Faeze Brahman

The task of organizing a shuffled set of sentences into a coherent text is important in NLP and has been used to evaluate a machine’s understanding of causal and temporal relations. We present Reorder-BART (RE-BART), a sentence ordering framework which leverages a pre-trained transformer-based model to identify a coherent order for a given set of shuffled sentences. We reformulate the task as a conditional text-to-marker generation setup where the input is a set of shuffled sentences with sentence-specific markers and output is a sequence of position markers of the ordered text. Our framework achieves the state-of-theart performance across six datasets in Perfect Match Ratio (PMR) and Kendall’s tau (τ ) metric. We perform evaluations in a zero-shot setting, showcasing that our model is able to generalize well across other datasets. We additionally perform a series of experiments to understand the functioning and explore the limitations of our framework.1

Despite several advantages of online education, lack of effective student-instructor interaction, especially when students need timely help, poses significant pedagogical challenges. Motivated by this, we address the problems of automatically identifying posts that express confusion or urgency from Massive Open Online Course (MOOC) forums. To this end, we first investigate the extent to which the tasks of confusion detection and urgency detection are correlated so as to explore the possibility of utilizing a multitasking set-up. We then propose two LSTM-based multitask learning frameworks to leverage shared information and transfer knowledge across these related tasks. Our experiments demonstrate that the approaches improve over single-task models. Our best-performing model is especially useful in identifying posts that express both confusion and urgency, which can be of particular relevance for forum curation.

Automatically generating stories is a challenging problem that requires producing causally related and logical sequences of events about a topic. Previous approaches in this domain have focused largely on one-shot generation, where a language model outputs a complete story based on limited initial input from a user. Here, we instead focus on the task of interactive story generation, where the user provides the model mid-level sentence abstractions in the form of cue phrases during the generation process. This provides an interface for human users to guide the story generation. We present two content-inducing approaches to effectively incorporate this additional information. Experimental results from both automatic and human evaluations show that these methods produce more topically coherent and personalized stories compared to baseline methods.

ABSTRACT Global warming and pressing concern about CO2 emission along with increasing 12 fuel and oil cost have brought about great challenges for energy companies and 13 homeowners. In this regard, a potential candidate solution is widely used for 14 Distributed Energy Resources, which are capable of providing high quality, 15 low-cost heat and power to off-grid or remote facilities. To appropriately manage 16 thermal and electrical energy, a Smart Energy Management System (SEMS) with 17 hierarchical control scheme has been presented. The developed SEMS model 18 results in mixed integer non-linear programming optimization problem with the 19 objective function of minimizing the operation cost as well as considering 20 emissions. Moreover, the optimization problem has been solved for deterministic 21 and stochastic scheduling algorithms. The novelty of this work is basically reliant 22 on using data mining approach to reduce forecasting error. Several case studies 23 have been carried out to evaluate the performance of proposed data mining method 24 on both energy cost and expected cost.

The black-box nature of neural models has motivated a line of research that aims to generate natural language rationales to explain why a model made certain predictions. Such rationale generation models, to date, have been trained on dataset-specific crowdsourced rationales, but this approach is costly and is not generalizable to new tasks and domains. In this paper, we investigate the extent to which neural models can reason about natural language rationales that explain model predictions, relying only on distant supervision with no additional annotation cost for human-written rationales. We investigate multiple ways to automatically generate rationales using pre-trained language models, neural knowledge models, and distant supervision from related tasks, and train generative models capable of composing explanatory rationales for unseen instances. We demonstrate our approach on the defeasible inference task, a nonmonotonic reasoning task in which an inference may be strengthened or...

Despite the progress made in recent years in addressing natural language understanding (NLU) challenges, the majority of this progress remains to be concentrated on resource-rich languages like English. This work focuses on Persian language, one of the widely spoken languages in the world, and yet there are few NLU datasets available for this language. The availability of high-quality evaluation datasets is a necessity for reliable assessment of the progress on different NLU tasks and domains. We introduce ParsiNLU, the first benchmark in Persian language that includes a range of language understanding tasks—reading comprehension, textual entailment, and so on. These datasets are collected in a multitude of ways, often involving manual annotations by native speakers. This results in over 14.5k new instances across 6 distinct NLU tasks. Additionally, we present the first results on state-of-the-art monolingual and multilingual pre-trained language models on this benchmark and compare...

ABSTRACT Energy crisis along with environmental concerns are some principal motivations for introducing “energy hubs” by integrating energy production, conversion and storage technologies such as combined cooling, heating and power systems (CCHPs), renewable energy resources (RESs), batteries and thermal energy storages (TESs). In this paper, a residential energy hub model is proposed which receives electricity, natural gas and solar radiation at its input port to supply required electrical, heating and cooling demands at the output port. Augmenting the operational flexibility of the proposed hub in supplying the required demands, an inclusive demand response (DR) program including load shifting, load curtailing and flexible thermal load modeling is employed. A thermal and electrical energy management is developed to optimally schedule major household appliances, production and storage components (i.e. CCHP unit, PHEV and TES). For this purpose, an optimization problem has been formulated and solved for three different case studies with objective function of minimizing total energy cost while considering customer preferences in terms of desired hot water and air temperature. Additionally, a multi-objective optimization is conducted to consider consumer's contribution to CO2, NOx and SOx emissions. The results indicate the impact of incorporating DR program, smart PHEV management and TES on energy cost reduction of proposed energy hub model.