Search

The vast nature of scientific publications brings out the importance of Literature-Based Discovery (LBD) research that is highly beneficial to accelerate knowledge acquisition and the research development process. LBD is a knowledge discovery workflow that automatically detects significant, implicit knowledge associations hidden in fragmented knowledge areas by analysing existing scientific literature. Therefore, the LBD output not only assists in formulating scientifically sensible, novel research hypotheses but also encourages the development of cross-disciplinary research. In this systematic review, we provide an in-depth analysis of the computational techniques used in the LBD process using a novel, up-to-date, and detailed classification. Moreover, we also summarise the key milestones of the discipline through a timeline of topics. To provide a general overview of the discipline, the review outlines LBD validation checks, major LBD tools, application areas, domains, and generalisability of LBD methodologies. We also outline the insights gathered through our statistical analysis that capture the trends in LBD literature. To conclude, we discuss the prevailing research deficiencies in the discipline by highlighting the challenges and opportunities of future LBD research.

Good software documentation encourages good software engineering, but the meaning of "good" documentation is vaguely defined in the software engineering literature. To clarify this ambiguity, we draw on work from the data and information quality community to propose a framework that decomposes documentation quality into ten dimensions of structure, content, and style. To demonstrate its application, we recruited technical editors to apply the framework when evaluating examples from several genres of software documentation. We summarise their assessments -- for example, reference documentation and README files excel in quality whereas blog articles have more problems -- and we describe our vision for reasoning about software documentation quality and for the expansion and potential of a unified quality framework., to appear in the Visions and Reflections Track of the ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering 2020

Cyberbullying is a prevalent social problem that inflicts detrimental consequences to the health and safety of victims such as psychological distress, anti-social behaviour, and suicide. The automation of cyberbullying detection is a recent but widely researched problem, with current research having a strong focus on a binary classification of bullying versus non-bullying. This paper proposes a novel approach to enhancing cyberbullying detection through role modeling. We utilise a dataset from ASKfm to perform multi-class classification to detect participant roles (e.g. victim, harasser). Our preliminary results demonstrate promising performance including 0.83 and 0.76 of F1-score for cyberbullying and role classification respectively, outperforming baselines.

This paper describes the systems our team (AdelaideCyC) has developed for SemEval Task 12 (OffensEval 2020) to detect offensive language in social media. The challenge focuses on three subtasks – offensive language identification (subtask A), offense type identification (subtask B), and offense target identification (subtask C). Our team has participated in all the three subtasks. We have developed machine learning and deep learning-based ensembles of models. We have achieved F1-scores of 0.906, 0.552, and 0.623 in subtask A, B, and C respectively. While our performance scores are promising for subtask A, the results demonstrate that subtask B and C still remain challenging to classify.

Lecture videos are amongst the most widely used instructional methods in present Massive Open Online Courses (MOOCs) and other digital educational platforms. As the main form of instruction, students’ engagement behaviour with videos directly impacts the students’ success or failure and accordingly, in-video dropouts positively correlate with dropout from MOOCs. The primary focus of previous video learning analytics studies is on analysing video engagement behaviour using explicit factors (i.e. views or annotations). Limited research studies focus on implicit video learning analytics (e.g. pause, seek, content type) and their impact on students’ success, with existing studies addressing video interactions and their relationship with visual transitions. We aim to explore the association between video interactions and non-visual (i.e. verbal or discourse) transition. This research focuses on text (e.g. cohesion and syntactic complexity) and spoken (e.g. speaking rate) discourse features of lecture videos. We conduct a fine-grained analysis of 3.4 million video interactions of two AdelaideX MOOCs – Programming (Code101x) and Cyber, Surveillance and Security (Cyber101x). According to our results, some discourse features (e.g. lexical diversity and causal connectives) demonstrate statistically significant correlation with video interactions. We present insights for educational video design implications based on discourse processing theories.

README files play an essential role in shaping a developer's first impression of a software repository and in documenting the software project that the repository hosts. Yet, we lack a systematic understanding of the content of a typical README file as well as tools that can process these files automatically. To close this gap, we conduct a qualitative study involving the manual annotation of 4,226 README file sections from 393 randomly sampled GitHub repositories and we design and evaluate a classifier and a set of features that can categorize these sections automatically. We find that information discussing the `What' and `How' of a repository is very common, while many README files lack information regarding the purpose and status of a repository. Our multi-label classifier which can predict eight different categories achieves an F1 score of 0.746. To evaluate the usefulness of the classification, we used the automatically determined classes to label sections in GitHub README files using badges and showed files with and without these badges to twenty software professionals. The majority of participants perceived the automated labeling of sections based on our classifier to ease information discovery. This work enables the owners of software repositories to improve the quality of their documentation and it has the potential to make it easier for the software development community to discover relevant information in GitHub README files.

Detecting interesting, cross-disciplinary knowledge associations hidden in scientific publications can greatly assist scientists to formulate and validate scientifically sensible novel research hypotheses. This will also introduce new areas of research that can be successfully linked with their research discipline. Currently, this process is mostly performed manually by exploring the scientific publications, requiring a substantial amount of time and effort. Due to the exponential growth of scientific literature, it has become almost impossible for a scientist to keep track of all research advances. As a result, scientists tend to deal with fragments of the literature according to their specialisation. Consequently, important and hidden associations among these fragmented knowledge that can be linked to produce significant scientific discoveries remain unnoticed. This doctoral work aims to develop a novel knowledge discovery approach that suggests most promising research pathways by analysing the existing scientific literature.

Concept maps are effective tools that assist learners in organising and representing knowledge. Recent efforts in the area of concept mapping work toward semi- or fully automated approaches to extract concept maps from various text sources such as text books. The motivation for this research is twofold: novice learners require substantial assistance from experts in constructing their own maps, introducing additional hurdles, and alternatively, the workload required by academics in manually constructing expert maps is substantial and repetitive. A key limitation of an automated concept map generation is the lack of an evaluation framework to measure the quality of concept maps. The most common evaluation mechanism is measuring the overlap between machine-generated elements (e.g. concepts) with expert maps using relevancy measures such as precision and recall. However, in the educational context, the majority of knowledge presented is relevant to the learner, resulting in a large amount of information being retrieved for knowledge organisation. Therefore, this paper introduces a machine-based approach to evaluate the relative importance of knowledge by comparing with human judgment. We introduce three ranking models and conclude that the structural features are positively correlated with human experts (rs ~ 1) for courses with rich content and good structure (well-fitted).

Concept maps are commonly used tools for organising and representing knowledge in order to assist meaningful learning. Although the process of constructing concept maps improves learners’ cognitive structures, novice students typically need substantial assistance from experts. Alternatively, expert-constructed maps may be given to students, which increase the workload of academic staff. To overcome this issue, automated concept map extraction has been introduced. One of the key limitations is the lack of an evaluation framework to measure the quality of machine-extracted concept maps. At present, researchers in this area utilise human experts’ judgement or expert-constructed maps as the gold standard to measure the relevancy of extracted knowledge. However, in the educational context, particularly in course materials, the majority of knowledge presented is relevant to the learner, resulting excess of information for knowledge organisation. Therefore, this paper introduces a novel machine-based approach which studies the relative importance of knowledge and organise knowledge hierarchically according to their importance. We compare machine’s prediction with human judgment based on experts’ knowledge and perception. Our ranking model utilise linguistic and structural features such as nouns, degree of co-occurrence and proximity. The results show that the machine’s predication positively correlates with human (rs~1) for well-fitted courses.