Annotation Model Research Articles

Review of Intonation in Romance, by Sónia Frota and Pilar Prieto (Eds.). Oxford: Oxford University Press

The edited volume, Intonation in Romance, comprises eleven chapters: nine content chapters summarise the results of detailed prosodic analysis of intonation patterns across varieties of a particular Romance language, and are framed by an introduction and conclusion by the editors. The languages treated include those whose intonation systems have received much attention (Catalan, French, Italian, Portuguese and Spanish), alongside less-studied languages (Friulian, Occitan, Romanian and Sardinian). All chapters used the same methods of data collection and analysis: parallel data across languages was elicited with a common set of dialogue completion tasks; intonation patterns in the data are analysed on the basis of a shared implementation of the Tone and Break Indices (ToBI) model of prosodic annotation, within the Autosegmental-Metrical framework. These shared methods inform direct comparison of Romance intonation patterns and systems, within and across languages, to identify the scope and potential causes of variation, as well as avenues for future research.

Fast image captioning using LSTM

Computer vision and natural language processing have been some of the long-standing challenges in artificial intelligence. In this paper, we explore a generative automatic image annotation model, which utilizes recent advances on both fronts. Our approach makes use of a deep-convolutional neural network to detect image regions, which later will be fed to recurrent neural network that is trained to maximize the likely-hood of the target sentence description of the given image. During our experimentation we found that better accuracy and training was achieved when the image representation from our detection model is coupled with the input word embedding, we also found out most of the information from the last layer of detection model vanishes when it is fed as thought vector for our LSTM decoder. This is mainly because the information within the last fully connected layer of the YOLO model represents the class probabilities for the detected objects and their bounding box and this information is not rich enough. We trained our model on coco benchmark for 60 h on 64,000 training and 12,800-validation dataset achieving 23% accuracy. We also realized a significant training speed drop when we changed the number of hidden units in the LSTM layer from 1470 to 4096.

Verifying data secure flow in AUTOSAR models

This paper presents an approach for enhancing the design phase of AUTOSAR models when security annotations are required. The approach is based on information flow analysis and abstract interpretation. The analysis evaluates the correctness of the model by assessing if the flow of data is secure with respect to causal data dependencies within the model. To find these dependencies an exhaustive search through the model would be required. Abstract interpretation is used as a trade-off between the precision and complexity of the analysis. The approach also provides annotated models without oversizing the set of annotations.

Incorporation of PBRM1, BAP1, TP53 mutation status into the Memorial Sloan Kettering Cancer Center (MSKCC) risk model: A genomically annotated tool to improve stratification of patients (pts) with advanced renal cell carcinoma (RCC).

639 Background: The MSKCC risk model, an established prognostic tool fo r metastatic RCC, integrates clinical + laboratory data, but is ignorant to tumor genomics. Mutations in BAP1, PBRM1, TP53, cumulatively found in over 50% of pts, have prognostic value in RCC. We sought to study the use of integrating mutation status into the MSKCC model using two large clinical trial datasets. Methods: Pts had received first line sunitinib or pazopanib on the phase III COMPARZ (training set, n = 357) or the phase II RECORD3 trial (validation set, n = 130). Genes were evaluated by next generation sequencing using archival tissue. Association of mutation status and overall survival (OS) was tested by multivariate Cox regression analysis (MVA) in the training set. An annotated model was constructed combining the original clinical variables and mutation status for the 3 genes. We compared risk group assignment and concordance index (c-index) for the original vs. new model in training and validation set. Results: Mutation status for each gene: BAP1, TP53 and PBRM1 independently correlated with OS on MVA (p≤0.0035). Comparing the original (clinical only) to the annotated (clinical + genomics) model, risk categories changed in 139 pts (39%). The C-index was improved with integration of genomic information (0.595 original model - > 0.628 new model). The independent validation cohort confirmed improvement of c-index for predicting OS with integration of genomic data (c-index 0.622 original model - > 0.641 new model). Conclusions: Mutation status for BAP1, PBRM1, and TP53 has prognostic value in pts with advanced RCC. The annotated risk model alters risk status in over 1/3 of pts and improves accuracy of estimating outcomes in patients receiving first-line therapy. Clinical trial information: NCT00720941. [Table: see text]

Secure information sharing in social agent interactions using information flow analysis

When we wish to coordinate complex, cooperative tasks in open multi-agent systems, where each agent has autonomy and the agents have not been designed to work together, we need a way for the agents themselves to determine the social norms that govern collective behaviour. An effective way to define social norms for agent communication is through the use of interaction models such as those expressed in the Lightweight Coordination Calculus (LCC), a compact executable specification language based on logic programming and pi-calculus. Open multi-agent systems have experienced growing popularity in the multi-agent community and gain importance as large scale distributed systems become more widespread. A major practical limitation to such systems is security, because the very openness of such systems opens the doors to adversaries to exploit vulnerabilities introduced through acceptance of social norms.This paper addresses a key vulnerability of security of open multi-agent systems governed by formal models of social norms (as exemplified by LCC). A fundamental limitation of conventional security mechanisms (e.g. access control and encryption) is the inability to prevent information from being propagated. Focusing on information leakage in choreography systems using LCC, we suggest a framework to detect insecure information flows. A novel security-typed LCC language is proposed to prevent information leakage.Both static (design-time) and dynamic (run-time) security type checking are employed to guarantee no information leakage can occur in annotated agent interaction models. The proposed security type system is discussed and then formally evaluated by proving its properties. Two disadvantages of the pure dynamic analysis are its late detection and its inability to detect implicit information flows. We overcome these issues by performing static analysis. The proposed security type system supports non-interference, i.e. high-security input to the program never affect low-security output. However, it disregards information leaks due to the termination of the program.

Improved ontology-based similarity calculations using a study-wise annotation model.

A typical use case of ontologies is the calculation of similarity scores between items that are annotated with classes of the ontology. For example, in differential diagnostics and disease gene prioritization, the human phenotype ontology (HPO) is often used to compare a query phenotype profile against gold-standard phenotype profiles of diseases or genes. The latter have long been constructed as flat lists of ontology classes, which, as we show in this work, can be improved by exploiting existing structure and information in annotation datasets or full text disease descriptions. We derive a study-wise annotation model of diseases and genes and show that this can improve the performance of semantic similarity measures. Inferred weights of individual annotations are one reason for this improvement, but more importantly using the study-wise structure further boosts the results of the algorithms according to precision-recall analyses. We test the study-wise annotation model for diseases annotated with classes from the HPO and for genes annotated with gene ontology (GO) classes. We incorporate this annotation model into similarity algorithms and show how this leads to improved performance. This work adds weight to the need for enhancing simple list-based representations of disease or gene annotations. We show how study-wise annotations can be automatically derived from full text summaries of disease descriptions and from the annotation data provided by the GO Consortium and how semantic similarity measure can utilize this extended annotation model. Database URL: https://phenomics.github.io/

Integration of Image Feature and Word Relevance: Toward Automatic Image Annotation in Cyber-Physical-Social Systems

Image annotation is challenging due to the explosive increase of image data in cyber-physical-social systems. Because of the semantic gap between images and corresponding labels, it has attracted extensive attentions in recent years. However, most existing research neglects the imbalanced distribution of different classes and the internal relevance of image labels. Besides, the weak image labeling affects the annotation performance to some extent. To address these issues, we propose a learning model for image annotation through integrating deep features and label relevance of images. Specifically, we first employ a convolutional neural-network approach to extract the deep features of images and utilize the synthetic minority oversampling technique to deal with the problem of class imbalance. Furthermore, we exploit the correlations, including symbiotic and semantic relationships of labels, to compute the relevance of label sets. Then, we incorporate this relevance into one classifier to reconstruct the complete label sets, and learn the mapping from image features to the reconstructed label sets by the other classifier. In addition, a joint convex loss function is proposed, which combines the two classifiers via co-regularization and compels them to be consistent. We evaluate the proposed method on two benchmark data sets. The experimental results demonstrate that our method outperforms several state-of-the-art solutions.

Leveraging Unlabeled Data for Emotion Recognition With Enhanced Collaborative Semi-Supervised Learning

One of the major obstacles that has to be faced when applying automatic emotion recognition to realistic human–machine interaction systems is the scarcity of labeled data for training a robust model. Motivated by this concern, this paper seeks to utmost exploit unlabeled data that are pervasively available in the real-world and easy to be collected, by means of novel semi-supervised learning (SSL) approaches. Conventional SSL methods such as self-training, suffer from their inherent drawback of error accumulation, i.e., the samples that are misclassified by the system are continuously employed to train the model in the following learning iterations. To address this major issue, we first propose an enhanced learning strategy, by which we re-evaluate the previously automatically labeled samples in each learning iteration, in order to update the training set by correcting the mislabeled samples. We further exploit multiple modalities and models in the SSL system, by using collaborative SSL, where all modalities and models are considered simultaneously; samples are selected by means of minimizing the joint entropy. This strategy is supposed to not only improve the performance of the model for data annotation and consequently enhance the trustability of the automatically labeled data, but also to elevate the diversity of selected data. To evaluate the effectiveness of the proposed approaches, we performed extensive experiments on the remote collaborative and affective database, which includes multimodal recordings of spontaneous affective interactions of dyads. The empirical results show that the proposed approaches significantly outperform recently well-established SSL methods.

A Technique for Semantic Annotation and Retrieval of E-Learning Objects

The primary objective of my research is to design and develop semantic annotation and retrieval model for e-learning document. In training phase, the documents from different domains are taken and the informative words from each document are obtained based on balanced mutual information and frequency of contents in each document. We then use the informative words to identify the superordinates and the objects. The superordinates, the informative words and the objects from each document will give the relation and properties of each document. The relation and properties of each document are then used to cluster the documents. In the testing phase, we give a query or a document as input to the system to retrieve the relevant documents. If a document is given as input, the relation and properties of that document are first identified and it is used to retrieve the relevant documents.

Discourse Markers in Speech: Distinctive Features and Corpus Annotation

It is generally acknowledged that discourse markers are used differently in speech and writing, yet many general descriptions and most annotation frameworks are written-based, thus partially unfit to be applied in spoken corpora. This paper identifies the major distinctive features of discourse markers in spoken language, which can be associated with problems related to their scope and structure, their meaning and their tendency to co-occur. The description is based on authentic examples and is followed by methodological recommendations on how to deal with these phenomena in more exhaustive, speech-friendly annotation models.

Quantitative RNA-seq meta-analysis of alternative exon usage in C. elegans.

Almost 20 years after the completion of the C. elegans genome sequence, gene structure annotation is still an ongoing process with new evidence for gene variants still being regularly uncovered by additional in-depth transcriptome studies. While alternative splice forms can allow a single gene to encode several functional isoforms, the question of how much spurious splicing is tolerated is still heavily debated. Here we gathered a compendium of 1682 publicly available C. elegans RNA-seq data sets to increase the dynamic range of detection of RNA isoforms, and obtained robust measurements of the relative abundance of each splicing event. While most of the splicing reads come from reproducibly detected splicing events, a large fraction of purported junctions is only supported by a very low number of reads. We devised an automated curation method that takes into account the expression level of each gene to discriminate robust splicing events from potential biological noise. We found that rarely used splice sites disproportionately come from highly expressed genes and are significantly less conserved in other nematode genomes than splice sites with a higher usage frequency. Our increased detection power confirmed trans-splicing for at least 84% of C. elegans protein coding genes. The genes for which trans-splicing was not observed are overwhelmingly low expression genes, suggesting that the mechanism is pervasive but not fully captured by organism-wide RNA-seq. We generated annotated gene models including quantitative exon usage information for the entire C. elegans genome. This allows users to visualize at a glance the relative expression of each isoform for their gene of interest.

Automatic Image Annotation Model Using LSTM Approach

Automatic Image Annotation Model Using LSTM Approach

Multi-level semantic annotation and unified data integration using semantic web ontology in big data processing

The potential applications of big data need semantic annotation and unified integration of heterogeneous data. This paper proposes MOUNT a multi-level annotation and integration framework that significantly process the heterogeneous dataset by exploiting the semantic knowledge to improve the query processing in the large scale infrastructure. The multi-level annotation proposes the coarse-grained and fine-grained annotation models. The coarse-grained annotation employs Yago and SEeds SEarch to categorize the domain information on the big data and fine-grained annotation enables semantic enrichment. Moreover, the MOUNT approach integrates the structured and unstructured data to form the global resource description framework ontology. Moreover, it facilitates the query processing by translating the natural language user query into structured triples. The experimental results prove that the MOUNT approach yields a better performance in terms of result accuracy by 94%.

Automatic evaluation of reading aloud performance in children

Evaluating children's reading aloud proficiency is typically a task done by teachers on an individual basis, where reading time and wrong words are marked manually. A computational tool that assists with recording reading tasks, automatically analyzing them and outputting performance related metrics could be a significant help to teachers. Working towards that goal, this work presents an approach to automatically predict the overall reading aloud ability of primary school children by employing automatic speech processing methods. Reading tasks were designed focused on sentences and pseudowords, so as to obtain complementary information from the two distinct assignments. A dataset was collected with recordings of 284 children aged 6–10 years reading in native European Portuguese. The most common disfluencies identified include intra-word pauses, phonetic extensions, false starts, repetitions, and mispronunciations. To automatically detect reading disfluencies, we first target extra events by employing task-specific lattices for decoding that allow syllable-based false starts as well as repetitions of words and sequences of words. Then, mispronunciations are detected based on the log likelihood ratio between the recognized and target words. The opinions of primary school teachers were gathered as ground truth of overall reading aloud performance, who provided 0–5 scores closely related to the expected performance at the end of each grade. To predict these scores, various features were extracted by automatic annotation and regression models were trained. Gaussian process regression proved to be the most successful approach. Feature selection from both sentence and pseudoword tasks give the closest predictions, with a correlation of 0.944 compared to the teachers' grading. Compared to the use of manual annotation, where the best models obtained give a correlation of 0.949, there was a relative decrease of only 0.5% for using automatic annotations to extract features. The error rate of predicted scores relative to ground truth also proved to be smaller than the deviation of evaluators’ opinion per child.

Large scale automatic image annotation based on convolutional neural network

Automatic image annotation is one of the most important challenges in computer vision, which is critical to many real-world researches and applications. In this paper, we focus on the issue of large scale image annotation with deep learning. Firstly, considering the existing image data, especially the network images, most of the labels of themselves are inaccurate or imprecise. We propose a Multitask Voting (MV) method, which can improve the accuracy of original annotation to a certain extent, thereby enhancing the training effect of the model. Secondly, the MV method can also achieve the adaptive label, whereas most existing methods pre-specify the number of tags to be selected. Additionally, based on convolutional neural network, a large scale image annotation model MVAIACNN is constructed. Finally, we evaluate the performance with experiments on the MIRFlickr25K and NUS-WIDE datasets, and compare with other methods, demonstrating the effectiveness of the MVAIACNN.

A Novel Image Annotation Method based on Kernel Methods for Structured Prediction

With the development of image processing and storage technology, rapid classification and annotation of huge volumes of digital images have been attracted much attention. However, the complex and ambiguous relationship between images and concept classes poses significant challenges on building effective annotation models. Structured machine learning methods have been studied to tackle the problem of complex relationship between concept classes for prediction, which have been proved effective for image understanding tasks. We proposed a novel image annotation model based on structured machine learning, by introducing a learned kernel function in the sample space, aiming at capturing the underlying distribution of concept classes of the training data set. The model is evaluated on two benchmark data sets and the results show that the model is promising compared to current state-of-the-art methods.

An ontological model for the reality-based 3D annotation of heritage building conservation state

Abstract The conservation and restoration of historical monuments require a diagnostic analysis carried out by a multidisciplinary team. The results of the diagnosis include data produced by different techniques and protocols, which are used by conservation scientists to assess the built heritage. Nowadays, together with the aforementioned data, a great deal of heterogeneous information is also available, including descriptive and contextual information, as well as 2D/3D geometrical restitution of the studied object. However, the integration of these diverse data into a unique information model capable of fully describing the building conservation state, as well as integrating future data, is still an open issue within the Cultural Heritage community. It is of paramount importance to correlate these data and spatialize them in order to provide scientists in charge of our heritage with a practical and easy means to explore the information used during their assessment, as well as a way to record their scientific observation and share them within their community of practice. In order to resolve this issue, we developed a correlation pipeline for the integration of the semantic, spatial and morphological dimension of a built heritage. The pipeline uses an ontological model for recording and integrating multidisciplinary observations of the conservation state into structural data spatialized into a semantic-aware 3D representation. The pipeline was successfully tested on the Saint Maurice church of Caromb in the south of France, integrating into a unique spatial representation information about material and alteration phenomena, providing users with a means to correlate, and more importantly retrieve several types of information.

Abstract 3845: Establishment and characterization of patient-derived xenografts of clear cell carcinoma of the ovary

Abstract Background; The patient-derived xenograft (PDX) model is likely to reflect original human cancer biology compared to cultured cell lines, since tumor tissues are directly implanted into animals. PDX models of ovarian cancer have been developed, but were not fully characterized at the molecular level, especially for clear cell carcinoma (CCC), which is more prevalent in Japan than western countries, exhibits chemoresistant phenotype and poor survival. The development of new treatment strategies and the better understanding of biology of CCC both depend on clinically relevant animal models. In particular, molecular annotated PDX models are useful for the preclinical investigation of anticancer drugs and individualized anticancer therapy. Methods; We transplanted 19 parent (primary or metastatic) tumors directly into NOG mice. The histologic characteristics were compared between parental tumors and PDX ones. Among CCC pairs, gene expression analysis, gene mutation analysis were conducted on parental tumors and corresponding PDX tumors. Response to chemotherapeutic agents was analyzed using PDX tumors of CCC, and correlated with clinical outcome. Results; Six of nineteen (31.6%) transplanted tumors could be passaged. Gene expression profiles revealed that engrafted tumors expressed more proliferation-related genes than those of non-englafted ones. Among six PDX models established, two were found to be CCC. Two PDX tumors of CCC maintained the histologic characteristics of parental CCC tumors. Gene expression profile revealed similar results between primary CCC and PDX tumors of CCC except that PDXs lost expression of immune-related genes.. Gene mutational analysis showed similar pattern between primary CCC and PDX of CCC. Drug-sensitivity test showed resistance to carboplatin in two PDX tumors of CCC, which is consistent with their clinical outcomes. Conclusions. PDXs of CCC were established and found to retain histologic and clinical features of original tumors, and can be utilized to screen new target drugs for platinum-resistant CCC. Citation Format: Hidemichi Watari, Daisuke Endo, Kouhei Honda, Mihoko Kunitomo, Toshiyuki Nomura, Noriaki Sakuragi. Establishment and characterization of patient-derived xenografts of clear cell carcinoma of the ovary [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3845. doi:10.1158/1538-7445.AM2017-3845

Semantic interoperability for big-data in heterogeneous IoT infrastructure for healthcare

Interoperability remains a major burden to the developers of Internet of Things systems. It is due to IoT devices are extremely heterogeneous regarding basic communication protocols, data formats, and technologies. Furthermore, due to the absence of worldwide satisfactory standards, Interoperability tools remains imperfect. In this paper, we have proposed Semantic Interoperability Model for Big-data in IoT (SIMB-IoT) to deliver semantic interoperability among heterogeneous IoT devices in health care domain. This model is used to recommend medicine with side effects for different symptoms collected from heterogeneous IoT sensors. Two datasets are taken for the analysis of big-data. One dataset contains diseases with drug details and the second dataset contains medicines with side effects. Information between physician and patient are semantically annotated and transferred in a meaningful way. A Lightweight Model for Semantic annotation of Big-data using heterogeneous devices in IoT is proposed to provide annotations for big data. Resource Description Framework (RDF) is a semantic web framework that is recycled to communicate things using Triples to make it semantically significant. RDF annotated patients’ data and made it semantically interoperable. SPARQL query is used to extract records from RDF graph. Tableau, Gruff-6.2.0, and Mysql tools are used in simulation in this article.

PSearch: a framework for semantic annotated process model search

Process models play an important role in modern information systems and in service-based applications. Given the sizes of current process model repositories, there is a critical need for techniques allowing to retrieve process models fulfilling user needs. While the problem of process similarity search has been extensively studied in the literature, existing techniques do not use semantic annotations that are proposed by the semantic web services languages and models. In this paper, we propose a framework, called PSearch, for semantic annotated process models discovery. Given a user query formulated as a process model, Psearch uses a fast search technique that returns a ranked list of similar models in the repository. In a second step, a fine-grained matching method can be applied between the query and one or several matching candidates.