DiMBERT: Learning Vision-Language Grounded Representations with Disentangled Multimodal-Attention

Vision-and-language (V-L) tasks require the system to understand both vision content and natural language, thus learning fine-grained joint representations of vision and language (a.k.a. V-L representations) is of paramount importance. Recently, various pre-trained V-L models are proposed to learn V-L representations and achieve improved results in many tasks. However, the mainstream models process both vision and language inputs with the same set of attention matrices. As a result, the generated V-L representations are entangled in one common latent space . To tackle this problem, we propose DiMBERT (short for Di sentangled M ultimodal-Attention BERT ), which is a novel framework that applies separated attention spaces for vision and language, and the representations of multi-modalities can thus be disentangled explicitly. To enhance the correlation between vision and language in disentangled spaces, we introduce the visual concepts to DiMBERT which represent visual information in textual format. In this manner, visual concepts help to bridge the gap between the two modalities. We pre-train DiMBERT on a large amount of image–sentence pairs on two tasks: bidirectional language modeling and sequence-to-sequence language modeling. After pre-train, DiMBERT is further fine-tuned for the downstream tasks. Experiments show that DiMBERT sets new state-of-the-art performance on three tasks (over four datasets), including both generation tasks (image captioning and visual storytelling) and classification tasks (referring expressions). The proposed DiM (short for Di sentangled M ultimodal-Attention) module can be easily incorporated into existing pre-trained V-L models to boost their performance, up to a 5% increase on the representative task. Finally, we conduct a systematic analysis and demonstrate the effectiveness of our DiM and the introduced visual concepts.

Task Modelling for Interactive System Design

Task models have been used for decades in interactive system design and there are several mature modelling approaches with corresponding tool support. However, in our own work, we have also experienced their limitations, especially in situations where task models are partial ancillary models and not primary artifacts. This was one of the motivations for this paper, which presents a systematic examination of literature to better understand the current place of task models in the continual evolution of user-centred software development practices. While overview work in this domain typically focuses on the analysis of representative task modelling notations and/or tools and relies on foundation papers, we apply a mixed top-down and bottom-up approach to identify relevant themes and trends in the use of task models over the last twenty-years. The paper identifies and discusses dominant patterns of use as well as gaps. It provides a comprehensive framing of both past and present trends in task modelling and supports those who want to incorporate task modelling in their own work. From this we identify areas of research that should receive greater attention in order to address future considerations.

Suitable Camera and Rotation Navigation for People with Visual Impairment on Looking for Something Using Object Detection Technique

For people with visual impairment, smartphone apps that use computer vision techniques to provide visual information have played important roles in supporting their daily lives. However, they can be used under a specific condition only. That is, only when the user knows where the object of interest is. In this paper, we first point out the fact mentioned above by categorizing the tasks that obtain visual information using computer vision techniques. Then, in looking for something as a representative task in a category, we argue suitable camera systems and rotation navigation methods. In the latter, we propose novel voice navigation methods. As a result of a user study comprised of seven people with visual impairment, we found that (1) a camera with a wide field of view such as an omnidirectional camera was preferred, and (2) users have different preferences in navigation methods.

Marcello Piacentinis Casa Madre dei Mutilati in Rom

The Casa Madre dei Mutilati in Rome is one of the lesser-known works of the architect Marcello Piacentini (1881–1960). Its history is closely linked to the dramatic events of the First World War and to the resulting association of Italian war invalids. The architecture reacts to the self-interpretation of the war invalids as living martyrs and has a meaningful and representative task. In the short period before the consolidation of Mussolini’s fascist regime, a modern style is manifested here, which attempts to combine classicism and avant-garde before the typical abstraction of the thirties. The tradition of Italian architectural history is still present without falling into a monumental formalism.

Law, Comparatism, and Epistemic Governance: There Is Critique and Critique

How many scholarly fields have experienced the disappointing fate of comparative law and continued in the grip of a demonstrably indigent epistemology for decades on end? After the early postmodernity witnessed their protracted servitude toLes Grands systèmes'sjejune classifications, facile correspondences, and meagre interpretive return — a predicament which, implausibly, endures in countries as diverse as Brazil, France, and Russia — law's comparatists began taking their epistemic orders from Hamburg and the Hamburgher diaspora. For fifty years or so, they have been gorged on a diet ofRechtsdogmatik, scientism, objectivity, neutrality, truth, and assorted shibboleths. As if these epistemic delusions were not problematic enough, the earlier, obsolete model was eventually revived although tweaked to focus on traditions instead of systems (or families). While critics were occasionally moved to chastise threadbare Hanseatic knowledge-claims — some expressing their concern in conspicuous venues, others harnessing prestigious institutional affiliations — comparative law's orthodoxy, somewhat extraordinarily, has hitherto been able to operate unencumbered by any epistemic challenge whose monographic exposition would have proved decidedly pre-eminent. It is the great merit of Gunter Frankenberg'sComparative Law As Critique, in crucial respects an account at once capital and extensive, that it interrupts, finally, the longstanding deployment of comparative law's mainstream imposture. Frankenberg's refutation is thus well worth restating, and the first part of this review wishes loyally to apply itself to this important representative task not least by affording the author much latitude to express himself in his own voice. Yet, Frankenberg's considerable critical integrity notwithstanding, this essay holds that his epistemic transgression remains too diffident. Specifically, five key concerns at least warranted more subversive epistemic commitments than Frankenberg allows. In the wake ofComparative Law As Critique, the second part of this commentary addresses these contentions with a view to making a case both for comparative law asstrongcritique and for the paradigmatic epistemic turn that has been persistently deferred within the field.

Game-Based Approaches, Pedagogical Principles and Tactical Constraints: Examining Games Modification

The purpose of this study was to analyze the effect of modification strategies based on the pedagogical principles of the Teaching Games for Understanding approach on tactical constraints of four 3v3 soccer small-sided games. The Game performance of 21 U-10 players was analyzed in a game similar to the adult game; one based on keeping-the-ball; another on penetrating-the-defense; and one on attacking-the-goal. Results showed that the modification of tactical problems had a significantly different effect on tactical-context adaptation (p < .005) and for developing passing, dribbling, shooting and getting free skills (p < .005). Small-sided games focused on keeping-the-ball and attacking-the-goal revealed a tactical complexity that was significantly different to the rest of the games (p < .005). With regard to the further investigation of the quality representative task design, these findings highlight the importance of knowing the effects that modification has on tactical constraints and the tactical complexity/technical difficulty involved in developing behaviors.

Reshaping Teachers' Mathematical Perceptions: Analysis of a Professional Development Task

As the focus of mathematics education moves from memorization toward reasoning and problem solving, professional development for in-service teachers must model these activities while simultaneously increasing participants' mathematical knowledge. We examine a representative task from a mathematics professional development course that uses rational number operation as an opportunity for problem solving and modeling. Transcripts exemplify the growth teachers make in deeply understanding the content–division of fractions–while engaging in guided reinvention and classroom discourse. We propose 4 interconnected qualities of this task that allow participants to engage in and reflect on the process of guided reinvention: (1) authentic context with multiple solution methods, including visual; (2) cognitive dissonance; (3) deep engagement; and (4) impact on mathematical knowledge for teaching.