Search

This article presents our recent work towards development of a controller for the new class of modernized excavators. There is an increasing demand to develop autonomous earth-moving machines that are: more efficient than the conventional ones and capable of operating with minimal supervision by operators. A fractional-order pro-portional-integral-derivative (FOPID) control is applied to a newly designed electro-hydrostatic actuator (EHA) circuit for these machines. We further employ the Oustaloup recursive method for which the parameters of the controller are chosen based on the modified Nelder-Mead optimization algorithm. Our experimental results indicate that the proposed controller is capable of controlling the speed of the link with a negligible error. This work paves the road for development of autonomous, energy efficient, off-road mobile machines.

We present a bilingual hybrid board game (English and Portuguese) as an innovative educational method to aid students in understanding concepts related to organic acids and bases. The proposed game promotes interaction between students in an enjoyable activity to review and learn the content, requiring minimal supervision from the professor.

We introduce a framework that predicts the goals behind observable human action in video. Motivated by evidence in developmental psychology, we leverage video of unintentional action to learn video representations of goals without direct supervision. Our approach models videos as contextual trajectories that represent both low-level motion and high-level action features. Experiments and visualizations show our trained model is able to predict the underlying goals in video of unintentional action. We also propose a method to "automatically correct" unintentional action by leveraging gradient signals of our model to adjust latent trajectories. Although the model is trained with minimal supervision, it is competitive with or outperforms baselines trained on large (supervised) datasets of successfully executed goals, showing that observing unintentional action is crucial to learning about goals in video. Project page: https://aha.cs.columbia.edu/

In this work, we attempt the segmentation of cardiac structures in late gadolinium-enhanced (LGE) magnetic resonance images (MRI) using only minimal supervision in a two-step approach. In the first step, we register a small set of five LGE cardiac magnetic resonance (CMR) images with ground truth labels to a set of 40 target LGE CMR images without annotation. Each manually annotated ground truth provides labels of the myocardium and the left ventricle (LV) and right ventricle (RV) cavities, which are used as atlases. After multi-atlas label fusion by majority voting, we possess noisy labels for each of the targeted LGE images. A second set of manual labels exists for 30 patients of the target LGE CMR images, but are annotated on different MRI sequences (bSSFP and T2-weighted). Again, we use multi-atlas label fusion with a consistency constraint to further refine our noisy labels if additional annotations in other modalities are available for a given patient. In the second step, we train a deep convolutional network for semantic segmentation on the target data while using data augmentation techniques to avoid over-fitting to the noisy labels. After inference and simple post-processing, we achieve our final segmentation for the targeted LGE CMR images, resulting in an average Dice of 0.890, 0.780, and 0.844 for LV cavity, LV myocardium, and RV cavity, respectively.

This study evaluates a cooperative learning approach for teaching anatomy to health science students incorporating small group and peer instruction based on the jigsaw method first described in the 1970's. Fifty-three volunteers participated in abdominal anatomy workshops. Students were given time to become an "expert" in one of four segments of the topic (sub-topics) by allocating groups to work-stations with learning resources: axial computerized tomography (CT) of abdominal structures, axial CT of abdominal blood vessels, angiograms and venograms of abdominal blood vessels and structures located within abdominal quadrants. In the second part of workshop, students were redistributed into "jigsaw" learning groups with at least one "expert" at each workstation. The "jigsaw" learning groups then circulated between workstations learning all sub-topics with the "expert" teaching others in their group. To assess abdominal anatomy knowledge, students completed a quiz pre- and post- workshop. Students increased their knowledge with significant improvements in quiz scores irrespective of prior exposure to lectures or practical classes related to the workshop topic. The evidence for long-term retention of knowledge, assessed by comparing end-semester examination performance of workshop participants with workshop nonparticipants, was less convincing. Workshop participants rated the jigsaw workshop highly for both educational value and enjoyment and felt the teaching approach would improve their course performance. The jigsaw method improved anatomy knowledge in the short-term by engaging students in group work and peer-led learning, with minimal supervision required. Reported outcomes suggest that cooperative learning approaches can lead to gains in student performance and motivation to learn. Anat Sci Educ 00: 000-000. © 2018 American Association of Anatomists.

Humanitarian disasters have been on the rise in recent years due to the effects of climate change and socio-political situations such as the refugee crisis. Technology can be used to best mobilize resources such as food and water in the event of a natural disaster, by semi-automatically flagging tweets and short messages as indicating an urgent need. The problem is challenging not just because of the sparseness of data in the immediate aftermath of a disaster, but because of the varying characteristics of disasters in developing countries (making it difficult to train just one system) and the noise and quirks in social media. In this paper, we present a robust, low-supervision social media urgency system that adapts to arbitrary crises by leveraging both labeled and unlabeled data in an ensemble setting. The system is also able to adapt to new crises where an unlabeled background corpus may not be available yet by utilizing a simple and effective transfer learning methodology. Experimentally, our transfer learning and low-supervision approaches are found to outperform viable baselines with high significance on myriad disaster datasets.

Graphical abstract, Drinking water distribution systems with long hydraulic retention times (HRTs) commonly encounter rapid microbiological-mediated monochloramine decay that results in microbial regrowth and nitrification with reduction in alkalinity. In this paper, we report the design and operation of a field-based pilot-scale distribution system (PDS) operated at flows that simulate long HRTs (∼10 days) to promote rapid microbiological monochloramine decay over long periods. The PDS is designed to enable the testing of chemical treatment for the control of nitrification and monochloramine decay. The PDS has two identical cylindrical polyethylene tanks (DS1 & DS2), each of 1 m diameter and 1.8 m height (∼1 kL) holding 900 m of polyethylene (PE) tubing with sampling points every 300 m intervals. Microbial mediated decay (determined by the Fm test) of monochloramine occurred as treated (alum coagulated and flocculated) and chloraminated water passed through the DSs. In this manuscript we report: • An inexpensive, flexible and compact system that can be readily set-up at a full-scale water treatment plant, requiring minimal supervision for operation. • A ‘draw & fill’ system for achieving control on HRT’s through the pipes.

This article considers the problem of correcting errors made by English as a Second Language writers from a machine learning perspective, and addresses an important issue of developing an appropriate training paradigm for the task, one that accounts for error patterns of non-native writers using minimal supervision. Existing training approaches present a trade-off between large amounts of cheap data offered by the native-trained models and additional knowledge of learner error patterns provided by the more expensive method of training on annotated learner data. We propose a novel training approach that draws on the strengths offered by the two standard training paradigms—of training either on native or on annotated learner data—and that outperforms both of these standard methods. Using the key observation that parameters relating to error regularities exhibited by non-native writers are relatively simple, we develop models that can incorporate knowledge about error regularities based on a small annotated sample but that are otherwise trained on native English data. The key contribution of this article is the introduction and analysis of two methods for adapting the learned models to error patterns of non-native writers; one method that applies to generative classifiers and a second that applies to discriminative classifiers. Both methods demonstrated state-of-the-art performance in several text correction competitions. In particular, the Illinois system that implements these methods ranked at the top in two recent CoNLL shared tasks on error correction. 1 We conduct further evaluation of the proposed approaches studying the effect of using error data from speakers of the same native language, languages that are closely related linguistically, and unrelated languages. 2

Current grammatical error correction (GEC) models typically consider the task as sequence generation, which requires large amounts of annotated data and limit the applications in data-limited settings. We try to incorporate contextual information from pre-trained language model to leverage annotation and benefit multilingual scenarios. Results show strong potential of Bidirectional Encoder Representations from Transformers (BERT) in grammatical error correction task.

The use of autonomous underwater vehicles is often hampered by stringent power constraints that limit the duration of their deployment. Here, we present a novel autonomous underwater charging system for robotic fish. The system features a charging station designed with form-fit claws to facilitate the robotic fish docking. A controller is implemented externally to monitor the battery level of the robotic fish, swimming autonomously in two dimensions. When the battery level is low, the controller commands the robotic fish to approach the charging station using video feedback from an overhead camera. The approach angle of the robotic fish to the charging station is optimized through a series of experiments, assessing the success of both docking and electrical contact, as well as the time to approach the charging station. To demonstrate the feasibility of the autonomous charging system, the robotic fish is monitored as it cycles through periods of charging and discharging while swimming in the tank. The proposed system is expected to find application in laboratory and educational settings, where robotic fish should operate with minimal supervision and assistance from technical staff over extended time periods.

This paper introduces Task 2 of the DCASE2019 Challenge, titled "Audio tagging with noisy labels and minimal supervision". This task was hosted on the Kaggle platform as "Freesound Audio Tagging 2019". The task evaluates systems for multi-label audio tagging using a large set of noisy-labeled data, and a much smaller set of manually-labeled data, under a large vocabulary setting of 80 everyday sound classes. In addition, the proposed dataset poses an acoustic mismatch problem between the noisy train set and the test set due to the fact that they come from different web audio sources. This can correspond to a realistic scenario given by the difficulty in gathering large amounts of manually labeled data. We present the task setup, the FSDKaggle2019 dataset prepared for this scientific evaluation, and a baseline system consisting of a convolutional neural network. All these resources are freely available., DCASE2019 Workshop

In this paper we present our audio tagging system for the DCASE 2019 Challenge Task 2. We propose a model consisting of a convolutional front end using log-mel-energies as input features, a recurrent neural network sequence encoder and a fully connected classifier network outputting an activity probability for each of the 80 considered event classes. Due to the recurrent neural network, which encodes a whole sequence into a single vector, our model is able to process sequences of varying lengths. The model is trained with only little manually labeled training data and a larger amount of automatically labeled web data, which hence suffers from label noise. To efficiently train the model with the provided data we use various data augmentation to prevent overfitting and improve generalization. Our best submitted system achieves a label-weighted label-ranking average precision (lwlrap) of 75.5% on the private test set which is an absolute improvement of 21.7% over the baseline. This system scored the second place in the teams ranking of the DCASE 2019 Challenge Task 2 and the fifth place in the Kaggle competition “Freesound Audio Tagging 2019” with more than 400 participants. After the challenge ended we further improved performance to 76.5% lwlrap setting a new state-of-the-art on this dataset.

The task of classifying X-ray data is a problem of both theoretical and clinical interest. Whilst supervised deep learning methods rely upon huge amounts of labelled data, the critical problem of achieving a good classification accuracy when an extremely small amount of labelled data is available has yet to be tackled. In this work, we introduce a novel semi-supervised framework for X-ray classification which is based on a graph-based optimisation model. To the best of our knowledge, this is the first method that exploits graph-based semi-supervised learning for X-ray data classification. Furthermore, we introduce a new multi-class classification functional with carefully selected class priors which allows for a smooth solution that strengthens the synergy between the limited number of labels and the huge amount of unlabelled data. We demonstrate, through a set of numerical and visual experiments, that our method produces highly competitive results on the ChestX-ray14 data set whilst drastically reducing the need for annotated data.

Cardiac rehabilitation is a program where the patient is supervised by a physical therapist while doing the exercise routine designed by their doctor. In the Philippines, a stand-alone equipment that can automate the doctor’s prescribed exercise routine while actively monitoring the patient’s vital sign is needed. The equipment could enable a safe and effective recovery with minimal supervision from the medical practitioners. This project aims to provide an automated aerobic exercise for cardiovascular rehabilitation patients through an adaptive cardiovascular training device. The automated exercise is based on the patient's heart rate which would be effectuated through the automatic adjustment of the load resistance of the cycling equipment. The equipment also includes blood pressure and oxygen saturation monitors to actively monitor the patient’s condition during the exercise.

Social media has become an inevitable part of individuals personal and business lives. Its benefits come with various negative consequences. One major concern is the prevalence of detrimental online behavior on social media, such as online harassment and cyberbullying. In this study, we aim to address the computational challenges associated with harassment detection in social media by developing a machine learning framework with three distinguishing characteristics. (1) It uses minimal supervision in the form of expert-provided key phrases that are indicative of bullying or non-bullying. (2) It detects harassment with an ensemble of two learners that co-train one another; one learner examines the language content in the message, and the other learner considers the social structure. (3) It incorporates distributed word and graph-node representations by training nonlinear deep models. The model is trained by optimizing an objective function that balances a co-training loss with a weak-supervision loss. We evaluate the effectiveness of our approach using post-hoc, crowdsourced annotation of Twitter, Ask.fm, and Instagram data, finding that our deep ensembles outperform previous non-deep methods for weakly supervised harassment detection.

Introduction Direct observation of procedural skills (DOPS) are tools designed by the Joint Advisory Group (JAG) to assess competence in endoscopy. These were expanded in July 2016 (new DOPS) to include those specific to paediatric gastroscopy (OGD). However, paediatric OGD DOPS assessments have not been validated. Methods We performed a prospective UK-wide analysis of formative paediatric OGD DOPS submitted to the JETS e-Portfolio over one-year (August 2016–2017). Scores were averaged across procedural domains (pre-procedural, procedural, post-procedural and endoscopic non-technical skills – ENTS), and according to skillset (cognitive or technical), and compared with the overall performance score using Spearman’s correlation. Overall scores were also compared by trainee grade. Results 157 DOPS assessments were completed by 20 unique trainers for 17 trainees. Overall performance score comprised: 1: Maximal supervision (4.5%), 2: Significant supervision (14.0%), 3: Minimal supervision (24.8%) and 4: Competent (56.7%). By domain, overall competence correlated most with mean scores for the ‘Insertion and Withdrawal’ domain (rho: 0.884, p Conclusions Competencies in paediatric OGD, as assessed within DOPS, vary in their correlation with overall competence. Strong correlations were reported for the ‘Insertion and Withdrawal’ domain, ‘Management of Complications’ items, and ‘’Technical’ skillsets. As assessors are completing the new DOPS in a consistent manner, with scores improving by seniority, this provides validity evidence for the new paediatric OGD DOPS.

Among the statistical approaches for math word problem solving, template based approaches have shown to be more robust against a wide spectrum of math word problems, while other approaches target simple arithmetic problems that compose of only one operation or equation. However, even template based systems are poor in performance for questions that contain different units to describe the same measurement. This paper presents a unit conflict resolution system to improve the performance and accuracy of template based systems under minimal supervision. To illustrate the importance of unit conflict resolution for math word problems, we have annotated a new dataset of 385 algebra word problems. We evaluate the performance of our approach both on a benchmark dataset and this new dataset. Experimental results show that integration of our system to an existing automatic math word problem solver outperforms state-of-the-art results when the dataset contains different units to describe the same measurement.

BACKGROUND:\ud The global strategy for elimination of lymphatic filariasis is by annual mass drug administration (MDA). Effective implementation of this strategy in endemic areas reduces Wuchereria bancrofti in the blood of infected individuals to very low levels. This minimises the rate at which vectors successfully pick microfilariae from infected blood, hence requiring large mosquito numbers to detect infections. The aim of this study was to assess the feasibility of using trained community vector collectors (CVCs) to sample large mosquito numbers with minimal supervision at low cost for potential scale-up of this strategy.\ud \ud METHODS:\ud CVCs and supervisors were trained in mosquito sampling methods, i.e. human landing collections, pyrethrum spray collections and window exit traps. Mosquito sampling was done over a 13-month period. Validation was conducted by a research team as quality control for mosquitoes sampled by CVCs. Data were analyzed for number of mosquitoes collected and cost incurred by the research team and CVCs during the validation phase of the study.\ud \ud RESULTS:\ud A total of 31,064 and 8720 mosquitoes were sampled by CVCs and the research team, respectively. We found a significant difference (F(1,13) = 27.1606, P = 0.0001) in the total number of mosquitoes collected from southern and northern communities. Validation revealed similar numbers of mosquitoes sampled by CVCs and the research team, both in the wet (F(1,4) = 1.875, P = 0.309) and dry (F(1,4) = 2.276, P = 0.258) seasons in the southern communities, but was significantly different for both wet (F(1,4) = 0.022, P = 0.005) and dry (F(1,4 ) = 0.079, P = 0.033) seasons in the north. The cost of sampling mosquitoes per season was considerably lower by CVCs compared to the research team (15.170 vs 53.739 USD).\ud \ud CONCLUSIONS:\ud This study revealed the feasibility of using CVCs to sample large numbers of mosquitoes with minimal supervision from a research team at considerably lower cost than a research team for lymphatic filariasis xenomonitoring. However, evaluation of the selection and motivation of CVCs, acceptability of CVCs strategy and its epidemiological relevance for lymphatic filariasis xenomonitoring programmes need to be assessed in greater detail.

Higher educational institutions aim to develop students who take responsibility of their own learning. Such students can manage and assess their own learning and they often perform better than others. Experts believe that one way to develop independent learners are through directed independent learning activities which students are required to complete with some or minimal supervision from instructors. This paper investigates students' perceptions on the use of digitized self-learning resources in enhancing students' independent learning. The participants of this study were 37 Chemistry students enrolled in the Australian Matriculation (AUSMAT) Programme at Sunway College, Malaysia. As part of their directed independent learning, participants were provided with paper-based self-learning material containing a compilation of examination questions on Oxidation and Reduction (Redox) chapter, marking key and solution videos provided in a QR code, as well as, similar past year questions for practice. Data collection tools were a survey questionnaire which consisted of 5-point Likert scale and open-ended questions and teacher reflective notes. Analysis of data revealed that even though most students favoured the use of digitized self-learning resources to support their learning, some were reluctant and preferred the use of more traditional methods of learning. The study provides valuable insights into the importance of preparing students for independent learning and the challenges in preparing digitized self-learning resources that would encourage students to become autonomous learners.

As surgery residents graduate and begin their careers as junior attending surgeons, the question of whether a surgeon can complete a case alone still lingers. Allowing autonomy during residency answers this question. The purpose of this study was to gather input from general surgery residency program directors on how they achieve autonomy for residents in their programs. An online survey of 18 questions was sent to all general surgery residency program directors in the United States between April and June of 2013 via e-mail. Questions were asked regarding classification of autonomy, percentage of case completed by the resident independently, and in what area a resident worked with minimal supervision. Of the 202 delivered, 85 program directors were responded (42%). Seventy-eight per cent of programs classified a resident as surgeon junior whether the resident completed more than 50 per cent of the case. Most classified autonomy as either the resident completing >75 per cent of a case (41%) or completing the critical steps of a surgery (41%). Eighty-eight per cent stated that chief residents completed the majority of cases under supervision, whereas only 12 per cent stated the chief had autonomy in the operating room and also acted as teaching assistant. While, 60 per cent stated their chief residents did not work in any area of the hospital independently. Despite differences in how autonomy is defined among programs, most program directors feel that their chief residents do not achieve complete autonomy. Programs should allow their residents to work in a progressive responsibility as they progress into their fourth and fifth years of residency to achieve autonomy.

We present an approach to effectively use millions of images with noisy annotations in conjunction with a small subset of cleanly-annotated images to learn powerful image representations. One common approach to combine clean and noisy data is to first pre-train a network using the large noisy dataset and then fine-tune with the clean dataset. We show this approach does not fully leverage the information contained in the clean set. Thus, we demonstrate how to use the clean annotations to reduce the noise in the large dataset before fine-tuning the network using both the clean set and the full set with reduced noise. The approach comprises a multi-task network that jointly learns to clean noisy annotations and to accurately classify images. We evaluate our approach on the recently released Open Images dataset, containing ~9 million images, multiple annotations per image and over 6000 unique classes. For the small clean set of annotations we use a quarter of the validation set with ~40k images. Our results demonstrate that the proposed approach clearly outperforms direct fine-tuning across all major categories of classes in the Open Image dataset. Further, our approach is particularly effective for a large number of classes with wide range of noise in annotations (20-80% false positive annotations)., CVPR 2017

This study is a prospective case series analyzing the effects of a comprehensive lifestyle intervention program in three patients with chronic paraplegia having major risks for the cardiometabolic syndrome (CMS). Individuals underwent an intense 6-month program of circuit resistance exercise, nutrition using a Mediterranean diet and behavioral support, followed by a 6-month extension (maintenance) phase involving minimal support. The primary goal was a 7% reduction of body mass. Other outcomes analyzed insulin resistance using the HOMA-IR model, and plasma levels of fasting triglycerides and high-density lipoprotein cholesterol. All participants achieved the goal for 7% reduction of body mass and maintained the loss after the MP. Improvements were observed in 2/3 subjects for HOMA-IR and high-density lipoprotein cholesterol. All participants improved their risk for plasma triglycerides. We conclude, in a three-person case series of persons with chronic paraplegia, a lifestyle intervention program involving circuit resistance training, a calorie-restrictive Mediterranean-style diet and behavioral support, results in clinically significant loss of body mass and effectively reduced component risks for CMS and diabetes. These results were for the most part maintained after a 6-month MP involving minimal supervision.

In this era, the development of new technological is helping in simplifications of everyday processes including the industrial processes. One such application of technology is in the food industry for both in manufacturing and also in quality assessment and management of synthesized food products. In this field one of the major problems is contamination of both raw materials and products produced in the food industry. A number of chemical sensors are commercially available for field measurements of chemical species (e.g., portable gas chromatographs, surface-wave acoustic sensors, optical instruments), few have been adapted for use in environments for long-term monitoring or remediation applications due to the requirement of significant amounts of manual labor and supervision to operate. Instead, we seek low-cost sensors that have detection limit at very low level concentrations and operated with minimal supervision, which yield continuous real-time monitoring capabilities. In recent, nanotechnology has helped in the development of better contaminant sensors, i.e., sensors that can detect contaminations in food raw materials and food products. Biosensor is low-cost, quick, and reliable compared to existing methods and it can be modified to detect the presence of single or multiple substances using very small samples. Many of the biosensors were available for detecting and monitoring of toxins, contaminants, and residues in food products. Biosensors, also called molecular sensors, can detect the presence of specific chemicals, viruses, or molecules in air or liquids. Some of the sensors/materials like Wireless Networks, microwave coplanar, smart polymer, and other novel biosensors were used to detect the contamination in food products. This chapter presents a review of sensors and technologies that are capable of detecting and monitoring contaminant in the food industry. The review also focuses on the effect of pH, odor, taste, temperature, microbial nature, and other physicochemical parameters in the food products as well sensor quality and its life span. Along with effects, nano-encapsulated enzymes can be used to check and degrade the possible contaminants.

We present a novel approach to object classification and detection which requires minimal supervision and which combines visual texture cues and shape information learned from freely available unlabeled web search results. The explosion of visual data on the web can potentially make visual examples of almost any object easily accessible via web search. Previous unsupervised methods have utilized either large scale sources of texture cues from the web, or shape information from data such as crowdsourced CAD models. We propose a two-stream deep learning framework that combines these cues, with one stream learning visual texture cues from image search data, and the other stream learning rich shape information from 3D CAD models. To perform classification or detection for a novel image, the predictions of the two streams are combined using a late fusion scheme. We present experiments and visualizations for both tasks on the standard benchmark PASCAL VOC 2007 to demonstrate that texture and shape provide complementary information in our model. Our method outperforms previous web image based models, 3D CAD model based approaches, and weakly supervised models.

In this paper, we describe a system for automatic construction of user disease progression timelines from their posts in online support groups using minimal supervision. In recent years, several online support groups have been established which has led to a huge increase in the amount of patient-authored text available. Creating systems which can automatically extract important medical events and create disease progression timelines for users from such text can help in patient health monitoring as well as studying links between medical events and users’ participation in support groups. Prior work in this domain has used manually constructed keyword sets to detect medical events. In this work, our aim is to perform medical event detection using minimal supervision in order to develop a more general timeline construction system. Our system achieves an accuracy of 55.17%, which is 92% of the performance achieved by a supervised baseline system.

CASE Steven is an 11-year-old boy who was adopted from an orphanage in Eastern Europe 8 years ago, when he was 3 years old along with his biological sister who is 1 year older. You have cared for him in your practice since that time seeing them annually for well child care. His mother reports that the first 2 years after adoption were very difficult, and Steven has never really bonded with her or her husband. Currently, he is reported to steal possessions of family members and instigate arguments with his older sister and act defiantly. The parents have put locks on their bedroom door due to fears of violence, primarily by Steven's older sister. Steven's mother reports that she and her husband no longer attempt to "parent" the children and provide minimal supervision to avoid conflict. In school, Steven is reported to have symptoms of inattention and hyperactivity that improved with treatment with stimulant medication and some learning challenges, but otherwise he is described as a lovely boy who has good relationships with his teachers.They present to urgent care clinic when mom stated "We have had enough." That day both children were caught taking $10 from the mother's wallet. On further discussion, it was discovered that they had done this several times over the last month to buy ice cream for neighborhood friends to "welcome them" to the neighborhood. Steven's mother presents today stating that the family has exhausted therapy services and are experiencing financial hardship secondary to investing time and money into behavioral and psychiatric services for Steven and his sister. They want to dissolve the adoption and have the children removed from their home. They seek guidance on how to do this. What would be your initial approach with the parents?

In this paper, we compare delexicalized transfer and minimally supervised parsing techniques on 32 different languages from Universal Dependencies treebank collection. The minimal supervision is in adding handcrafted universal grammatical rules for POS tags. The rules are incorporated into the unsupervised dependency parser in forms of external prior probabilities. We also experiment with learning this probabilities from other treebanks. The average attachment score of our parser is slightly lower then the delexicalized transfer parser, however, it performs better for languages from less resourced language families (non-Indo-European) and is therefore suitable for those, for which the treebanks often do not exist.

Objectives ACVS Diplomates were surveyed to identify a complete list of core surgical skills, determine the expected frequency of use of these skills, and ascertain the level of skills proficiency expected of entry-level practicing veterinarians. Study Design Internet-based survey. Sample Population Randomly chosen ACVS Diplomates (n = 750). Methods Rating scale items were used to construct the main portion of the survey with 26 skills that were scored by participants based on rankings of proficiency expected of entry-level veterinarians, as well as expected frequency of use of these skills. Additional demographic questions were included to gather information regarding gender, practice type, year of veterinary school, year of ACVS membership, job description, primary surgical discipline, practice location, average number of procedures performed on a weekly basis, and number of new graduates mentored in the last 5 years. Results Of 337 Diplomates (44.9%) who responded to the survey, >60% expected entry-level graduates to have good skills with minimal supervision or complete skills for 21 of the 26 skills categories listed. More than 60% of respondents ranked 7 skills with a lower frequency score (seldom or occasionally use). Orthopedic instrument handling and fixation skills as well as electrosurgical and laser skills received some of the lowest expected proficiency and frequency rankings. No additional skills categories were identified in the open-ended survey questions. Conclusions A complete list of core surgical skills was identified. There was broad consensus between boarded surgeons irrespective of their practice type, experience, or discipline for ranked proficiency and frequency scores among the core surgical skills expected of entry-level veterinarians.

There remains a lag in the development and implementation of effective strategies to serve lesbian, gay, bisexual, transgender, and questioning (LGBTQ) adolescents in schools. Although practitioners and scholars have begun to discuss how school counselors can provide effective services for LGBTQ students, there is minimal supervision literature to guide this work. This article presents an application of supervision that integrates LGBTQ Responsive School Counseling across the supervision of a comprehensive developmental school counseling program. Using the American School Counseling Association's National Model as a guideline, the authors provide a vehicle for supervisors to structure their work with professional school counselors.

PurposeThis paper aims to redefine the peer reference model – using student assistants to staff the reference desk – by expanding it to include upper‐level undergraduate students from various disciplines providing full reference service with minimal supervision as well as performing paraprofessional‐level duties.Design/methodology/approachSet within a backdrop of budget cuts, severe time and staffing limitations, and increasing demands for library services, the paper assesses the processes and results of implementing this new model of peer reference at a medium‐sized academic library.FindingsThe findings suggest that undergraduate students are not only capable but perhaps optimal at providing high‐quality reference service to their peers.Originality/valueThis paper redefines the peer reference model, using a case study carried out in a US university library.

243 Background: Patients prescribed oral chemotherapy receive minimal supervision and support for adherence. We developed a smartphone mobile app consisting of medication reminders, symptom reporting, clinician communication, and educational resources for improving adherence and symptom management. In this study, we sought to identify potential correlates of app usage. Methods: From 12/2014 to 08/2016, 181 patients with diverse malignancies prescribed oral chemotherapy enrolled in a randomized trial to test the efficacy of a smartphone mobile app to improve adherence and symptoms. Medication adherence was monitored using electronic pill caps (MEMScaps). At baseline and 12 weeks, participants completed a measure of treatment satisfaction (FACIT-TS). We gathered clinical and treatment data from the electronic health record. Through the app we collected data on engagement: total number of minutes, days of usage, and number of completed symptom reports. We calculated Pearson correlations to identify associations between app engagement, using a latent variable for overall engagement, with adherence, treatment satisfaction and healthcare utilization. Results: 91 patients were randomly assigned to receive the mobile app. Over the 12-week study period, participants used the mobile app for a mean of 59 minutes (SD = 68) and 21.75 discrete days (SD = 21.24 days), completing 15.92 symptom reports (SD = 14.15) on average. Higher app engagement was associated with having a lower proportion of missed doses (r = -.248, p = .035) over the study period. Patients who became less satisfied with their communication with physicians (r = -.278, p = .019) and nurses (r = -.238, p = .047) used the app more. Higher app engagement was associated with fewer emergency department visits resulting in hospitalizations (r = -.252, p = .029). Conclusions: Among patients prescribed oral chemotherapy, higher usage of a mobile app targeting adherence was related to fewer missed doses and ED visits leading to hospitalization. Fostering patient engagement with the app may also serve to buffer difficulties in patient-clinician communication about care. Clinical trial information: NCT02157519.

A significant challenge that faces any teacher of introductory programming is the diversity of the class. At one extreme there will be students who have never programmed before, while at the other there will be students who have many years experience of programming. Handling this diversity is difficult. The temptation for the instructor is often to focus on the novice group and to assume that the others will get by with minimal supervision. This is understandable, but it can be risky. There is a very real risk that the neglected group of experienced programmers become bored and disengage from the course. At the worst, they can lose motivation and fail or drop out altogether. This paper describes and presents the outcomes of a project aimed at challenging the more experienced programmers in four introductory programming classes at four different UK institutions. The project took the form of a competition in which students were asked to devise and solve a series of programming challenges.

Entity classification, like many other important problems in NLP, involves learning classifiers over sparse highdimensional feature spaces that result from the conjunction of elementary features of the entity mention and its context. In this paper we develop a low-rank regularization framework for training maxentropy models in such sparse conjunctive feature spaces. Our approach handles conjunctive feature spaces using matrices and induces an implicit low-dimensional representation via low-rank constraints. We show that when learning entity classifiers under minimal supervision, using a seed set, our approach is more effective in controlling model capacity than standard techniques for linear classifiers.

Nearly all work in unsupervised grammar induction aims to induce unlabeled dependency trees from gold part-of-speechtagged text. These clean linguistic classes provide a very important, though unrealistic, inductive bias. Conversely, induced clusters are very noisy. We show here, for the first time, that very limited human supervision (three frequent words per cluster) may be required to induce labeled dependencies from automatically induced word clusters.

The 2011/12 Antarctic field season saw the first use of a new British Antarctic Survey (BAS) ice-shelf hot-water drill system on the Larsen C and George VI ice shelves. Delivering 90 L min−1 at 80°C, a total of five holes >30 cm in diameter at three locations were successfully drilled through almost 400 m of ice to provide access to the underlying ocean, including the first access beneath the Larsen C ice shelf. These access holes enabled the deployment of instruments to measure sea-water conductivity, temperature, depth and microstructure, the collection of water samples and up to 2.9 m long sediment cores, before long-term oceanographic moorings were deployed. The simple modular design allowed for Twin Otter aircraft deployment, rapid assembly and commissioning of the system, which proved highly reliable with minimal supervision. A number of novel solutions to various operational sub-ice-shelf profiling and mooring deployment issues were successfully employed through the hot-water drilled access holes to aid the positioning, recovery and deployment of instruments. With future activities now focusing on the Filchner–Ronne Ice Shelf, the drill has been upgraded from its current 500 m capability to 1000 m with additional drill hose and further generator, pumping and heating modules.

The authors undertook a series of empirical studies to examine how students experience homework at various grade levels. The research casts a different light on the century-old practice of doing homework, suggesting it is the quintessential job of childhood. Homework creates a situation where the child must complete assigned tasks under minimal supervision and after little initial training. Doing well on that job gets one further along in school. As in the workplace, when children move from beginners to experts with homework, they demonstrate responsibility and become skilled at managing tasks. These positive outcomes enhance the intended deepening of students' subject matter knowledge. However, the authors assert that another virtue of homework is that it can prepare children for jobs they will have one day; it may develop an aptitude for gainful employment.

As students progress through the midwifery programme, more and more study will be self-directed or independent—none more so than the dissertation which is likely to be the largest assessment, worked on over a number of months. Keeping on track can be difficult and a number of students leave this until the last minute due to poor planning or working on other pieces of work. The aim of this article is to give students hints and tips for keeping on track when working with minimal supervision.

Summary Background Minimal faculty member supervision of students refers to a method of instruction in which the patient–student encounter is not directly supervised by a faculty member, and presents a feasible solution in clinical teaching. It is unclear, however, how such practices are perceived by patients and how they affect student learning. Context We aimed to assess patient and medical student perceptions of clinical teaching with minimal faculty member supervision. Questionnaires focusing on the perception of students’ performance were administered to patients pre- and post-consultation. Students’ self-perceptions on their performance were obtained using a questionnaire at the end of the consultation. Innovation Before encounters with students, 22 per cent of the 95 patients were not sure if they would feel comfortable or trust the students; after the consultation, almost all felt comfortable (97%) and relied on the students (99%). The 81 students surveyed agreed that instruction with minimal faculty member supervision encouraged their participation and engagement (86%). They expressed interest in knowing patients’ opinions about their performance (94%), and they felt comfortable about being assessed by the patients (86%). Implications The minimal faculty member supervision model was well accepted by patients. Responses from the final-year students support the use of assessments that incorporate feedback from patients in their overall clinical evaluations.

Over the recent past, information extraction (IE) systems such as Nell and ReVerb have attained much success in creating large knowledge resources with minimal supervision. But, these resources in general, lack schema information and contain facts with high degree of ambiguity which are often difficult to interpret. Whereas, Wikipedia-based IE projects like DBpedia and Yago are structured, have disambiguated facts with unique identifiers and maintain a well-defined schema. In this work, we propose a probabilistic method to integrate these two types of IE projects where the structured knowledge bases benefit from the wide coverage of the semi-supervised IE projects and the latter benefits from the schema information of the former.

This paper describes an approach for automatic construction of dictionaries for Named Entity Recognition (NER) using large amounts of unlabeled data and a few seed examples. We use Canonical Correlation Analysis (CCA) to obtain lower dimensional embeddings (representations) for candidate phrases and classify these phrases using a small number of labeled examples. Our method achieves 16.5% and 11.3% F-1 score improvement over co-training on disease and virus NER respectively. We also show that by adding candidate phrase embeddings as features in a sequence tagger gives better performance compared to using word embeddings., Comment: In 14th Conference of the European Chapter of the Association for Computational Linguistic, 2014

Objective.To investigate the effect of workplace neck/shoulder strength training with and without regular supervision on neck/shoulder pain and headache among office workers.Method.A 20-week cluster randomized controlled trial among 351 office workers was randomized into three groups: two training groups with the same total amount of planned exercises three times per week (1) with supervision (3WS) throughout the intervention period, (2) with minimal supervision (3MS) only initially, and (3) a reference group (REF). Main outcome is self-reported pain intensity in neck and shoulder (scale 0–9) and headache (scale 0–10).Results.Intention-to-treat analyses showed a significant decrease in neck pain intensity the last 7 days in 3MS compared with REF: −0.5 ± 0.2(P<0.02)and a tendency for 3WS versus REF: −0.4 ± 0.2 (P<0.07). Intensity of headache the last month decreased in both training groups: 3WS versus REF: −1.1 ± 0.2 (P<0.001) and 3MS versus REF: −1.1 ± 0.2 (P<0.001). Additionally, days of headache decreased 1.0 ± 0.5 in 3WS and 1.3 ± 0.5 in 3MS versus REF. There were no differences between the two training groups for any of the variables.Conclusion.Neck/shoulder training at the workplace reduced neck pain and headache among office workers independently of the extent of supervision. This finding has important practical implications for future workplace interventions.

In many parts of the world, livestock guardian dogs (LGDs) are a relatively new and increasingly popular method for controlling the impact of wild predators on livestock. On large grazing properties in Australia, LGDs are often allowed to range freely over large areas, with minimal supervision by their owners. How they behave in this situation is mostly unknown. We fitted free-ranging Maremma sheepdogs with GPS tracking collars on three properties in Victoria, Australia; on two properties, four sheep were also fitted with GPS collars. We investigated how much time the Maremmas spent with their livestock, how far they moved outside the ranges of their stock, and tested whether they use their ranges sequentially, which is an effective way of maintaining a presence over a large area. The 95% kernel isopleth of the Maremmas ranged between 31 and 1161 ha, the 50% kernel isopleth ranged between 4 and 252 ha. Maremmas spent on average 90% of their time in sheep paddocks. Movements away from sheep occurred mostly at night, and were characterised by high-speed travel on relatively straight paths, similar to the change in activity at the edge of their range. Maremmas used different parts of their range sequentially, similar to sheep, and had a distinct early morning and late afternoon peak in activity. Our results show that while free-ranging LGDs spend the majority of their time with livestock, movements away from stock do occur. These movements could be important in allowing the dogs to maintain large territories, and could increase the effectiveness of livestock protection. Allowing LGDs to range freely can therefore be a useful management decision, but property size has to be large enough to accommodate the large areas that the dogs use.

The aim of this study is to develop a new Smart Heat Therapy device that is able to expedite the process, increase the effectiveness and reduce the limitations of existing devices used in physiotherapy procedure to stimulate aching muscle due to improper blood circulation within certain parts of the body. This new compact, portable, economical and user friendly device will promote blood circulation by continuously applying low level heat energy to the body to enlarge the blood vessel at targeted area with water as the heating medium. The advanced feature of the device is its capability to deliver consistent amount of heat energy required for selected duration of therapy session prescribed to the patient. Hence the patient can select the temperature and time required during therapy session. The vital feature of the device is the incorporation of safety features that makes it suitable to be used under minimal supervision either in infirmaries or at home.