An approach framework of transfer learning, adversarial training and hierarchical multi-task learning - a case study of disinformation detection with offensive text

With society going online and disinformation getting accepted as a phenomena that we have to live with, there is a growing need to automatically detect offensive text on modern social media platforms. But the lack of enough balanced labeled data, constantly evolving socio-linguistic patterns and ever-changing definition of offensive text make it a challenging task. This is a common pattern witnessed in all disinformation detection tasks such as detection of propaganda, rumour, fake news, hate etc. The work described in this paper improves upon the existing body of techniques by bringing in an approach framework that can surpass the existing benchmarks. Firstly, it addresses the imbalanced and insufficient nature of available labeled dataset. Secondly, learning using relates tasks through multi-task learning has been proved to be an effective approach in this domain but it has the unrealistic requirement of labeled data for all related tasks. The framework presented here suitably uses transfer learning in lieu of multi-task learning to address this issue. Thirdly, it builds a model explicitly addressing the hierarchical nature in the taxonomy of disinformation being detected as that delivers a stronger error feedback to the learning tasks. Finally, the model is made more robust by adversarial training. The work presented in this paper uses offensive text detection as a case study and shows convincing results for the chosen approach. The framework adopted can be easily replicated in other similar learning tasks facing a similar set of challenges.

Finland, A Package Deal: Disciplinary Climate in Science Classes, Science Dispositions and Science Literacy

Finland’s educational prowess, though tempered by recent international assessments, has remained intact. This report focused on lessons that could be learned regarding secondary-level science education from the Program for International Student Assessment (PISA) 2015, science-focused assessment. That PISA iteration included not only science literacy but also students’ science dispositions (epistemology, enjoyment, interest, and self-efficacy) and the schools’ science climate measures (disciplinary climate and teaching support). Due to the hierarchical nature of the PISA data, multilevel models were employed in this Finnish study, involving 5582 students from 167 schools. Science dispositions (as outcome measures) were differently associated with teaching support and disciplinary climate (epistemology with neither; enjoyment and interest, with both). Science literacy (as an outcome measure) was associated with all four science dispositions, whether modeled with each science disposition separately or all four simultaneously. Science literacy was also associated with the disciplinary climate in science classes for all tested models. We concluded that, in the Finnish context, science dispositions and the disciplinary climate were predictive of science literacy. Furthermore, we presented evidence from the literature indicating that these conclusions may well extend to other international contexts.

Hierarchical organization of rhesus macaque behavior

Primatologists, psychologists and neuroscientists have long hypothesized that primate behavior is highly structured. However, fully delineating that structure has been impossible due to the difficulties of precision behavioral tracking. Here we analyzed a dataset consisting of continuous measures of the 3D position of fifteen body landmarks from two male rhesus macaques (Macaca mulatta) performing three different tasks in a large unrestrained environment over many hours. Using an unsupervised embedding approach on the tracked joints, we identified commonly repeated pose patterns, which we call postures. We found that macaques' behavior is characterized by 49 distinct identifiable postures, lasting an average of 0.6 seconds each. We found evidence that behavior is hierarchically organized, in that transitions between poses tend to occur within larger modules, which correspond to intuitively identifiably actions; these actions are in turn organized hierarchically. Our behavioral decomposition allows us to identify universal (cross-individual and cross-task) and unique (specific to each individual and task) principles of behavior. These results demonstrate the hierarchical nature of primate behavior and provide a method for the automated "ethogramming" of primate behavior.

Lineage Recording Reveals the Phylodynamics, Plasticity and Paths of Tumor Evolution

Tumor evolution is driven by the progressive acquisition of genetic and epigenetic alterations that enable uncontrolled growth, expansion to neighboring and distal tissues, and therapeutic resistance. The study of phylogenetic relationships between cancer cells provides key insights into these processes. Here, we introduced an evolving lineage-tracing system with a single-cell RNA-seq readout into a mouse model of Kras;Trp53(KP)-driven lung adenocarcinoma which enabled us to track tumor evolution from single transformed cells to metastatic tumors at unprecedented resolution. We found that loss of the initial, stable alveolar-type2-like state was accompanied by transient increase in plasticity. This was followed by adoption of distinct fitness associated transcriptional programs which enable rapid expansion and ultimately clonal sweep of rare, stable subclones capable of metastasizing to distant sites. Finally, we showed that tumors develop through stereotypical evolutionary trajectories, and perturbing additional tumor suppressors accelerates tumor progression by creating novel evolutionary paths. Overall, our study elucidates the hierarchical nature of tumor evolution, and more broadly enables the in-depth study of tumor progression.

What Neural Oscillations Can(not) Do for Syntactic Structure Building

Understanding what someone says requires relating words in the sentence to one another as instructed by grammatical rules of language. In recent years, a neurophysiological basis for this process has become a prominent topic of discussion in cognitive neuroscience. Current proposals about the neural mechanisms of syntactic structure building converge in assigning a key role to neural oscillations but differ in the exact function assigned to them. We discuss two types of approaches – oscillations for chunking and oscillations for multi-scale information integration – and evaluate their merits and limitations considering a fundamentally hierarchical nature of syntactic representations in natural language. We highlight insights that can provide a tangible starting point for a wide-scope neurocognitive model of syntactic structure building.

Towards a functional classification of behaviour: a taxonomy based on outcomes

The classification of behaviour has historically been done using one of the two approaches, either through the hypothetical causes (such as ‘instincts’, ‘drives’ and ‘needs’) or through the cataloguing of the observable form of behaviour using an ethogram. This article offers an alternative framework for classification of behaviour based upon only the behavioural outcomes. The framework is specified from first principles of a state-space approach, allowing us to discuss intermediate outcomes that may have instrumental value. This approach could provide a firmer foundation to consider the hierarchical nature of goals and allows us to address both the ‘how’ and the ‘why’ questions within a single framework. This taxonomy is designed to complement rather than replace existing attempts; the classification of behaviour by outcome is orthogonal to questions of the mechanisms of decision making or of the implementation of actions. This article specifies nine basic classes of behaviour and provides precise definitions for each of these. We then develop a formal language for the description of observed activities, the representation of behavioural hierarchies and for the analysis of possibility sets for achieving future goals. We follow up with some critique and discussion of the problems such a framework poses.

Striking a Balance: Sibling Emotionality and the Negotiation of Power in an Eighteenth-Century Noble Family of the Austrian Netherlands

This article examines the emotional dynamics between six siblings of the Merode family in the eighteenth century and the ways in which they used emotional expression as a mechanism to communicate and negotiate existing forms of power within their sibling group. Using the extensive personal correspondence between the siblings, the article explores how the hierarchical nature of their relationship related to expectations of unity, arguing that the siblings had to find a balance between keeping the peace and challenging inequality related to gender, age, and marital status. Sibling emotionality played an important role in negotiating these unequal power dynamics.

'Major Technological Challenge'

While archival description has been standardized in North America since the 1990s, the online environment has introduced new challenges in preserving the hierarchical nature of archival description. Many archival databases, in an attempt to mimic library discovery layers and web search engines, have collapsed the archival aggregation into item-level records, which can erase the context and structure of the content described. This article examines the communication breakdown that happens between the archivist and end user via online archival description and proposes solutions to improve multi-level description including the use of predicates in links, increased online instructional support, and the inclusion of digitized assets in archival description databases.

Making use of our data

Engineers are acquiring data at an ever-increasing rate: data from computational design studies; measurements data from manufacturing processes, development tests, and products in service; contemporary data and legacy data. In this paper, two recommendations are made to allow engineers to make better use of these expanding databases. First, we should build on the hierarchical nature of our data; we can navigate and filter the database using high level descriptors such as design specifications and performance metrics, and then request comparative plots of detailed data such as line, contour and surface plots. Second, we can speed up the rate at which we learn from data by making the visualisations dynamic; in so doing, we enable virtual experiments to be performed that highlight connections between input parameters, output metrics and physical mechanisms. The embodiment of these two principles in the open source project, dbslice, is described. Three example applications (an aerodynamic design study for a compressor stator; the application of machine learning to aid navigation of large databases; and visualisation of a database of snapshots from an unsteady simulation) are presented. In each case, the hierarchical data and dynamic visualisations allow the user to explore the database and experience the connections and patterns within it. By Making Use of Our Data to interactively navigate existing and new design spaces in this way, engineers can accelerate their response to the challenges of future products.