While communicating with a user, a task-oriented dialogue system has to track the user's needs at each turn according to the conversation history. This process called dialogue state tracking (DST) is crucial because it directly informs the downstream dialogue policy. DST has received a lot of interest in recent years with the text-to-text paradigm emerging as the favored approach. In this review paper, we first present the task and its associated datasets. Then, considering a large number of recent publications, we identify highlights and advances of research in 2021-2022. Although neural approaches have enabled significant progress, we argue that some critical aspects of dialogue systems such as generalizability are still underexplored. To motivate future studies, we propose several research avenues.

Bilingual terminologies are important machine translation resources in the field of e-commerce, which are usually either manually translated or automatically extracted from parallel data. The human translation is costly and e-commerce parallel corpora is very scarce. However, the comparable data in different languages in the same commodity field is abundant. In this paper, we propose a novel framework of extracting e-commercial bilingual terminologies from comparable data. Benefiting from the cross-lingual pre-training in e-commerce, our framework can make full use of the deep semantic relationship between source-side terminology and target-side sentence to extract corresponding target terminology. Experimental results on various language pairs show that our approaches achieve significantly better performance than various strong baselines.

Federated learning (FL) is an emerging technique used to collaboratively train a global machine learning model while keeping the data localized on the user devices. The main obstacle to FL's practical implementation is the Non-Independent and Identical (Non-IID) data distribution across users, which slows convergence and degrades performance. To tackle this fundamental issue, we propose a method (ComFed) that enhances the whole training process on both the client and server sides. The key idea of ComFed is to simultaneously utilize client-variance reduction techniques to facilitate server aggregation and global adaptive update techniques to accelerate learning. Our experiments on the Cifar-10 classification task show that ComFed can improve state-of-the-art algorithms dedicated to Non-IID data.

This study proposes the use of generative models (GANs) for augmenting the EuroSAT dataset for the Land Use and Land Cover (LULC) Classification task. We used DCGAN and WGAN-GP to generate images for each class in the dataset. We then explored the effect of augmenting the original dataset by about 10% in each case on model performance. The choice of GAN architecture seems to have no apparent effect on the model performance. However, a combination of geometric augmentation and GAN-generated images improved baseline results. Our study shows that GANs augmentation can improve the generalizability of deep classification models on satellite images.

Abstract-- To safely deploy legged robots in the real world it is necessary to provide them with the ability to reliably detect unexpected contacts and accurately estimate the corresponding contact force. In this paper, we propose a collision detection and identification pipeline for a quadrupedal manipulator. We first introduce an approach to estimate the collision time span based on band-pass filtering and show that this information is key for obtaining accurate collision force estimates. We then improve the accuracy of the identified force magnitude by compensating for model inaccuracies, unmodeled loads, and any other potential source of quasi-static disturbances acting on the robot. Quadrupedal robots have recently become sufficiently advanced to be deployed in unknown and unstructured environments, where they could operate alongside humans or phase (e.g., detecting a collision when there is none) or other robots.

In this work, we consider the problem of procedural content generation for video game levels. Prior approaches have relied on evolutionary search (ES) methods capable of generating diverse levels, but this generation procedure is slow, which is problematic in real-time settings. Reinforcement learning (RL) has also been proposed to tackle the same problem, and while level generation is fast, training time can be prohibitively expensive. We propose a framework to tackle the procedural content generation problem that combines the best of ES and RL. In particular, our approach first uses ES to generate a sequence of levels evolved over time, and then uses behaviour cloning to distil these levels into a policy, which can then be queried to produce new levels quickly. We apply our approach to a maze game and Super Mario Bros, with our results indicating that our approach does in fact decrease the time required for level generation, especially when an increasing number of valid levels are required.

Inverse text normalization (ITN) is an essential post-processing step in automatic speech recognition (ASR). It converts numbers, dates, abbreviations, and other semiotic classes from the spoken form generated by ASR to their written forms. One can consider ITN as a Machine Translation task and use neural sequence-to-sequence models to solve it. Unfortunately, such neural models are prone to hallucinations that could lead to unacceptable errors. To mitigate this issue, we propose a single-pass token classifier model that regards ITN as a tagging task. The model assigns a replacement fragment to every input token or marks it for deletion or copying without changes. We present a dataset preparation method based on the granular alignment of ITN examples. The proposed model is less prone to hallucination errors. The model is trained on the Google Text Normalization dataset and achieves state-of-the-art sentence accuracy on both English and Russian test sets. One-to-one correspondence between tags and input words improves the interpretability of the model's predictions, simplifies debugging, and allows for post-processing corrections. The model is simpler than sequence-to-sequence models and easier to optimize in production settings. The model and the code to prepare the dataset is published as part of NeMo project.

Although previous epidemiological studies have examined the potential risk factors that increase the likelihood of acquiring Helicobacter pylori infections, most of these analyses have utilized conventional statistical models, including logistic regression, and have not benefited from advanced machine learning techniques. We examined H. pylori infection risk factors among school children using machine learning algorithms to identify important risk factors as well as to determine whether machine learning can be used to predict H. pylori infection status. We applied feature selection and classification algorithms to data from a school-based cross-sectional survey in Ethiopia. The data set included 954 school children with 27 sociodemographic and lifestyle variables. We conducted five runs of tenfold cross-validation on the data. We combined the results of these runs for each combination of feature selection (e.g., Information Gain) and classification (e.g., Support Vector Machines) algorithms. The XGBoost classifier had the highest accuracy in predicting H. pylori infection status with an accuracy of 77%—a 13% improvement from the baseline accuracy of guessing the most frequent class (64% of the samples were H. Pylori negative.) K-Nearest Neighbors showed the worst performance across all classifiers. A similar performance was observed using the F1-score and area under the receiver operating curve (AUROC) classifier evaluation metrics. Among all features, place of residence (with urban residence increasing risk) was the most common risk factor for H. pylori infection, regardless of the feature selection method choice. Additionally, our machine learning algorithms identified other important risk factors for H. pylori infection, such as; electricity usage in the home, toilet type, and waste disposal location. Using a 75% cutoff for robustness, machine learning identified five of the eight significant features found by traditional multivariate logistic regression. However, when a lower robustness threshold is used, machine learning approaches identified more H. pylori risk factors than multivariate logistic regression and suggested risk factors not detected by logistic regression. This study provides evidence that machine learning approaches are positioned to uncover H. pylori infection risk factors and predict H. pylori infection status. These approaches identify similar risk factors and predict infection with comparable accuracy to logistic regression, thus they could be used as an alternative method.

Millions of shipments of nuclear and other radiological materials are moved in the U.S. every year for good reasons, including health care, power generation, research and manufacturing. But there remains the threat that bad actors in possession of stolen or illegally produced nuclear materials or weapons will try to smuggle them across borders for nefarious purposes. Texas A&M University researchers are making it harder for them to succeed. If border agents intercept illicit nuclear materials, investigators need to know who produced them and where they came from. Fortunately, nuclear materials carry certain forensic markers that can reveal valuable information, much like fingerprints can identify criminals.

It is a gripping scene of an orca viciously ripping out the liver of a nine-foot-long great white shark, as two other killer whales excitedly watch the once blue waters of South Africa's Mossel Bay turn blood red before the shark sinks to a the bottom of the sea – never to be seen again. The wild story was captured by a drone camera soaring above and now gives scientists a better understanding about why these apex-predators seem to be fleeing from this regions that was once the shark capital of the world. Orcas are known to feast on a great white shark liver, as to organ is are large, fatty and has become the whale's favorite dish – eight shark carcasses washing ashore the Western Cape in 2017 and all were missing their liver. The footage is part of marine biologist Alison Towner's long-term work with great whites. She shared on her Instagram page that the clip is'one of the most incredible pieces of natural history ever captured on film. The clip which is the first to show an orca eating a great white, is set to air on Discovery's Shark House Thursday night at 9pm ET, which is a day before the highly anticipated Shark Week begins.