a Conditional independence test-based method was used to learn the classifiers.},
  year = {2016},
  journal = {15th IEEE International Conference on Machine Learning and Applications (ICMLA)},
  month = {18/12 - 20/12},
  publisher = {IEEE},
  doi = {10.1109/ICMLA.2016.0105},
}
Casini G, Meyer T. Using Defeasible Information to Obtain Coherence. Fifteenth International Conference on Principles of Knowledge Representation and Reasoning (KR). 2016. doi:https://dl.acm.org/doi/10.5555/3032027.3032097.

We consider the problem of obtaining coherence in a propositional knowledge base using techniques from Belief Change. Our motivation comes from the field of formal ontologies where coherence is interpreted to mean that a concept name has to be satisfiable. In the propositional case we consider here, this translates to a propositional formula being satisfiable. We define belief change operators in a framework of nonmonotonic preferential reasoning. We show how the introduction of defeasible information using contraction operators can be an effective means for obtaining coherence.

@proceedings{360,
  author = {Giovanni Casini and Thomas Meyer},
  title = {Using Defeasible Information to Obtain Coherence},
  abstract = {We consider the problem of obtaining coherence in a propositional knowledge base using techniques from Belief Change. Our motivation comes from the field of formal ontologies where coherence is interpreted to mean that a concept name has to be satisfiable. In the propositional case we consider here, this translates to a propositional formula being satisfiable. We define belief change operators in a framework of nonmonotonic preferential reasoning. We show how the introduction of defeasible information using contraction operators can be an effective means for obtaining coherence.},
  year = {2016},
  journal = {Fifteenth International Conference on  Principles of Knowledge Representation and Reasoning (KR)},
  pages = {537-540},
  month = {25/04 - 29/04},
  publisher = {AAAI Press},
  doi = {https://dl.acm.org/doi/10.5555/3032027.3032097},
}
Rens G, Casini G, Meyer T. On Revision of Partially Specified Convex Probabilistic Belief Bases. European Conference on Artificial Intelligence (ECAI). 2016. https://www.researchgate.net/publication/307577667_On_Revision_of_Partially_Specified_Convex_Probabilistic_Belief_Bases.

We propose a method for an agent to revise its incomplete probabilistic beliefs when a new piece of propositional information is observed. In this work, an agent’s beliefs are represented by a set of probabilistic formulae – a belief base. The method involves determining a representative set of ‘boundary’ probability distributions consistent with the current belief base, revising each of these probability distributions and then translating the revised information into a new belief base. We use a version of Lewis Imaging as the revision operation. The correctness of the approach is proved. An analysis of the approach is done against six rationality postulates. The expressivity of the belief bases under consideration are rather restricted, but has some applications. We also discuss methods of belief base revision employing the notion of optimum entropy, and point out some of the benefits and difficulties in those methods. Both the boundary distribution method and the optimum entropy methods are reasonable, yet yield different results.

@proceedings{359,
  author = {Gavin Rens and Giovanni Casini and Thomas Meyer},
  title = {On Revision of Partially Specified Convex Probabilistic Belief Bases},
  abstract = {We propose a method for an agent to revise its incomplete probabilistic beliefs when a new piece of propositional information is observed. In this work, an agent’s beliefs are represented by a set of probabilistic formulae – a belief base. The method involves determining a representative set of ‘boundary’ probability distributions consistent with the current belief base, revising each of these probability distributions and then translating the revised information into a new belief base. We use a version of Lewis Imaging as the revision operation. The correctness of the approach is proved. An analysis of the approach is done against six rationality postulates. The expressivity of the belief bases under consideration are rather restricted, but has some applications. We also discuss methods of belief base revision employing the notion of optimum entropy, and point out some of the benefits and difficulties in those methods. Both the boundary distribution method and the optimum entropy methods are reasonable, yet yield different results.},
  year = {2016},
  journal = {European Conference on Artificial Intelligence (ECAI)},
  pages = {921-929},
  month = {29/08 - 2/09},
  publisher = {IO Press},
  url = {https://www.researchgate.net/publication/307577667_On_Revision_of_Partially_Specified_Convex_Probabilistic_Belief_Bases},
}
Van Niekerk DR. Syllabification for Afrikaans speech synthesis. Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference (PRASA-RobMech). 2016. doi:10.1109/RoboMech.2016.7813143.

This paper describes the continuing development of a pronunciation resource for speech synthesis of Afrikaans by augmenting an existing pronunciation dictionary to include syllable boundaries and stress. Furthermore, different approaches for grapheme to phoneme conversion and syllabification derived from the dictionary are evaluated. Cross-validation experiments suggest that joint sequence models are effective at directly modelling pronunciations including syllable boundaries. Finally, some informal observations and demonstrations are presented regarding the integration of this work into a typical text-to-speech system.

@proceedings{285,
  author = {Daniel Van Niekerk},
  title = {Syllabification for Afrikaans speech synthesis},
  abstract = {This paper describes the continuing development of a pronunciation resource for speech synthesis of Afrikaans by augmenting an existing pronunciation dictionary to include syllable boundaries and stress. Furthermore, different approaches for grapheme to phoneme conversion and syllabification derived from the dictionary are evaluated. Cross-validation experiments suggest that joint sequence models are effective at directly modelling pronunciations including syllable boundaries. Finally, some informal observations and demonstrations are presented regarding the integration of this work into a typical text-to-speech system.},
  year = {2016},
  journal = {Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference (PRASA-RobMech)},
  pages = {31-36},
  address = {Stellenbosch, South Africa},
  isbn = {978-1-5090-3335-5},
  doi = {10.1109/RoboMech.2016.7813143},
}
Kleynhans N, Hartman W, Van Niekerk DR, et al. Code-switched English Pronunciation Modeling for Swahili Spoken Term Detection. Procedia Computer Science. 2016;81. doi:10.1016/j.procs.2016.04.040.

We investigate modeling strategies for English code-switched words as found in a Swahili spoken term detection system. Code switching, where speakers switch language in a conversation, occurs frequently in multilingual environments, and typically deteriorates STD performance. Analysis is performed in the context of the IARPA Babel program which focuses on rapid STD system development for under-resourced languages. Our results show that approaches that specifically target the modeling of code-switched words, significantly improve the detection performance of these words.

@article{271,
  author = {Neil Kleynhans and William Hartman and Daniel Van Niekerk and Charl Van Heerden and Richard Schwartz and Stavros Tsakalidis and Marelie Davel},
  title = {Code-switched English Pronunciation Modeling for Swahili Spoken Term Detection},
  abstract = {We investigate modeling strategies for English code-switched words as found in a Swahili spoken term detection system. Code
switching, where speakers switch language in a conversation, occurs frequently in multilingual environments, and typically deteriorates STD performance. Analysis is performed in the context of the IARPA Babel program which focuses on rapid STD
system development for under-resourced languages. Our results show that approaches that specifically target the modeling of
code-switched words, significantly improve the detection performance of these words.},
  year = {2016},
  journal = {Procedia Computer Science},
  volume = {81},
  pages = {128-135},
  publisher = {Elsevier B.V.},
  address = {Yogyakarta, Indonesia},
  isbn = {1877-0509},
  doi = {10.1016/j.procs.2016.04.040},
}
Leenen L, Meyer T. Semantic Technologies and Big Data: Analytics for Cyber Defence. International Journal of Cyber Warfare and Terrorism. 2016;6(3).

The Governments, military forces and other organisations responsible for cybersecurity deal with vast amounts of data that has to be understood in order to lead to intelligent decision making. Due to the vast amounts of information pertinent to cybersecurity, automation is required for processing and decision making, specifically to present advance warning of possible threats. The ability to detect patterns in vast data sets, and being able to understanding the significance of detected patterns are essential in the cyber defence domain. Big data technologies supported by semantic technologies can improve cybersecurity, and thus cyber defence by providing support for the processing and understanding of the huge amounts of information in the cyber environment. The term big data analytics refers to advanced analytic techniques such as machine learning, predictive analysis, and other intelligent processing techniques applied to large data sets that contain different data types. The purpose is to detect patterns, correlations, trends and other useful information. Semantic technologies is a knowledge representation paradigm where the meaning of data is encoded separately from the data itself. The use of semantic technologies such as logic-based systems to support decision making is becoming increasingly popular. However, most automated systems are currently based on syntactic rules. These rules are generally not sophisticated enough to deal with the complexity of decisions required to be made. The incorporation of semantic information allows for increased understanding and sophistication in cyber defence systems. This paper argues that both big data analytics and semantic technologies are necessary to provide counter measures against cyber threats. An overview of the use of semantic technologies and big data technologies in cyber defence is provided, and important areas for future research in the combined domains are discussed.

@article{229,
  author = {Louise Leenen and Thomas Meyer},
  title = {Semantic Technologies and Big Data: Analytics for Cyber Defence},
  abstract = {The Governments, military forces and other organisations responsible for cybersecurity deal with
vast amounts of data that has to be understood in order to lead to intelligent decision making. Due
to the vast amounts of information pertinent to cybersecurity, automation is required for processing
and decision making, specifically to present advance warning of possible threats. The ability to detect
patterns in vast data sets, and being able to understanding the significance of detected patterns are
essential in the cyber defence domain. Big data technologies supported by semantic technologies
can improve cybersecurity, and thus cyber defence by providing support for the processing and
understanding of the huge amounts of information in the cyber environment. The term big data
analytics refers to advanced analytic techniques such as machine learning, predictive analysis, and
other intelligent processing techniques applied to large data sets that contain different data types. The
purpose is to detect patterns, correlations, trends and other useful information. Semantic technologies
is a knowledge representation paradigm where the meaning of data is encoded separately from the
data itself. The use of semantic technologies such as logic-based systems to support decision making
is becoming increasingly popular. However, most automated systems are currently based on syntactic
rules. These rules are generally not sophisticated enough to deal with the complexity of decisions
required to be made. The incorporation of semantic information allows for increased understanding and
sophistication in cyber defence systems. This paper argues that both big data analytics and semantic
technologies are necessary to provide counter measures against cyber threats. An overview of the
use of semantic technologies and big data technologies in cyber defence is provided, and important
areas for future research in the combined domains are discussed.},
  year = {2016},
  journal = {International Journal of Cyber Warfare and Terrorism},
  volume = {6},
  issue = {3},
}
van Niekerk L, Watson B. The Development and Evaluation of an Electronic Serious Game Aimed at the Education of Core Programming Skills. 2016;MA. http://hdl.handle.net/10019.1/100119.

No Abstract

@phdthesis{207,
  author = {L. van Niekerk and Bruce Watson},
  title = {The Development and Evaluation of an Electronic Serious Game Aimed at the Education of Core Programming Skills},
  abstract = {No Abstract},
  year = {2016},
  volume = {MA},
  url = {http://hdl.handle.net/10019.1/100119},
}
Kala JR, Viriri S, Moodley D. Leaf Classification Using Convexity Moments of Polygons. International Symposium on Visual Computing. 2016.

Research has shown that shape features can be used in the process of object recognition with promising results. However, due to a wide variety of shape descriptors, selecting the right one remains a difficult task. This paper presents a new shape recognition feature: Convexity Moment of Polygons. The Convexity Moments of Polygons is derived from the Convexity measure of polygons. A series of experimentations based on FLAVIA images dataset was performed to demonstrate the accuracy of the proposed feature compared to the Convexity measure of polygons in the field of leaf classification. A classification rate of 92% was obtained with the Convexity Moment of Polygons, 80% with the convexity Measure of Polygons using the Radial Basis function neural networks classifier (RBF).

@proceedings{161,
  author = {J.R. Kala and S. Viriri and Deshen Moodley},
  title = {Leaf Classification Using Convexity Moments of Polygons},
  abstract = {Research has shown that shape features can be used in the process of object recognition with promising results. However, due to a wide variety of shape descriptors, selecting the right one remains a difficult task. This paper presents a new shape recognition feature: Convexity Moment of Polygons. The Convexity Moments of Polygons is derived from the Convexity measure of polygons. A series of experimentations based on FLAVIA images dataset was performed to demonstrate the accuracy of the proposed feature compared to the Convexity measure of polygons in the field of leaf classification. A classification rate of 92% was obtained with the Convexity Moment of Polygons, 80% with the convexity Measure of Polygons using the Radial Basis function neural networks classifier (RBF).},
  year = {2016},
  journal = {International Symposium on Visual Computing},
  pages = {300-339},
  month = {14/12-16/12},
  isbn = {978-3-319-50832-0},
}
Coetzer W, Moodley D, Gerber A. Eliciting and Representing High-Level Knowledge Requirements to Discover Ecological Knowledge in Flower-Visiting Data. PLoS ONE . 2016;11(11). http://pubs.cs.uct.ac.za/archive/00001127/01/journal.pone.0166559.pdf.

Observations of individual organisms (data) can be combined with expert ecological knowledge of species, especially causal knowledge, to model and extract from flower–visiting data useful information about behavioral interactions between insect and plant organisms, such as nectar foraging and pollen transfer. We describe and evaluate a method to elicit and represent such expert causal knowledge of behavioral ecology, and discuss the potential for wider application of this method to the design of knowledge-based systems for knowledge discovery in biodiversity and ecosystem informatics.

@article{159,
  author = {Willem Coetzer and Deshen Moodley and Aurona Gerber},
  title = {Eliciting and Representing High-Level Knowledge Requirements to Discover Ecological Knowledge in Flower-Visiting Data},
  abstract = {Observations of individual organisms (data) can be combined with expert ecological knowledge of species, especially causal knowledge, to model and extract from flower–visiting data useful information about behavioral interactions between insect and plant organisms, such as nectar foraging and pollen transfer. We describe and evaluate a method to elicit and represent such expert causal knowledge of behavioral ecology, and discuss the potential for wider application of this method to the design of knowledge-based systems for knowledge discovery in biodiversity and ecosystem informatics.},
  year = {2016},
  journal = {PLoS ONE},
  volume = {11},
  pages = {1-15},
  issue = {11},
  url = {http://pubs.cs.uct.ac.za/archive/00001127/01/journal.pone.0166559.pdf},
}
Waltham M, Moodley D. An Analysis of Artificial Intelligence Techniques in Multiplayer Online Battle Arena Game Environments. Annual Conference of the South African Institute of Computer Scientists and Information Technologists (SAICSIT 2016). 2016. doi: http://dx.doi.org/10.1145/2987491.2987513.

The 3D computer gaming industry is constantly exploring new avenues for creating immersive and engaging environments. One avenue being explored is autonomous control of the behaviour of non-player characters (NPC). This paper reviews and compares existing artificial intelligence (AI) techniques for controlling the behaviour of non-human characters in Multiplayer Online Battle Arena (MOBA) game environments. Two techniques, the fuzzy state machine (FuSM) and the emotional behaviour tree (EBT), were reviewed and compared. In addition, an alternate and simple mechanism to incorporate emotion in a behaviour tree is proposed and tested. Initial tests of the mechanism show that it is a viable and promising mechanism for effectively tracking the emotional state of an NPC and for incorporating emotion in NPC decision making.

@proceedings{157,
  author = {Michael Waltham and Deshen Moodley},
  title = {An Analysis of Artificial Intelligence Techniques in Multiplayer Online Battle Arena Game Environments},
  abstract = {The 3D computer gaming industry is constantly exploring new avenues for creating immersive and engaging environments. One avenue being explored is autonomous control of the behaviour of non-player characters (NPC). This paper reviews and compares existing artificial intelligence (AI) techniques for controlling the behaviour of non-human characters in Multiplayer Online Battle Arena (MOBA) game environments. Two techniques, the fuzzy state machine (FuSM) and the emotional behaviour tree (EBT), were reviewed and compared. In addition, an alternate and simple mechanism to incorporate emotion in a behaviour tree is proposed and tested. Initial tests of the mechanism show that it is a viable and promising mechanism for effectively tracking the emotional state of an NPC and for incorporating emotion in NPC decision making.},
  year = {2016},
  journal = {Annual Conference of the South African Institute of Computer Scientists and Information Technologists (SAICSIT 2016)},
  pages = {45},
  month = {26/09-28/09},
  publisher = {ACM},
  address = {Johannesburg},
  isbn = {978-1-4503-4805-8},
  doi = {http://dx.doi.org/10.1145/2987491.2987513},
}
Clark A, Moodley D. A System for a Hand Gesture-Manipulated Virtual Reality Environment. Annual Conference of the South African Institute of Computer Scientists and Information Technologists (SAICSIT 2016). 2016. doi:http://dx.doi.org/10.1145/2987491.2987511.

Extensive research has been done using machine learning techniques for hand gesture recognition (HGR) using camera-based devices; such as the Leap Motion Controller (LMC). However, limited research has investigated machine learning techniques for HGR in virtual reality applications (VR). This paper reports on the design, implementation, and evaluation of a static HGR system for VR applications using the LMC. The gesture recognition system incorporated a lightweight feature vector of five normalized tip-to-palm distances and a k-nearest neighbour (kNN) classifier. The system was evaluated in terms of response time, accuracy and usability using a case-study VR stellar data visualization application created in the Unreal Engine 4. An average gesture classification time of 0.057ms with an accuracy of 82.5% was achieved on four distinct gestures, which is comparable with previous results from Sign Language recognition systems. This shows the potential of HGR machine learning techniques applied to VR, which were previously applied to non-VR scenarios such as Sign Language recognition.

@proceedings{156,
  author = {A. Clark and Deshen Moodley},
  title = {A System for a Hand Gesture-Manipulated Virtual Reality Environment},
  abstract = {Extensive research has been done using machine learning techniques for hand gesture recognition (HGR) using camera-based devices; such as the Leap Motion Controller (LMC). However, limited research has investigated machine learning techniques for HGR in virtual reality applications (VR). This paper reports on the design, implementation, and evaluation of a static HGR system for VR applications using the LMC. The gesture recognition system incorporated a lightweight feature vector of five normalized tip-to-palm distances and a k-nearest neighbour (kNN) classifier. The system was evaluated in terms of response time, accuracy and usability using a case-study VR stellar data visualization application created in the Unreal Engine 4. An average gesture classification time of 0.057ms with an accuracy of 82.5% was achieved on four distinct gestures, which is comparable with previous results from Sign Language recognition systems. This shows the potential of HGR machine learning techniques applied to VR, which were previously applied to non-VR scenarios such as Sign Language recognition.},
  year = {2016},
  journal = {Annual Conference of the South African Institute of Computer Scientists and Information Technologists (SAICSIT 2016)},
  pages = {10},
  month = {26/09-28/09},
  publisher = {ACM},
  address = {Johannesburg},
  isbn = {978-1-4503-4805-8},
  doi = {http://dx.doi.org/10.1145/2987491.2987511},
}
Van Heerden CJ, Kleynhans N, Davel MH. Improving the Lwazi ASR baseline. Interspeech. 2016. doi:http://dx.doi.org/10.21437/Interspeech.2016-1412.

We investigate the impact of recent advances in speech recognition techniques for under-resourced languages. Specifically, we review earlier results published on the Lwazi ASR corpus of South African languages, and experiment with additional acoustic modeling approaches. We demonstrate large gains by applying current state-of-the-art techniques, even if the data itself is neither extended nor improved. We analyze the various performance improvements observed, report on comparative performance per technique – across all eleven languages in the corpus – and discuss the implications of our findings for under-resourced languages in general.

@proceedings{153,
  author = {Charl Van Heerden and Neil Kleynhans and Marelie Davel},
  title = {Improving the Lwazi ASR baseline},
  abstract = {We investigate the impact of recent advances in speech recognition techniques for under-resourced languages. Specifically, we review earlier results published on the Lwazi ASR corpus of South African languages, and experiment with additional acoustic modeling approaches. We demonstrate large gains by applying current state-of-the-art techniques, even if the data itself is neither extended nor improved. We analyze the various performance improvements observed, report on comparative performance per technique – across all eleven languages in the corpus – and discuss the implications of our findings for under-resourced languages in general.},
  year = {2016},
  journal = {Interspeech},
  pages = {3529-3538},
  month = {08/09-12/09},
  address = {San Francisco, USA},
  doi = {http://dx.doi.org/10.21437/Interspeech.2016-1412},
}
Lapalme J, Gerber A, van der Merwe A, Zachman J, de Vries M, Hinkelmann K. Exploring the future of enterprise architecture: A Zachman perspective. 2016. http://www.sciencedirect.com/science/journal/01663615/79.

No Abstract

@misc{150,
  author = {James Lapalme and Aurona Gerber and Alta van der Merwe and John Zachman and Marne de Vries and Knut Hinkelmann},
  title = {Exploring the future of enterprise architecture: A Zachman perspective.},
  abstract = {No Abstract},
  year = {2016},
  url = {http://www.sciencedirect.com/science/journal/01663615/79},
}
Harmse H, Britz K, Gerber A. Armstrong Relations for Ontology Design and Evaluation. 2016. http://ceur-ws.org/Vol-1577/.

No Abstract

@misc{149,
  author = {Henriette Harmse and Katarina Britz and Aurona Gerber},
  title = {Armstrong Relations for Ontology Design and Evaluation},
  abstract = {No Abstract},
  year = {2016},
  url = {http://ceur-ws.org/Vol-1577/},
}
  • CSIR
  • DSI
  • Covid-19