Search

The distributed representation of symbols is one of the key technologies in machine learning systems today, playing a pivotal role in modern natural language processing. Traditional word embeddings associate a separate vector with each word. While this approach is simple and leads to good performance, it requires a lot of memory for representing a large vocabulary. To reduce the memory footprint, the default embedding layer in spaCy is a hash embeddings layer. It is a stochastic approximation of traditional embeddings that provides unique vectors for a large number of words without explicitly storing a separate vector for each of them. To be able to compute meaningful representations for both known and unknown words, hash embeddings represent each word as a summary of the normalized word form, subword information and word shape. Together, these features produce a multi-embedding of a word. In this technical report we lay out a bit of history and introduce the embedding methods in spaCy in detail. Second, we critically evaluate the hash embedding architecture with multi-embeddings on Named Entity Recognition datasets from a variety of domains and languages. The experiments validate most key design choices behind spaCy's embedders, but we also uncover a few surprising results.

This exploratory article highlights how problem-based learning (PBL) provided the pedagogical framework used to design and interpret learning analytics from [GAME], a collaborative game-based learning environment centered on supporting science inquiry. Results showed that students learned the targeted content after engaging with the game-based learning environment. Clusters based on the PCA revealed four main ways of engaging in the game-based learning environment: (1) high collaborative sensemaking, (2) low collaborative sensemaking, (3) high self-directed actions and low collaborative sensemaking, and (4) moderate collaborative sensemaking. Although students who scored lower spent more time in chat and engaging in voting tasks, qualitative analysis revealed that these groups engaged in metacognitive discussions that left less time for discussions about the content, whereas groups that scored higher in the post-test engaged in more in-depth scientific discussions. These findings have implications for designing adaptive support that responds to students’ interactions with in-game activities.

This symposium will provide opportunities for discussion about how Artificial Intelligence can support ambitious learning practices in CSCL. To the extent that CSCL can be a lever for educational equitable educational change, AI needs to be able to support the kinds of practices that afford agency to students and teachers. However, AI also brings to the fore the need to consider equity and ethics. This interactive session will provide opportunities to discuss these issues in the context of the examples presented here.

Game-based learning environments hold significant potential for supporting K-12 computer science (CS) education by providing CS learning experiences embedded within engaging virtual worlds. However, many game-based learning environments do not adaptively support individual students based on their specific knowledge and skills. Often, this is because creating game levels is highly labor-intensive, which limits the number of levels created to support student learning. Procedural content generation (PCG) is a promising direction for addressing this challenge by dynamically creating game levels that address specific student needs without requiring extensive development effort. In this paper, we investigate a PCG framework driven by answer set programming (ASP), a variant of logic programming that utilizes well-formed logical rules to express constraints for valid game levels. We demonstrate how variations in CS learning objectives and game-playing skills can be incorporated into ASP-based rules to generate learner-adaptive levels in a middle-grades CS game-based learning environment. Evaluations of the generated levels suggest that the ASP-based level generator not only reliably generates desired CS educational game levels but also synthesizes a large set of diverse game levels. The findings suggest that the ASP-based PCG approach has considerable promise for creating highly engaging and adaptive game-based learning experiences for K-12 CS education.

Accurately recognizing learner affect is critically important for enabling affect-responsive learning environments to support student learning and engagement. Multimodal affect detection combining sensor-based and sensor-free approaches has shown significant promise in both laboratory and classroom settings. However, important questions remain regarding which data channels are most predictive and how they should be combined. In this paper, we investigate a multimodal affect detection framework that integrates motion tracking-based posture data and interaction-based trace data to recognize the affective states of students engaged with a game-based learning environment for emergency medical training. We compare several machine learning-based affective models using competing feature-level and decision-level multimodal data fusion approaches. Results indicate that multimodal affect detectors induced using joint feature representations from posture-based and interaction-based data channels yield improved accuracy relative to unimodal models across several learner-centered affective states. These findings point toward implications for the design of multimodal affect-responsive learning environments that support learning and engagement.

Game-based learning environments enable students to engage in authentic, inquiry-based learning. Reflective thinking serves a critical role in inquiry-based learning by encouraging students to think critically about their knowledge and experiences in order to foster deeper learning processes. Free-response reflection prompts can be embedded in game-based learning environments to encourage students to engage in reflection and externalize their reflection processes, but automatically assessing student reflection presents significant challenges. In this paper, we present a framework for automatically assessing students’ written reflection responses during inquiry-based learning in Crystal Island, a game-based learning environment for middle school microbiology. Using data from a classroom study involving 153 middle school students, we compare the effectiveness of several computational representations of students’ natural language responses to reflection prompts—GloVe, ELMo, tf-idf, unigrams—across several machine learning-based regression techniques (i.e., random forest, support vector machine, multi-layer perceptron) to assess the depth of student reflection responses. Results demonstrate that assessment models based on ELMo deep contextualized word representations yield more accurate predictions of students’ written reflection depth than competing techniques. These findings point toward the potential of leveraging automated assessment of student reflection to inform real-time adaptive support for inquiry-based learning in game-based learning environments.

Collaborative game-based learning environments integrate game-based learning and collaborative learning. These environments present students with a shared objective and provide them with a means to communicate, which allows them to share information, ask questions, construct explanations, and work together toward their shared goal. A key challenge in collaborative learning is that students may engage in unproductive discourse, which may affect learning activities and outcomes. Collaborative game-based learning environments that can detect this off-task behavior in real-time have the potential to enhance collaboration between students by redirecting the conversation back to more productive topics. This paper investigates the use of dialogue analysis to classify student conversational utterances as either off-task or on-task. Using classroom data collected from 13 groups of four students, we trained off-task dialogue models for text messages from a group chat feature integrated into Crystal Island: EcoJourneys, a collaborative game-based learning environment for middle school ecosystem science. We evaluate the effectiveness of the off-task dialogue models, which use different word embeddings (i.e., word2vec, ELMo, and BERT), as well as predictive off-task dialogue models that capture varying amounts of contextual information from the chat log. Results indicate that predictive off-task dialogue models that incorporate a window of recent context and represent the sequential nature of the chat messages achieve higher predictive performance compared to models that do not leverage this information. These findings suggest that off-task dialogue models for collaborative game-based learning environments can reliably recognize and predict students’ off-task behavior, which introduces the opportunity to adaptively scaffold collaborative dialogue.

Engagement plays a critical role in visitor learning in museums. Devising computational models of visitor engagement shows significant promise for enabling adaptive support to enhance visitors’ learning experiences and for providing analytic tools for museum educators. A salient feature of science museums is their capacity to attract diverse visitor populations that range broadly in age, interest, prior knowledge, and socio-cultural background, which can significantly affect how visitors interact with museum exhibits. In this paper, we introduce a Bayesian hierarchical modeling framework for predicting learner engagement with Future Worlds, a tabletop science exhibit for environmental sustainability. We utilize multi-channel data (e.g., eye tracking, facial expression, posture, interaction logs) captured from visitor interactions with a fully-instrumented version of Future Worlds to model visitor dwell time with the exhibit in a science museum. We demonstrate that the proposed Bayesian hierarchical modeling approach outperforms competitive baseline techniques. These findings point toward significant opportunities for enriching our understanding of visitor engagement in science museums with multimodal learning analytics.

In this case report, the authors describe a patient who underwent an endovascular abdominal aortic aneurysm repair complicated by more than a 2-year delay in healing of the left inguinal fold access site. Providers initially suspected a chronic infection or foreign body reaction, but eventually the patient was diagnosed with superficial granulomatous pyoderma. Once the correct etiology was determined and appropriate treatment begun, the access site healed within 3 weeks.

This paper presents Geomancer, an open-source framework for geospatial feature engineering. It simplifies the acquisition of geospatial attributes for downstream, large-scale machine learning tasks. Geomancer leverages any geospatial dataset stored in a data warehouse, users need only to define the features (Spells) they want to create, and cast them on any spatial dataset. In addition, these features can be exported into a JSON file (SpellBook) for sharing and reproducibility. Geomancer has been useful to some of our production use-cases such as property value estimation, area valuation, and more. It is available on Github, and can be installed from PyPI.

LUX-Head & Neck 2 investigators., [Importance] Locoregionally advanced head and neck squamous cell cancer (HNSCC) is treated curatively; however, risk of recurrence remains high among some patients. The ERBB family blocker afatinib has shown efficacy in recurrent or metastatic HNSCC., [Objective] To assess whether afatinib therapy after definitive chemoradiotherapy (CRT) improves disease-free survival (DFS) in patients with HNSCC., [Design, Setting, and Participants] This multicenter, phase 3, double-blind randomized clinical trial (LUX-Head & Neck 2) studied 617 patients from November 2, 2011, to July 4, 2016. Patients who had complete response after CRT, comprising radiotherapy with cisplatin or carboplatin, with or without resection of residual disease, for locoregionally advanced high- or intermediate-risk HNSCC of the oral cavity, hypopharynx, larynx, or oropharynx were included in the study. Data analysis was of the intention-to-treat population., [Interventions] Patients were randomized (2:1) to treatment with afatinib (40 mg/d) or placebo, stratified by nodal status (N0-2a or N2b-3) and Eastern Cooperative Oncology Group performance status (0 or 1). Treatment continued for 18 months or until disease recurrence, unacceptable adverse events, or patient withdrawal., [Main Outcomes and Measures] The primary end point was DFS, defined as time from the date of randomization to the date of tumor recurrence or secondary primary tumor or death from any cause. Secondary end points were DFS at 2 years, overall survival (defined as time from the date of randomization to death), and health-related quality of life., [Results] A total of 617 patients were studied (mean [SD] age, 58 [8.4] years; 528 male [85.6%]). Recruitment was stopped after a preplanned interim futility analysis on July 4, 2016, on recommendation from an independent data monitoring committee. Treatment was discontinued. Median DFS was 43.4 months (95% CI, 37.4 months to not estimable) in the afatinib group and not estimable (95% CI, 40.1 months to not estimable) in the placebo group (hazard ratio, 1.13; 95% CI, 0.81-1.57; stratified log-rank test P = .48). The most common grade 3 and 4 drug-related adverse effects were acneiform rash (61 [14.8%] of 411 patients in the afatinib group vs 1 [0.5%] of 206 patients in the placebo group), stomatitis (55 [13.4%] in the afatinib group vs 1 [0.5%] in the placebo group), and diarrhea (32 [7.8%] in the afatinib group vs 1 [0.5%] in the placebo group)., [Conclusions and Relevance] This study’s findings indicate that treatment with afatinib after CRT did not improve DFS and was associated with more adverse events than placebo in patients with primary, unresected, clinically high- to intermediate-risk HNSCC. The use of adjuvant afatinib after CRT is not recommended., [Trial Registration] ClinicalTrials.gov identifier: NCT01345669., The study was funded by Boehringer Ingelheim.

A high percentage of cytologic specimens obtained from pulmonary nodules, masses, and cystic lesions are malignant. Indeed the majority of cytologic studies of material obtained from the lower respiratory tract are performed for the investigation as to whether or not a pulmonary lesion is malignant. The lungs are the site of a large number of primary and metastatic malignancies as well as a smaller number of benign neoplasms and localized nonneoplastic conditions. Cytologic recognition of an imaging finding as definitively malignant is extraordinarily important for development of a therapeutic plan. Given today’s therapeutic requirements, a number of specific questions must be answered including (1) whether a lesion is malignant or benign, (2) whether the lesion represents a primary carcinoma or some other lesion, (3) whether the carcinoma is small cell or non-small cell in type, and (4) whether a non-small cell carcinoma is an adenocarcinoma or a squamous cell carcinoma. A number of molecular and immunohistochemical studies are often required for the proper characterization of primary pulmonary carcinomas. The malignancy risk for the malignant category is approximately 90%.

Predicting protein functions is a fundamental task with applications in medicine and healthcare. However, the accelerating pace of protein-discovery renders slow and expensive biochemical techniques unsustainable. Machine learning is suitable for such data-intensive task, but the presence of noise in protein datasets adds another level of difficulty. Hence, we propose a deep learning system based on a stacked denoising autoencoder that extracts robust features to improve predictive performance. We then feed the resulting features to a multilabel support-vector machine for classification. We evaluated on two protein benchmarks, and experimental results show that our system consistently produced the best performance against techniques that do not have a denoising or feature learning capability. This research demonstrates that learning robust representations from raw data can benefit the process of predicting protein functions.

Human immunodeficiency virus type 1 (HIV-1) is the result of cross-species transmission of simian immunodeficiency virus from chimpanzees (SIVcpz). SIVcpz is a chimeric virus which shares common ancestors with viruses infecting red-capped mangabeys and a subset of guenon species. The epidemiology of SIV infection in hominoids is characterized by low prevalences and an uneven geographic distribution. Surveys in Cameroon indicated that two closely related members of the guenon species subset, mustached guenons and greater spot-nosed guenons, infected with SIVmus and SIVgsn, respectively, also have low rates of SIV infections in their populations. Compared to that for other monkeys, including red-capped mangabeys and closely related guenon species, such an epidemiology is unusual. By intensifying sampling of geographically distinct populations of mustached and greater spot-nosed guenons in Gabon and including large sample sets of mona guenons from Cameroon, we add strong support to the hypothesis that the paucity of SIV infections in wild populations is a general feature of this monophyletic group of viruses. Furthermore, comparative phylogenetic analysis reveals that this phenotype is a feature of this group of viruses infecting phylogenetically disparate hosts, suggesting that this epidemiological phenotype results from infection with these HIV-1-related viruses rather than from a common host factor. Thus, these HIV-1-related viruses, i.e., SIVcpz and the guenon viruses which share an ancestor with part of the SIVcpz genome, have an epidemiology distinct from that found for SIVs in other African primate species. IMPORTANCE Stable virus-host relationships are established over multiple generations. The prevalence of viral infections in any given host is determined by various factors. Stable virus-host relationships of viruses that are able to cause persistent infections and exist with high incidences of infection are generally characterized by a lack of morbidity prior to host reproduction. Such is the case for cytomegalovirus (CMV) and Epstein-Barr virus (EBV) infections of humans. SIV infections of most African primate species also satisfy these criteria, with these infections found at a high prevalence and with rare cases of clinical disease. In contrast, SIVcpz, the ancestor of HIV-1, has a different epidemiology, and it has been reported that infected animals suffer from an AIDS-like disease in the wild. Here we conclusively demonstrate that viruses which are closely related to SIVcpz and infect a subset of guenon monkeys show an epidemiology resembling that of SIVcpz.

Power management systems employ appliance recognition such that the burden of manually configuring the system for each appliance is lifted from the user. This research then aims to develop an appliance recognition functionality through current readings gathered from a data acquisition (DAQ) device consisting of Hall Effect current sensors, and through a machine learning classification algorithm called k-nearest neighbors. Ten appliances were tested, comprising of 6,500 samples of test data in the four outlets tested. The average accuracy for the trials is 92.73%. In addition, the appliance recognition functionality was embedded to a cloud-based power management system following an Internet of Things (IoT) architecture. In the end, the developed system can gather data from plugged appliances, perform recognition, and carry out various power management functionalities such as monitoring and appliance-level smart-recommendations.

The neoplastic benign interpretation category indicates that the cytological specimen is sufficiently cellular and representative, with or without the context of clinical, imaging, and ancillary studies, to be diagnostic of a benign neoplasm. The most common entity is this category is serous cystadenoma; other rare entities include neuroendocrine microadenoma (< 0.5 cm) and lymphangioma. Surgical resection is optional and largely dependent on the size of the lesion and patient symptoms.

International audience

Specimens are assigned to the negative category when the specimen does not contain any material indicative of malignancy or any degree of cellular atypia. Specimens containing only normal pancreatic tissue are assigned to this category only when there is no discrete mass or cystic lesion. Correlation with imaging findings is required for placing a specimen in this category. Specific nonneoplastic entities placed in this category include acute pancreatitis, chronic pancreatitis, autoimmune pancreatitis, pseudocyst, lymphoepithelial cyst, and ectopic spleen.

The categories “Atypical” and “Suspicious for Malignancy” are indeterminate categories containing a spectrum of cellular changes ranging from minor degrees of nuclear enlargement, hyperchromasia, and anisonucleosis (atypical) to cellular specimens with features falling just short of those necessary for the diagnosis of a high grade malignancy (suspicious). The “Atypical” category is used for atypia in endoscopic ultrasound fine needle aspiration specimens of solid lesions, nonspecific cyst fluids, and duct brushing specimens. Specimens with cytologic features falling short of those diagnostic of pancreatic neuroendocrine tumor and solid and papillary neoplasm (SPN) are placed in the “Atypical” category. Brushing specimens of strictures containing features of low-grade dysplasia (BilIN I to II) are placed in the “Atypical” category while high-grade dysplasia, e. g. biliary intraepithelial neoplasia III, is placed in the “Suspicious for Malignancy” category.

The diagnostic category “Suspicious for Malignancy” is utilized for endoscopic ultrasound fine-needle aspiration specimens obtained from solid lesions and brushing specimens of strictures containing cytologic features falling just short of those necessary for a definitive diagnosis of a high-grade malignancy (pancreatic ductal carcinoma, acinar carcinoma or metastasis). “Suspicious for Malignancy” is the appropriate category for brushing specimens with features of high-grade dysplasia (biliary intraepithelial neoplasia III). Specimens categorized as “Suspicious for Malignancy” have a high risk for malignancy. Fluorescence in situ hybridization (FISH) ancillary testing or next-generation sequencing may be appropriate to further define specimens placed in this category.

Over the past decade, there has been growing interest in real-time assessment of student engagement and motivation during interactions with educational software. Detecting symptoms of disengagement, such as offtask behavior, has shown considerable promise for understanding students' motivational characteristics during learning. In this paper, we investigate the affective role of off-task behavior by analyzing data from student interactions with CRYSTAL ISLAND, a narrative-centered learning environment for middle school microbiology. We observe that off-task behavior is associated with reduced student learning, but preliminary analyses of students' affective transitions suggest that off-task behavior may also serve a productive role for some students coping with negative affective states such as frustration. Empirical findings imply that some students may use off-task behavior as a strategy for self-regulating negative emotional states during learning. Based on these observations, we introduce a supervised machine learning procedure for detecting whether students' off-task behaviors are cases of emotion self-regulation. The method proceeds in three stages. During the first stage, a dynamic Bayesian network (DBN) is trained to model the valence of students' emotion self-reports using collected data from interactions with the learning environment. In the second stage, a novel simulation process uses the DBN to generate alternate futures by modeling students' affective trajectories as if they had engaged in fewer off-task behaviors than they did during their actual learning interactions. The alternate futures are compared to students' actual traces to produce labels denoting whether students' off-task behaviors are cases of emotion self-regulation. In the final stage, the generated emotion self-regulation labels are predicted using off-the-shelf classifiers and features that can be computed in run-time settings. Results suggest that this approach shows promise for identifying cases of off-task behavior that are emotion self-regulation. Analyses of the first two phases suggest that trained DBN models are capable of accurately modeling relationships between students' off-task behaviors and self-reported emotional valence in CRYSTAL ISLAND. Additionally, the proposed simulation process produces emotion self-regulation labels with high levels of reliability. Preliminary analyses indicate that support vector machines, bagged trees, and random forests show promise for predicting the generated emotion self-regulation labels, but room for improvement remains. The findings underscore the methodological potential of considering alternate futures when modeling students' emotion self-regulation processes in narrative-centered learning environments., The file is in PDF format. If your computer does not recognize it, simply download the file and then open it with your browser.

In this paper the WASABI Affect Simulation Architecture is introduced, in which a virtual human’s cognitive reasoning capabilities are combined with simulated embodiment to achieve the simulation of primary and secondary emotions. In modeling primary emotions we follow the idea of “Core Affect” in combination with a continuous progression of bodily feeling in three-dimensional emotion space (PAD space), that is only subsequently categorized into discrete emotions. In humans, primary emotions are understood as onto-genetically earlier emotions, which directly influence facial expressions. Secondary emotions, in contrast, afford the ability to reason about current events in the light of experiences and expectations. By technically representing aspects of their connotative meaning in PAD space, we not only assure their mood-congruent elicitation, but also combine them with facial expressions, that are concurrently driven by the primary emotions. An empirical study showed that human players in the Skip-Bo scenario judge our virtual human MAX significantly older when secondary emotions are simulated in addition to primary ones.

In order to enable collaboration and exchange of modules for generating multimodal communicative behaviours of robots and virtual agents, the SAIBA initiative envisions the definition of two representa- tion languages. One of these is the Functional Markup Language (FML). This language specifies the communicative intent behind an agent's be- haviour. Currently, several research groups have contributed to the dis- cussion on the definition of FML. The discussion reveals agreement on many points but it also points out important issues that need to be dealt with. This paper summarises the current state of affairs in thinking about FML.

In the framework of our Reactive Virtual Trainer (RVT) project, we are developing an Intelligent Virtual Agent (IVA) capable to act similarly to a real trainer. Besides presenting the physical exercises to be performed, she keeps an eye on the user. She provides feedback whenever appropriate, to in-troduce and structure the exercises, to make sure that the exercises are per-formed correctly, and also to motivate the user. In this paper we talk about the corpora we collected, serving a basis to model repetitive exercises on high level. Then we discuss in detail how the actual performance of the user is com-pared to what he should be doing, what strategy is used to provide feedback, and how it is generated. We provide preliminary feedback from users and out-line further work.

During face-to-face interactions, listeners use backchannel feedback such as head nods as a signal to the speaker that the communication is working and that they should continue speaking. Predicting these backchannel opportunities is an important milestone for building engaging and natural virtual humans. In this paper we show how sequential probabilistic models (e.g., Hidden Markov Model (HMM) or Conditional Random Fields (CRF)) can automatically learn from a database of human-to-human interactions to predict listener backchannels using the speaker multimodal output features (e.g., prosody, spoken words and eye gaze). The main challenges addressed in this paper are automatic selection of the relevant features and optimal feature representation for probabilistic models. For prediction of visual backchannel cues (i.e., head nods), our prediction model shows a statistically significant improvement over a previously published approach based on hand-crafted rules.

The GIVE Challenge is a new Internet based evaluation effort for natural language generation systems. In this paper, we motivate and describe the software infrastructure that we developed to support this challenge.

Conventionally poorly functioning hydronephrotic kidneys have been removed if they are symptomatic. In our unit, patients are offered renal artery embolization as an alternative treatment option. Patients and Methods: Fifteen patients (11 male, 4 female) with a mean age of 32.9 yr (20–51 yrs) have undergone renal artery embolization for symptomatic hydronephrosis with poor function. Mean follow-up was 64.13 weeks (range 14–200). All patients had loin pain and hydronephrosis. Twelve patients had primary pelvi-ureteric junction obstruction (PUJO). Two patients had poorly functioning hydronephrotic kidneys secondary to chronic calculous obstruction. One patient had chronic pain in an obstructed but reasonably functioning kidney following a previous pyeloplasty for PUJO which demanded intervention. Mean split function on renography was 11% (range 0–46%). Selective renal artery embolization was carried out under antibiotic cover using a 7 Fr balloon occlusion catheter and absolute alcohol, steel coils, and polyvinyl alcohol particles.Results: Nine patients developed post-embolization syndrome of self-limiting pain and pyrexia with no evidence of sepsis. One patient required readmission with this condition. One patient developed a hematoma at the puncture site. Mean hospital stay was 2.3 days. Fourteen patients are happy with the result and are completely pain free. One patient has minor discomfort but is delighted with the result. Nine patients have had follow-up ultrasound confirming resolution of the hydronephrosis. Conclusion: Renal artery embolization is an effective, safe, well-tolerated minimally invasive treatment option in end-stage hydronephrosis and we routinely offer it as an alternative to nephrectomy.

Recent years have witnessed significant progress in intelligent user interfaces. Emerging from the intersection of AI and human-computer interaction, research on intelligent user interfaces is experiencing a renaissance, both in the overall level of activity and in raw research achievements. Research on intelligent user interfaces exploits developments in a broad range of foundational AI work, ranging from knowledge representation and computational linguistics to planning and vision. Because intelligent user interfaces are designed to facilitate problem-solving activities where reasoning is shared between users and the machine, they are currently transitioning from the laboratory to applications in the workplace, home, and classroom.

'Best of ITS 98' Special Issue. The following papers are revised, extended and IJAIED-reviewed versions of the most highly-rated papers presented at ITS 98.; Lifelike animated agents for knowledge-based learning environments can provide timely, customized advice to support learnersâ problem-solving activities. By drawing on a rich repertoire of emotive behaviors to exhibit contextually appropriate facial expressions and emotive gestures, these agents could exploit the visual channel to more effectively communicate with learners. To address these issues, this article proposes the emotive-kinesthetic behavior sequencing framework for dynamically sequencing lifelike pedagogical agentsâ full-body emotive expression. By exploiting a rich behavior space populated with emotive behaviors and structured by pedagogical speech act categories, a behavior sequencing engine operates in realtime to select and assemble contextually appropriate expressive behaviors. This framework has been implemented in a lifelike pedagogical agent, Cosmo, who exhibits full-body emotive behaviors in response to learners' problem-solving activities. (http://aied.inf.ed.ac.uk/members99/archive/vol_10/lester/full.html)

An edge detection function is defined on a very broad class of pictures. It departs from previous approaches to edge detection, which are based on differences in intensity or other properties between adjacent areas of a picture, in using the spatial and similarity relations among individual entities, called elements, to determine the locations and probabilities of edges. With auxiliary definitions of similarity, it may be applied to digitized pictures, treating picture points as elements, or to scenes of diversified shapes distributed in any way - although only scenes of disks are discussed in this paper. This generality permits recursive use of the edge detector in the discrimination of visual textures. An algorithm for the efficient computation of the function is described. The program runs at the University of Wisconsin in a larger system, which includes T.V. input and additional programs that cluster edges into boundaries.

Introduction - This thesis attempts to formulate the literary theories of Stendhal, not so much by analysing his novels as by searching out and synthesizing his own remarks on the principles which guided him. The author has had access to the 79 volumes of the complete edition of Le Divan of Stendhal's works. By his literary ideas S. is fundamentally of the eighteenth century; his modernity and durability are due to the fact that he recognized and applied the most fecund literary principles of that century. Chapter I - La Relativlté du Beau -The ideas of Relativity and Progress, the two great contributions to aesthetics of the 18th century, are Stendhal's basic principles. There is no absolute form of beauty since taste varies according to race, climate and the level of progress of the age. Chapter II - Le Romanticisme - Since there is no absolute beauty each generation has the right to create the form of the beau which pleases it. In Italy Stendhal participates in the Romantic struggle and adopts the word Romanticisme to express his idea of modern beauty. He attempts to destroy the authority of Racine and discourage his imitators. Chapter III - La Tragédie romantique - Discusses Stendhal's ideas on the modern tragedy. The chapter is a summary of the essential doctrine of Racine et Shakespeare with the addition of all relevant material from his other writings. He helped the French Romantic poets to achieve their victory but when it took the form of the drame in verse he abandoned the theatre and concerned himself with the novel. Chapter IV - Le Roman - Stendhal's evolution from a writer of comedies to a novelist. He abandons the comedy on perceiving that the Revolution has destroyed the formerly uniform good taste of the public. The novel as a genre is free of rules. He sees it as a mirror which faithfully reflects contemporary times, particularly those aspects of it which are new and not hitherto found in literature. The chief aim of the novelist is to reflect the psychology of his period. The clarity and simplicity of his style is enforced on him by the obscurity of this subject matter. He adopts a concise, rapid style since he perceives that the tempo of the human intelligence is steadily accelerating. Because of the bad taste of his generation he writes only for the Happy Few who are intelligent, cultivated and sensitive, hoping thus to go down to posterity and be read "in 1935". To this end he creates a style which will not date or age. Deliberately and systematically he constructed his own immortality. Chapter V - Le Lecteur comme Créateur - From Fénelon he derives his manner of reflecting the "petits faits vrais" like a mirror, without suggesting to the reader the conclusion or emotion to be drawn from the facts. His style seems dry and without sentiment since he writes for the imaginative who are able to create their own emotion from these facts. By suggesting, rather than describing, he is a precursor of the Symbolists. However Diderot, in the eighteenth century, had already recognized the advantage of allowing the spectator to participate in the creation of the beau.

BACKGROUND:Dengue virus (DENV) infects hundreds of thousands of people annually in Indonesia. However, DENV sequence data from the country are limited, as samples from outbreaks must be shipped across long-distances to suitably equipped laboratories to be sequenced. This approach is time-consuming, expensive, and frequently results in failure due to low viral load or degradation of the RNA genome. METHODS:We evaluated a method designed to address this challenge, using the 'Primal Scheme' multiplex PCR tiling approach to rapidly generate short, overlapping amplicons covering the complete DENV coding-region, and sequencing the amplicons on the portable Nanopore MinION device. The resulting sequence data was assessed in terms of genome coverage, consensus sequence accuracy and by phylogenetic analysis. RESULTS:The multiplex approach proved capable of producing near complete coding-region coverage from all samples tested ([Formula: see text] = 99.96%, n = 18), 61% of which could not be fully amplified using the current, long-amplicon PCR, approach. Nanopore-generated consensus sequences were found to be between 99.17-99.92% identical to those produced by high-coverage Illumina sequencing. Consensus accuracy could be improved by masking regions below 20X coverage depth (99.69-99.92%). However, coding-region coverage was reduced at this depth ([Formula: see text] = 93.48%). Nanopore and Illumina consensus sequences generated from the same samples formed monophyletic clades on phylogenetic analysis, and Indonesian consensus sequences accurately clustered by geographical origin. CONCLUSION:The multiplex, short-amplicon approach proved superior for amplifying DENV genomes from clinical samples, particularly when the virus was present at low concentrations. The accuracy of Nanopore-generated consensus sequences from these amplicons was sufficient for identifying the geographic origin of the samples, demonstrating that the approach can be a useful tool for identifying and monitoring DENV clades circulating in low-resource settings across Indonesia. However, the inaccuracies in Nanopore-generated consensus sequences mean that the approach may not be appropriate for higher resolution transmission studies, particularly when more accurate sequencing technologies are available.

Background: Dengue virus (DENV) infects hundreds of thousands of people annually in Indonesia. However, DENV sequence data from the country are limited, as samples from outbreaks must be shipped across long-distances to suitably equipped laboratories to be sequenced. This approach is time-consuming, expensive, and frequently results in failure due to low viral load or degradation of the RNA genome. Methods: We evaluated a method designed to address this challenge, using the ‘Primal Scheme’ multiplex PCR tiling approach to rapidly generate short, overlapping amplicons covering the complete DENV coding-region, and sequencing the amplicons on the portable Nanopore MinION device. The resulting sequence data was assessed in terms of genome coverage, consensus sequence accuracy and by phylogenetic analysis. Results: The multiplex approach proved capable of producing near complete coding-region coverage from all samples tested (x¯ = 99.96%, n = 18), 61% of which could not be fully amplified using the current, long-amplicon PCR, approach. Nanopore-generated consensus sequences were found to be between 99.17–99.92% identical to those produced by high-coverage Illumina sequencing. Consensus accuracy could be improved by masking regions below 20X coverage depth (99.69–99.92%). However, coding-region coverage was reduced at this depth (x¯ = 93.48%). Nanopore and Illumina consensus sequences generated from the same samples formed monophyletic clades on phylogenetic analysis, and Indonesian consensus sequences accurately clustered by geographical origin. Conclusion: The multiplex, short-amplicon approach proved superior for amplifying DENV genomes from clinical samples, particularly when the virus was present at low concentrations. The accuracy of Nanopore-generated consensus sequences from these amplicons was sufficient for identifying the geographic origin of the samples, demonstrating that the approach can be a useful tool for identifying and monitoring DENV clades circulating in low-resource settings across Indonesia. However, the inaccuracies in Nanopore-generated consensus sequences mean that the approach may not be appropriate for higher resolution transmission studies, particularly when more accurate sequencing technologies are available.

BACKGROUND: Dengue virus (DENV) infects hundreds of thousands of people annually in Indonesia. However, DENV sequence data from the country are limited, as samples from outbreaks must be shipped across long-distances to suitably equipped laboratories to be sequenced. This approach is time-consuming, expensive, and frequently results in failure due to low viral load or degradation of the RNA genome. METHODS: We evaluated a method designed to address this challenge, using the 'Primal Scheme' multiplex PCR tiling approach to rapidly generate short, overlapping amplicons covering the complete DENV coding-region, and sequencing the amplicons on the portable Nanopore MinION device. The resulting sequence data was assessed in terms of genome coverage, consensus sequence accuracy and by phylogenetic analysis. RESULTS: The multiplex approach proved capable of producing near complete coding-region coverage from all samples tested ([Formula: see text] = 99.96%, n = 18), 61% of which could not be fully amplified using the current, long-amplicon PCR, approach. Nanopore-generated consensus sequences were found to be between 99.17-99.92% identical to those produced by high-coverage Illumina sequencing. Consensus accuracy could be improved by masking regions below 20X coverage depth (99.69-99.92%). However, coding-region coverage was reduced at this depth ([Formula: see text] = 93.48%). Nanopore and Illumina consensus sequences generated from the same samples formed monophyletic clades on phylogenetic analysis, and Indonesian consensus sequences accurately clustered by geographical origin. CONCLUSION: The multiplex, short-amplicon approach proved superior for amplifying DENV genomes from clinical samples, particularly when the virus was present at low concentrations. The accuracy of Nanopore-generated consensus sequences from these amplicons was sufficient for identifying the geographic origin of the samples, demonstrating that the approach can be a useful tool for identifying and monitoring DENV clades circulating in low-resource settings across Indonesia. However, the inaccuracies in Nanopore-generated consensus sequences mean that the approach may not be appropriate for higher resolution transmission studies, particularly when more accurate sequencing technologies are available.

In this work we concentrate on finding correlation between speech signal and occurrence of facial gestures with the goal of creating believable virtual humans. We propose a method to implement facial gestures as a valuable part of human behavior and communication. Information needed for the generation of the facial gestures is extracted from speech prosody by analyzing natural speech in real time. This work is based on the previously developed HUGE architecture for statistically based facial gesturing, and extends our previous work on automatic real time lip sync.

The HomeWork project is building an exemplar system to provide individualised experiences for individual and groups of children aged 6-7 years, their parents, teachers and classmates at school. It employs an existing set of broadcast video media and associated resources that tackle both numeracy and literacy at Key Stage 1. The system employs a learner model and a pedagogical model to identify what resource is best used with an individual child or group of children collaboratively at a particular learning point and at a particular location. The Coherence Compiler is that component of the system which is designed to impose an overall narrative coherence on the materials that any particular child is exposed to. This paper presents a high level vision of the design of the Coherence Compiler and sets its design within the overall framework of the HomeWork project and its learner and pedagogical models.