Search

Context-aware systems are ubiquitous computing systems capable to adapt themselves to a dynamically changing environment. Ensuring consistency in context-aware systems has proved a challenging task due to the inherent expressive power required to model dynamical systems. In the current work, we propose the use of the μ-calculus with converse, an expressive modal logic, for modeling and verifying consistency. In particular, we propose a consistency model for a context-aware communication system. Consistency is tested in terms of the satisfiability of a μ-calculus formula. We show this consistency verification method is correct and a complexity analysis is provided. We also describe an implementation with several experiments. [ABSTRACT FROM AUTHOR]

Automatic text summarization systems are nowadays of great help to extract relevant information from large corpora. Many solutions to the task have been proposed from the perspective of the optimization of a single-objective function, aiming at finding the global optimum. This is an unrealistic goal since when multiple objectives are considered a solution that optimizes one of the objectives may induce the opposite effect on the others. Recently other solutions have been proposed that involve multiple, conflicting objectives, but which eventually are aggregated into a scalar function thus resulting in a single-objective optimization problem. Furthermore, oftentimes a typical bag of words model is used and little effort has been made to include semantic relations between sentences to improve performance. In this paper a novel method for query-oriented summarization is proposed as a multiobjective optimization problem taking into account the Pareto front and based on an embedded representation of sentences. The method is evaluated with the TAC 2009 dataset. Experimental results show that the approach contributes to improve performance significantly. To the authors’ knowledge, the method is the first attempt to include embedded representations of sentences in a multiobjective optimization solution, which applies the Pareto approach to query-oriented summarization. [ABSTRACT FROM AUTHOR]

In this paper we study how the presence or absence of redundancy on multiple related texts can be used to compute sentence relevance for extractive multi-document summarization. Two types of redundancy can be found: intra-document and inter-document. By experimenting with them, different ideas can be extracted, for example: statements redundant between documents—which can be important by their popularity; statements that are not redundant—which can be important by their novelty; or statements redundant within each document—which can be important by being constantly addressed by a single author. We propose an unsupervised graph-based method that allows to generate summaries based on different strategies of redundancy. We present experiments on two DUC corpora of nine different strategies to extract information depending of how redundancy within a document and in different documents is managed. According to DUC gold standards, we found that a multi-document generic summary should contain the most redundant (popular) information between different sources while avoiding local intra-document redundancy. We implemented a mechanism to enrich sentence rankings with redundancy, improving the evaluation of summaries. [ABSTRACT FROM AUTHOR]

Errors and residuals are closely related measures of the deviation. An error is a deviation of the observed value (PEMT output) from the expected value (MT output), while the residual of the observed value is the difference between the observed and predicted value of quality. We propose an exploratory data technique representing an ideal instrument to evaluate and improve machine translation (MT) systems. The main contribution consists of a rigorous technique (a statistical method), novel to the research of MT evaluation given by residual analysis to identify differences between MT output and post-edited machine translation output regarding human translation (reference). The residual analysis of the automatic metrics can help us to discover significant differences between MT and PEMT and to identify questionable issues regarding the one reference. In this study, we show the usage of residuals in MT evaluation. Using residual analysis, we identified sentences, in which significant differences were found in the scores of automatic metrics between MT output and post-edited (PE) MT output from Slovak into English. [ABSTRACT FROM AUTHOR]

Technological advancement in the area of computing has led to production of huge amount of structured as well as unstructured data. This high dimensional data is very complex to process. Feature selection is one of the widely used techniques for preprocessing of this huge data in predictive analytics. Rough set based feature selection is an approach for handling the vagueness in data and works fine on discrete data but struggles in the continuous case as it requires discretization. This process of discretization leads to information loss. Solution for this problem was given by various authors in form of fuzzy rough set as well as intuitionistic fuzzy rough set based approaches for feature selection. Intuitionistic fuzzy set has certain benefits over the theory of traditional fuzzy sets such as its ability in a better expression of underlying information as well as its aptness to recite fragile ambiguities of the uncertainty of the objective world. The benefits offered by Intuitionistic fuzzy sets is due to the concurrent contemplation of positive, negative and hesitancy degrees for an object to belong to a set. In this paper, three novel approaches of feature reduction based on intuitionistic fuzzy rough set are presented. For this, a new intuitionistic fuzzy rough set model is established by defining a pair of lower and upper approximations. Furthermore, three new approaches of feature selection based on the degree of dependency by using score function, membership grade and cardinality of intuitionistic fuzzy numbers are introduced. Moreover, the basic results on lower and upper approximations based on rough sets are extended for intuitionistic fuzzy rough sets and analogous results are established. Moreover, a suitable algorithm is given based on our proposed approaches. Finally, the proposed algorithm is applied to an arbitrary example data set and comparison has been made with the previous fuzzy rough set based technique. The proposed algorithm is found to be better performing in terms of selected features. [ABSTRACT FROM AUTHOR]

The Computer Science department at Tecnológico Nacional de México-CENIDET, Mexico, with the collaboration of the Psychological department of the University of Castilla-La Mancha (UCLM), Spain, is on a developing process for the creation of an immersive virtual environment through virtual reality (VR) for the e-learning educational area. Such environment, works through electroencephalographic lectures (EEG) from the students, acquired by a Brain-Computer Interface (BCI), to adapt the virtual content to the profile and needs of the student on real-time basis. This system can detect the accuracy of attention and concentration levels on mental states, for the optimum development of the activities requested on an e-learning platform; if the student is not on a suitable concentration level, the system is able to induce the student to the requested mental state. The present document shows the proposal of different recommended activities that induce the mentioned mental states and the EEG response of each one. As well, the definition of the ideal learning emotional state that will be included as a part of the future works. It is important to mention that such activities are based on psychological researches that are dedicated to measure the levels of attention, concentration and other executive functions. [ABSTRACT FROM AUTHOR]

In this paper we measure the relationship between messages in the social media and the stock market prices. First, we measure the correlation and association between the amount of stock related tweets and different financial indicators such as prices, returns and transaction volume. Then, we analyze the content of the messages and test whether the tweets generated during different trends of price change (up, down or steady) can be distinguished by automatic classifiers. Our corpus consist on messages related to nine IT companies and also their daily prices and volume during trading hours for over a period of three months. Two textual representations were used, bag of words and word embeddings. The tweets were automatically tagged using two thresholds to bin the changes in price. We have found a correlation between the amount of daily messages and the volume of financial transactions. We also found negative association (more specifically, what we define as local trend association) between tweet volume and financial indicators that were not found by using only the correlation analysis. Our main contribution is that the messages generated during a positive, negative and neutral trend can be distinguished by state of the art classifiers. [ABSTRACT FROM AUTHOR]

Many multimedia objects accept an abstract representation as point sets, or point clouds, in the plane. Searching for objects in a collection is traduced to searching for matching point clouds. In this paper algorithms and data structures are given for indexing and searching point clouds. The indexes are implemented using off-the-shelf, popular, software components. Experimental tests were performed on large databases, including a synthetic database of 10 million point clouds (1000 points per cloud) and the MIR Flickr-1M database, which contains 1 million high-resolution images. The performance of the proposed indexes was evaluated according to: Average search time, construction time, recall@k, memory usage and performance under insertions and deletions. A thorough comparison was performed between the fastest method available in the literature and a repertoire of implementations. The most competitive index is three orders of magnitude faster than the state of the art, in the image database, with recall @ 1 ≥0.989 for 20% insertions and deletions. [ABSTRACT FROM AUTHOR]

This work presents the application of a convolutional neural network (CNN) used to identify emotions through taken images to students, which are learning Java language with an Intelligent Learning Environment. The CNN contains three convolutional layers, three max-pooling layers, and three neural networks with intermediate dropout connections. The CNN was trained using different emotional databases. One of them was a posed database (RaFD) and two of them were spontaneous databases created specially by us with a content focused on learning-centered emotions. The results show a comparison among three emotion recognition systems. One applying a local binary pattern approach with facial patches, another applying a geometry-based method, and the last one applying the convolutional network. The analysis presented satisfactory results; the CNN obtained a 95% accuracy for the RaFD database, an 88% accuracy for a learning-centered emotion database and a 74% accuracy for a second learning-centered emotion database. Results are compared against the classifiers support vector machine, k-nearest neighbors, and artificial neural network. [ABSTRACT FROM AUTHOR]

The advances in social robotics have extended the possibilities of their use in different applications and have also increased the sectors of users to which those applications can benefit. An attractive population of users is children. Recently, there has been a trend towards research in the design of interactive systems for children, as well as in the study of modeling the interaction between children and this type of systems. In this work, we present a study carried out with the objective of analyzing the affective response of children when interacting with a robot using speech-based communication. We collected data through an experiment using a Wizard of Oz scenario where we induced different affective reactions in the participants. Two type of data were collected and analyzed: 1) a set of evaluators manually created annotations of emotions and attitudes to determine the distribution of emotions during the experiments and evaluate how difficult is the training of automatic classifiers to discriminate different affective states from the acoustic properties of the children’s voices; 2) we used the children’s responses from a self-evaluation questionnaire about their perceptions and preferences towards the robots, modeled with different personalities, to assess whether there are relevant differences according to their different age’s range. We obtained a large children’s speech database that would be a valuable resource for the study of paralinguistic and interaction aspects. Despite the imbalance of the database, we were able to obtain good results for the classification of emotions and attitudes. We also find some relevant differences in how young and older children note the differences in the behaviors of the robots according to the modeled personality. Differences based on children’s age were also found in the preferences towards the two different robots. [ABSTRACT FROM AUTHOR]

In this paper a strategy for incorporating a flexible and reliable high-level inference module in service robots is presented. This module is a part of the robot’s cognitive architecture which coordinates perception, inference and action within the robot’s communication and interaction cycle. The present approach relies on an explicit representation of the structure of the task performed by the robot. There are three kinds of inferences that the robot can use opportunistically along the task: (1) diagnosis, (2) decision making and (3) planning; each kind can be used in specific situations of the task structure or performed in arbitrary situations with recovery purposes when there is an interaction failure. In this latter case the three kinds of inference are performed sequentially in what we call the daily-life inference cycle. The inference cycle allows the incorporation of basic emotions in the robot’s behavior. A case study incorporating these functionalities in the robot Golem-III is presented. The paper is concluded with a reflection on the opportunistic use of inference schemes to support flexible and robust behavior, including the expression of emotions, in service robots. [ABSTRACT FROM AUTHOR]

Belief merging aims to combine pieces of information from different (and possibly conflicting) sources so as to produce a single consistent belief base which retains as much information as possible. In this paper, we describe Merginator, a logic-based tool implementing three belief merging operators presented with specific characteristics that make them interesting for comparison: ΔΣ, ΔGMax and Δps. We describe these operators while solving a set of basic examples found in the literature that we translate into natural language to provide insights for Merginator users. We also propose two more complex consensus-seeking examples: an adaptation to the role play “Lost at Sea” and the subject of administrative response to air pollution in Hong Kong. Results show that all three operators provide a consensus to the scenarios, and Δps gives the most refined consensus. This demonstrates that belief merging, is a viable technique to support consensus decision-making in many domains. Merginator is open source software available at GitHub. [ABSTRACT FROM AUTHOR]

What if service robots could tell the story of the task they’ve just realized like a story? The aim of our work is to provide service robots with natural language capabilities to produce a Robot Experience Story for its human interlocutors. Robxp stories are narratives composed of the robot’s holistic perception of a recently performed task: navigation, visual perceptions and action descriptions. We contribute with a narrate dialog model specifying the composition of situations necessary for a service robot to transform its task history record into a narrative knowledge representation. We provide SitLog algorithms allowing to analyze the robot’s situation and behaviors sequence in order to generate a robxp story of the task. Both the dialogue model and the algorithms can be embedded as compositional behaviors in any other SitLog task structure. We instantiated our model into the Golem service robot framework on an experimental task. We believe Robxp stories generation could be integrated as a standard behavior for more complex service robot tasks. [ABSTRACT FROM AUTHOR]

Newborn cry features extraction for affections detection and classification has been intensively developed during the last ten to fifteen years. In this work, methods from the system identification area have been implemented in order to obtain ten Linear Predictive Coefficients (LPCs) plus a nonlinear one stated as Bilinear Intermittent Factor (BIF) per 20 ms analysis window for 40 normal and loss hearing (deaf) newborn cries each. In order to show the contribution of the nonlinear feature, a Kernel Discriminant Analysis (KDA) is performed and afterwards, two classifications tests employing Supported Vector machines (SVMs) as a standard and the Expectation Maximization (EM) algorithm over a Mixture of Experts (ME) operation, considering the BIF as an expert or parent of the LPCs, allows to obtain a 99.84% classification. [ABSTRACT FROM AUTHOR]

The bark is a very distinctive vocalization of the dog. It is very common and a mean for interaction with humans. However, the scope of their automated analysis by computational techniques, as well as the possible applications to which they can give rise have been little explored. In this study, we describe the process to develop an automatic classifier that can identify individual dogs based on their barks. We created a database with more than 6,000 barks applying positive and negative stimuli to dogs. We acoustically characterized the barking samples using a signal processing tool that extracts large sets of features. Based on these sets, we generated optimal subsets of features to feed machine learning algorithms which trained classification models. We evaluated such models and compared the classification performance of different algorithms. We analyzed the pertinence of training specific models per each breed. The classification obtained outperform the results previously reported in similar works. Our findings suggest that practical applications could be constructed on this kind of technology. [ABSTRACT FROM AUTHOR]

This paper presents a system for improving the quality of pronunciation error detection and correction for Qur’an recitation by Non-Arabic speakers. Most of the classical speech recognition systems are built using the Hidden Markov Model (HMM) with a Mixture of Gaussian Model (GMM). This paper attempts to enhance the GMM-HMM model’s performance by using Deep Neural Networks (DNNs). The major part of the work done in this paper is involved in the collection and processing of speakers’ data, and building and evaluation of baseline GMM system and the proposed DNN acoustic models for the Qur’an recitation framework. With the aim of solving some pronunciation problems and enhancing the overall performance of such a speech recognition system, we replace the mixture of Gaussians with a DNN. The DNN-HMM model outperforms the GMM-HMM model by 1.02% based on HTK’s word accuracy equation. By calculating the insertion results for both models, DNN-HMM showed progress by 2.59%. In addition, in substitution results, DNN-HMM shows progress with the confusion phonemes DAA by 15.09% and DHA by 17.28%. All experiments and results are presented and discussed in detail. [ABSTRACT FROM AUTHOR]

The study describes an experiment with different estimations of reliability. Reliability reflects the technical quality of the measurement procedure such as an automatic evaluation of Machine Translation (MT). Reliability is an indicator of accuracy, the reliability of measuring, in our case, measuring the accuracy and error rate of MT output based on automatic metrics (precision, recall, f-measure, Bleu-n, WER, PER, and CDER). The experiment showed metrics (Bleu-4 and WER) that reduce the overall reliability of the automatic evaluation of accuracy and error rate using entropy. Based on the results we can say, that the use of entropy for the estimation of reliability brings more accurate results than conventional estimations of reliability (Cronbach’s alpha and correlation). MT evaluation, based on n-grams or edit distance, using entropy could offer a new view on lexicon-based metrics in comparison to commonly used ones. [ABSTRACT FROM AUTHOR]

Course sequencing plays a major role in Intelligent Tutoring Systems because it determines the learning path of the student. However, it is difficult to define this order during early stages when there is no interaction with the student. The objective of this study is to determine the sequence of learning concepts considering an ontology and Wikipedia information. We used a text mining algorithm using Wikipedia to determine course sequencing. The knowledge base is formed by concepts and relationships in an ontology, in addition to Wikipedia articles of the same concepts. To evaluate the accuracy of the algorithm, we made a comparison against domain experts. According to the Pearson test, a correlation of 0.664 between the algorithm and experts was obtained, with a confidence level higher than 99%. The learning sequence can be defined with this method when we do not have evidence of student knowledge, to be later modified according to the interaction of the student. [ABSTRACT FROM AUTHOR]

The applications based on Web services (WS) composition are gaining momentum as an approach to build autonomous, heterogenous and distributed applications. Often a unique WS does not provide the required funcionality, therefore it is necessary to carry out a composition of WS in order to get the expected results. Optimizing the WS composition requires an efficient way to select the best Web services. Solving this problem can be a very complex task involving many criteria. In this work an approach is proposed to accomplish the WS selection from services previously registered and classified according to their quality of service (QoS) parameters. QoS parameters are characterized by an ontology and their values stored in a semantic database. The values of QoS corresponding to providers are collected automatically by an evaluation agent of QoS. In this way the service composition benefits from this proposal being easier and more efficient, taking into account only the best services and respecting the quality of service requirements established by the user. The results from the experiments performed with data involving real WS and artificially created data show that the proposed method is a feasible option to do a better selection of WS in the context of a WS composition. [ABSTRACT FROM AUTHOR]

Let K be a knowledge base (KB) and let φ be new information, both propositional formulas expressed in conjunctive form (CF). We propose a deterministic and correct algorithm for performing the belief revision of φ in K, denoted as: K ∘ φ. Our proposal satisfies subsets of AGM and KM postulates. We also present the soundness proof of our belief revision method, and the analysis of its time complexity. [ABSTRACT FROM AUTHOR]

Counting the number of words and lines that a user reads is important for many educational purposes – e.g., the reading speed is a key factor to improve learning, intelligent systems can suggest text that must be read to achieve a determined learning objective. The eye tracking technology is commonly used to analyze the user reading habits. Counting the number of read words could be hard when the readings are obtained from imprecise eye tracking data – e.g., eye tracking calibration difficulties. Approaches that find patterns from saccades and fixations usually fail to solve the problem in such conditions. This paper introduces the Cowl approach, which deals with the imprecision problem by associating the eye tracking data with points obtained from character recognition. To detect text lines truly read, the problem is stated as one of merging two hypothetical lines and it is solved by a Bayesian approach. Tests show that the proposed approach shows high performance, reaching average precision rates up to 0.866 for recall 0.976 – in the case of text with different orientations. [ABSTRACT FROM AUTHOR]

Nowadays, context-aware systems use data obtained from various sources to adapt and provide services of interest to users according to their needs, location or interaction with the corresponding environment. However, the use of heterogeneous sources creates a huge amount of data that may differ in format, transmission speed and may be affected by environmental noise. This generates some inconsistency in data, which must be detected in time to avoid erroneous analysis. This is done using data fusion, which is the action for integrating diverse sources to be analyzed according to a given context. In this work, we propose a scheme of data fusion of heterogeneous sources, supported by a distributed architecture and Bayesian inference as fusion method. As a practical experiment, data were collected from three DHT22 sensors, whose measurements were relative humidity and temperature. The purpose of the experiment was to analyze the variation of these measurements over 24 hours, and fusion them to obtain integrated data. This proposed of data fusion represents an important field of action for the knowledge generation of interest in context-aware systems, for example for the analysis of the environment in order to take advantage of the use of energy and provide a comfortable working environment for the users. [ABSTRACT FROM AUTHOR]

Facial Expression Recognition (FER) is a research area that has been interesting for computer science community in recent years. In this paper, we propose a methodology for the three stages of a FER system. In the pre-processing stage a method based on edge detectors and thresholding operators for eyebrow and mouth segmentation is proposed; the next stage is feature extraction, we propose using polynomials as features for describing eyebrows and mouth regions. Finally, in classification stage different supervised learners such as: Neural Networks, K-Nearest Neighbors and C4.5 decision trees are tested in order to obtain a model for classifying three out of six basic emotions (anger, happiness and surprise). According to our results, the proposed approach has acceptable accuracy for predicting new examples. [ABSTRACT FROM AUTHOR]

The goal of Author Profiling (AP) is to identify demographic aspects (e.g., age, gender) from a given set of authors by analyzing their written texts. Recently, the AP task has gained interest in many problems related to computer forensics, psychology, marketing, but specially in those related with social media exploitation. As known, social media data is shared through a wide range of modalities (e.g., text, images and audio), representing valuable information to be exploited for extracting valuable insights from users. Nevertheless, most of the current work in AP using social media data has been devoted to analyze textual information only, and there are very few works that have started exploring the gender identification using visual information. Contrastingly, this paper focuses in exploiting the visual modality to perform both age and gender identification in social media, specifically in Twitter. Our goal is to evaluate the pertinence of using visual information in solving the AP task. Accordingly, we have extended the Twitter corpus from PAN 2014, incorporating posted images from all the users, making a distinction between tweeted and retweeted images. Performed experiments provide interesting evidence on the usefulness of visual information in comparison with traditional textual representations for the AP task. [ABSTRACT FROM AUTHOR]

Books are an important source of knowledge to disseminate information. Researchers and academicians write books to propagate their innovative research or teachings amongst academic as well as non-academic audience. The number of books written every year is increasing rapidly. According to International Publisher Association (IPA) annual report 2015–2016, around 150 million different books were published worldwide in 2014–2015. Many e-commerce websites are also involved in selling books. A recent addition to book publishing world is e-books, which have really made it very simple to publish. While, availability of large number of books is good for readers, at the same time it is challenging to find a good book, particularly in scholarly settings. Researchers in the area of Scientometrics have attempted to view assessment of goodness of a scholarly book by measuring citations that a book receive. However, citations alone are not a true measure of a book’s impact. Many a times people use the knowledge in a book without actually citing it. Also use of books in classroom settings or for general reading often is not reflected in terms of citations. Therefore, it is important to obtain users’s opinion about a book from other forms of data. Fortunately, we have now some data of this sort available in form of reviews, downloads and social media mentions etc. Amazon and Goodreads, both of which provide the readers’ views about a book, are two good examples. This paper presents an exploratory research work on using these non-traditional data about books to assess impact of a book. A set of Scopus-indexed computer science books with good citations as well as some other popular books in computer science domain are used for analysis. The reviews of books have been crawled in an automated fashion from Amazon and Goodreads. Thereafter sentiment analysis is carried out the text of reviews. Results of sentiment analysis are compared and correlated with traditional impact assessment metrics. The experimental analysis does not show a coherent relationship between citation and online reviews. Also, majority of the online reviews are found to be positive for large number of books in the dataset. As a related exercise, the Scopus citation data and Google scholar citation data for books are also compared. A high value of correlation is observed in these two. Overall the exploratory analysis provides a useful insight into the problem of book impact assessment. [ABSTRACT FROM AUTHOR]

Scholarly articles are considered one of the primary medium for dissemination of inventions and discoveries. Traditionally, usefulness and popularity of a scholarly article has been measured in terms of citations it receives. However, in the changed research publishing landscape, where most of the publications are now available in digital form accessible through various digital libraries; new measures of measuring usefulness of scholarly articles have emerged. Nowadays, scholarly articles are easily available for access and download from various digital access portals. The use and popularity of these digital access portals has also made it possible to integrate various social media platforms with journal access and use. Most of the journals now maintain statistics about reads, number of downloads, social profile shares etc. Several newer platforms like ResearchGate, Academia and Mendeley have also become popular. Researchers now often share their articles on various such platforms and also use social media channels to disseminate their article to a wider audience. This transformed environment has allowed to track and measure usefulness and popularity of scholarly articles through alternative metrics (now popularly known as Altmetrics) as compared to traditional citation impact measures. Altmetrics attempts to derive impact of a scholarly article by using data from different kinds such as social network share, mentions, tweets etc. The use of Altmetrics varies widely from country to country and discipline to discipline. This paper attempts to present findings of an exploratory analysis of relevance of Altmetrics data through a case study of scholarly articles from India published during 2016 and indexed in Web of Science and also updated on ResearchGate. The results obtained provide an interesting insight on relatedness and correlation of presence of scholarly articles in Web of Science and ResearchGate. It is observed that about 61% papers indexed in Web of Science have an entry in ResearchGate. There are, however, disciplinary variations in presence of articles in ResearchGate. Only about 61% of the total disciplines in Web of Science are found to be covered in ResearchGate. [ABSTRACT FROM AUTHOR]

In this work, we present a morphological segmenter for the Mexican indigenous language Wixarika. Segmentation is fundamental for rich morphological languages, a common aspect of the native American languages, to improve other tasks like machine translation, dialogue systems, summarization, etc. On top of the agglutinative nature of the language, the low amount of resources and the lack of an orthographic standard among dialects add to the challenge. Our proposal is based on a probabilistic finite-state approach that exploits regular agglutinative patterns and requires little linguistic knowledge. We show that our approach outperforms unsupervised and semi-supervised methods in a low-resource context. The dataset used in this work was openly released for future work by the community. [ABSTRACT FROM AUTHOR]

This paper proposes an end-to-end Natural Language Generation approach to automatically create fiction stories using statistical language models. The proposed approach integrates the stages of macroplanning and the surface realisation, necessary to determine the content to write about together with the structure of the story, and the syntactic and lexical realisation of sentences to be generated, respectively. Moreover, the use of language models within the stages allows the generation task to be more flexible, as far as the adaptation of the approach to different languages, domains and textual genres is concerned. In order to validate our approach, two evaluations were performed. On the one hand, the influence of integrating position-specific language modelling in the macroplanning stage into the surface realisation module was evaluated. On the other hand, a user evaluation was performed to analyse the generated stories in a qualitative manner. Although there is still room for improvement, the results obtained from the first evaluation in conjunction with the user evaluation feedback shows that the combination of the aforementioned stages in an end-to-end approach is appropriate and have positive effects in the resulting generated text. [ABSTRACT FROM AUTHOR]

Scholarly search engines, reference management tools, and academic social networks enable modern researchers to organize their scientific libraries. Moreover, they often provide recommendations for scientific publications that might be of interest to researchers. Because of the exponentially increasing volume of publications, effective citation recommendation is of great importance to researchers, as it reduces the time and effort spent on retrieving, understanding, and selecting research papers. In this context, we address the problem of citation recommendation, i.e., the task of recommending citations for a new paper. Current research investigates this task in different settings, including cases where rich user metadata is available (e.g., user profile, publications, citations). This work focus on a setting where the user provides only the abstract of a new paper as input. Our proposed approach is to expand the semantic features of the given abstract using knowledge graphs – and, combine them with other features (e.g., indegree, recency) to fit a learning to rank model. This model is used to generate the citation recommendations. By evaluating on real data, we show that the expanded semantic features lead to improving the quality of the recommendations measured by nDCG@10. [ABSTRACT FROM AUTHOR]

A Verbal Phraseological Unit (VPU) is a phrase or expression that has either a literal or figurative (non-compositional) meaning. Identifying, in an automatic manner, which meaning is associated to a VPU in a certain context is an open problem in natural language processing, whose solution impacts in various real life applications. In this paper we present a method for the automatic verification of non-compositionality of a VPU based on the use of lexical domains. The methodology proposed is based in the degree of overlapping between the VPU domain and its context domain. This methdology is general in the sense that it opens the opportunity of applying different overlapping metrics, and different methods for calculating the domain of a given phrase. The experiments carried out show that this novel approach can be efficiently applied in high volumes of data with competitive results close to 65%. [ABSTRACT FROM AUTHOR]

In this paper we explore to what extent text parameters, such as average number of words per sentence, syllables per word, nouns per sentence, frequency of content words, etc. can successfully rank Russian academic texts for different age and grade levels. We provide a brief overview of previous research on readability of Russian texts and describe the corpus of school textbooks on Social Studies (from 5-th to 11-th grade) compiled by the authors. We share our experience of using a variety of quantitative text complexity metrics and evaluate the measures of existing Russian text complexity formulas. Based on the tests of a set of extended text features, we propose one innovative metric for better prediction of Russian text complexity, i.e. the number of adjectives. As the results obtained compare favorably with the previously published results on the established complexity metrics for Russian texts, the study encourages the development of valid, reliable and transparent complexity tools for Russian texts. [ABSTRACT FROM AUTHOR]

This work highlights how to transform information from invoice documents to semantic models, as an implementation of ontology modeling. The migration from printed paper to digital documents in the Mexican Government Offices in the last few years has brought significant opportunities for the usage of information technologies and applications. However, when changing digital document information into knowledge, there are still many gaps to be filled. This work proposes a solution to some issues regarding ontology modeling, specifically when mapping a document that follows some XML schema to an ontology under the OWL standard. The main contribution of this work is to provide new interpretations of the XML terms in the context of OWL, so that the XML Schema Definition (XSD) structures can be mapped into more complex OWL structures. A software tool developed to test and validate the information extraction strategies proposed is presented here. [ABSTRACT FROM AUTHOR]

Text Lines Segmentation (TLS) affects the performance of Manuscript Text Recognition (MTR) systems from document images. At the same time, the TLS task consists of two tasks: the first is Text Lines Localization (TLL) and the second is the Search of the Path that Divides neighboring Lines (SPDL) of handwritten text. The TLS task depends on the type of language, author’s writing style, pen type and document quality. In this paper, Projected Energy Map with Alpha blending (PEM-Alpha) is presented as an unsupervised method for the TLL task, which can work with lines that are touching or overlapping. In addition, SPDL-GA is proposed as a method for SPDL task which finds the line that best splits the text. The experimentation is carried out with a standard collection of historical multilingual documents. Through experimentation it is demostrated that the proposed methods outperform other state-of-the-art methods, even in documents with mixed languages. In addition, few parameters required by PEM-Alpha and SPDL-GA are automatically calculated. [ABSTRACT FROM AUTHOR]

There are several categories of criminal events. However, one of them is focused on people: criminal events against people. It directly affects some of the guarantee of a person or a family. These events are reported in digital media and, without neglecting, digital news media in Spanish. It is relevant to recognize criminal events against people to get useful information about the public security of citizens. Natural Language Processing has techniques that can be possible their identification. However, fine-grained linguistic analysis is required in order to carry out such task. This paper considers the enhancing the discovered patterns with linguistic information (morphological and POS categories) to recognize criminal events against people from Spanish newspapers. Six categories of criminal events are considered: killing, violation, assault, suicide, kidnapping and sexual exploitation. An experimentation is carried out with a gold standard data set of criminal events. The experimentation shows promising results. [ABSTRACT FROM AUTHOR]

This paper presents an approach to solve the author profiling, a text classification task, which consists in determining the demographic and psychological characteristics of an author (like age, gender and personality traits), from some samples of the author’s writing style. The main focus of the approach consists on the creation and enrichment of a co-occurrence graph using the link prediction theory in order to find an author’s profile considering a graph similarity technique (instead of a traditional supervised learning strategy). The proposed method is applied on the English language partition of the CLEF PAN 2015 author profiling task, producing competitive results that are close to the best results reported so far, given the same training and test corpora. The experimental results show that the addition of new edges to a graph representation based on the topological neighborhood of words can be a valuable asset to infer and discover patterns in texts that comes from social media. Additionally, the use of a graph similarity provides a novel way for analyzing how alike are the texts related to a specific demographic or personality aspect against the writing style of an author. [ABSTRACT FROM AUTHOR]

Argumentation in academic writing is a challenging task required to communicate clear ideas. Exposed ideas have to be supported by reasoned arguments. Arguments are composed of components such as premises and conclusions. In this paper, we present an approach to classify argumentative components using language models and machine learning algorithms on a new corpus of academic theses and research proposals. We explore the use of lexical, syntactic, semantic and indicator features to tackle this task. We found that lexical features provide the best efficacy for the classification. For language models, the best features were syntactical. But our experiments showed that a document occurrence representation with unigrams achieved the best accuracy. We also tested the conclusions about the representation and classifier on theses according to their study level (undergraduate, master, and doctoral). We analyzed the information gain of features and found patterns that are part of argumentative markers. [ABSTRACT FROM AUTHOR]

Given the large volumes of information that are generated every day in the Web, Adaptive Information Filtering systems have the potential to become a very useful tool to handle such information overload. These systems allow users to focus on documents that actually meet their information needs, while the system discards the rest. Traditionally, these systems assume that terms of a document are not related to each other, and therefore their efficacy is limited. To overcome this limitation, we propose a method for extracting different relations between the terms of the documents that satisfy the information needs of the users, in order to update the system modeling of such needs, and thereby improve its discrimination capability. These relations are based on the co-occurrence of terms at different levels of granularity, such as document, sentences or noun phrases. The experiments conducted indicate the potential of our proposal, which is capable of improving system efficacy, from the beginning and in the long run. [ABSTRACT FROM AUTHOR]

Resources such as labeled corpora are necessary to train automatic models within the natural language processing (NLP) field. Historically, a large number of resources regarding a broad number of problems are available mostly in English. One of such problems is known as Personality Identification where based on a psychological model (e.g. The Big Five Model), the goal is to find the traits of a subject’s personality given, for instance, a text written by the same subject. In this paper we introduce a new corpus in Spanish called Texts for Personality Identification (TxPI). This corpus will help to develop models to automatically assign a personality trait to an author of a text document. Our corpus, TxPI-u, contains information of 416 Mexican undergraduate students with some demographics information such as, age, gender, and the academic program they are enrolled. Finally, as an additional contribution, we present a set of baselines to provide a comparison scheme for further research. [ABSTRACT FROM AUTHOR]

Paraphrase plagiarism identification represents a very complex task given that plagiarized texts are intentionally modified through several rewording techniques. Accordingly, this paper introduces two new measures for evaluating the relatedness of two given texts: a semantically-informed similarity measure and a semantically-informed edit distance. Both measures are able to extract semantic information from either an external resource or a distributed representation of words, resulting in informative features for training a supervised classifier for detecting paraphrase plagiarism. Obtained results indicate that the proposed metrics are consistently good in detecting different types of paraphrase plagiarism. In addition, results are very competitive against state-of-the art methods having the advantage of representing a much more simple but equally effective solution. [ABSTRACT FROM AUTHOR]

Online social networks like Instagram has more than 600 million users and creates over 300 million new posts every day. All those data can be used to detect real world events. Many works have been proposed in the literature to detect such events using different techniques, but this task is still hard. It involves many challenges including the processing of large volumes of data, the lack of a ground truth and the need for an adaptive approach. In this sense, our work attempts to tackle these problems with a semi-supervised learning approach to overcome those challenges using times series from Instagram posts. Experimental studies demonstrate that similar time series can be used to generalize the knowledge and predict the occurrence of an event. Also, we demonstrate that Support Vector Regression is a good alternative to Gaussian Process Regression as the first provides good results using much less computing resources than the second. Moreover, we made our labeled dataset public, hoping it can be useful to other researchers as well. [ABSTRACT FROM AUTHOR]

Before the advent of the Internet era, code-mixing was mainly used in the spoken form. However, with the recent popular informal networking platforms such as Facebook, Twitter, Instagram, etc., in social media, code-mixing is being used more and more in written form. User-generated social media content is becoming an increasingly important resource in applied linguistics. Recent trends in social media usage have led to a proliferation of studies on social media content. Multilingual social media users often write native language content in non-native script (cross-script). Recently Banerjee et al. [9] introduced the code-mixed cross-script question answering research problem and reported that the ever increasing social media content could serve as a potential digital resource for less-computerized languages to build question answering systems. Question classification is a core task in question answering in which questions are assigned a class or a number of classes which denote the expected answer type(s). In this research work, we address the question classification task as part of the code-mixed cross-script question answering research problem. We combine deep learning framework with feature engineering to address the question classification task and enhance the state-of-the-art question classification accuracy by over 4% for code-mixed cross-script questions. [ABSTRACT FROM AUTHOR]

Feature selection is a crucial aspect in classification problems, especially in domains such as text classification, where usually there is a large number of features. Recently, a two-stage feature selection method for text classification which combines class-based and corpus-based feature selection, was introduced. Based on their experiments, the authors conclude what parameter values for both, corpus-based and class-based approaches, allow a feature selection which improves the traditional methods in text classification. In this paper, we revisit this two-stage feature selection method and based on several experiments we come to a different conclusion: the parameters suggested by the original work do not necessarily provide the best results. Based on our experiments, we conclude that by combining the best parameter value for each stage, for the specific corpus under study, the two stage selection method based on coverage policies provides a subset of features which allows to get statistically significant increase over the traditional methods in the success rates of the classifier. [ABSTRACT FROM AUTHOR]

Recently, the extraction of clinical events from unstructured medical texts has attracted much attention of the research community. Machine learning approaches are popular for this task, due to their ability to solve the problem of sequence tagging effectively. It has been suggested previously that simple features, such as word unigrams, part-of-speech tags, chunk tags, among others, are sufficient for this task. We show that more careful preprocessing and feature selection can significantly improve the results. We used conditional random field classifier with more linguistically oriented features and outperformed the current state-of-the-art approaches. We also show that the popular and much simpler Viterbi algorithm (hidden Markov model-based classification algorithm) can produce competitive results, when its parameters are tuned using specific optimization techniques. We evaluate these algorithms for the task of extraction of medical events from the corpus developed for SemEval shared Task 12: Clinical TempEval (Temporal Evaluation) 2016, namely, for its two subtasks: (i) event detection and (ii) event classification based on contextual modality. [ABSTRACT FROM AUTHOR]

In supervised classification, a training set is given to a classifier to learn a decision rule for classifying unseen cases. When large training sets are processed, the training stage becomes slow especially for instance-based learning. However, not all information in a training set is useful for classification because it could contain either redundant or noisy prototypes. Therefore a process for discarding useless prototypes is required; this process is known as prototype selection. In this work, we present some methods for selecting prototypes based on prototype relevance, which are accurate and fast for large datasets; in addition, our methods can be applied over datasets described by nominal features. We report experimental results showing the effectiveness of our methods as well as a comparison against other successful prototype selection methods. [ABSTRACT FROM AUTHOR]

In this paper, the use of collection term frequencies (i.e. the total number of occurrences of a term in a document collection) in the BM25 retrieval model is investigated by modifying its term frequency (TF) and inverse document frequency (IDF) components. Using selected examples extracted from TREC collections, it was observed that the informative nature, for retrieval purposes, of terms, either with the same TF (in a document) or IDF (in a collection) may be better revealed with the use of collection term frequencies (CTF). From three new heuristics based on those observations and deviations from a random Poisson model, collection term frequencies were integrated to TF and IDF factors. The novel formulations were tested by employing the TREC-1 to TREC-8 collections in the ad hoc task, for which BM25 was first developed and tested. Consistent and significant improvements were observed in mean average precision (MAP) reaching up to 17.67% for the TREC-8 dataset, and 7.16% averaged over all tested collections. These results were considerably better in comparison to other approaches surveyed aiming to improve BM25, proving in this way the effectiveness of the proposed heuristics and formulae. The proposed approach requires only additional offline pre-computations and does not entail extra computational complexity for retrieval while keeping the original spirit and parameter robustness of BM25. [ABSTRACT FROM AUTHOR]

We explore various machine learning-based classifiers applied to rule-based features for recognizing textual entailment. The features, extracted with a set of synthesized matching rules, reflect syntactic and semantic similarity between the text and the hypothesis. The fact that we use only seven relatively simple features makes our method suitable for low-resource languages. We test our method on the test sets of the RTE competitions and achieve accuracy of up to 69.13%. [ABSTRACT FROM AUTHOR]

Given the immediacy of social networks, commonly called sociodigital networks, it is necessary to develop methods to retrieve and interpret visually and in an organized way large amounts of information. Although there are tools that classify the information by using a visualization, generally in form of graphs, the identification of the topics around an event remains complicated. This article describes the use of dendrograms, as a different visual representation, by analyzing the frequency of the terms used in the tweets as well as the relationship between them. Thus, the use of semantic dendrograms facilitates the immediate identification of themes and subtopics of a given event by showing a clustering of these in the form of a tree. [ABSTRACT FROM AUTHOR]

Actionable Patterns are desired knowledge to be mined from large datasets. Action Rules are vital data mining method for gaining actionable knowledge from the datasets. They recommend actions which users can undertake to their advantage, or to accomplish their goal. Meta actions are the sub-actions to the Action Rules, which intends to change the attribute value of an object, under consideration, to attain the desirable value. The essence of this paper is to propose a new optimized and more promising system, in terms of speed and efficiency, for generating meta-actions by implementing Specific Action Rule discovery based on Grabbing strategy (SARGS) algorithm, and to apply that for Sentiment Analysis on Twitter data. We perform a comparative analysis of meta-actions generating algorithmic implementation in Apache Spark driven system, conventional Hadoop driven system and Single node machine using the Twitter social networking data and evaluate the results. We implement corpus based Sentimental Analysis of social networking data, and test the total time taken by the systems and their sub components for the data processing. Results show faster computational time for Spark system compared to Hadoop MapReduce and Single node machine for the meta-action generation methods. [ABSTRACT FROM AUTHOR]

Recently, available data has increased explosively in both number of samples and dimensionality. The huge number of high dimensional data generates the presence of noisy, redundant and irrelevant dimensions. Such dimensions can increase the time and computational cost in the learning process and even degenerate the performance of learning tasks. One of the ways to reduce dimensionality is by Feature Selection (FS). The aim of this paper is study the feature selection based on expert knowledge and traditional methods (filter, wrapper and embedded) and analyze their performance in classification tasks. Three datasets related to cancer domain in humans were used for feature selection: Breast Cancer (BC), Primary Tumor (PT) and Central Nervous System (CNS). C4.5, K-Nearest Neighbors, Support Vector Machine and Multi Layer Perceptron were trained with the best subset of features for each cancer dataset. The subset of features selected by the wrapper method presents the best average accuracy in the datasets BC and PT, while the subset of features selected by the embedded method reaches the highest average accuracy in the CNS dataset. [ABSTRACT FROM AUTHOR]

Social networks users often post their opinion after reading a news article. By analyzing these responses, it is possible to find diverse emotions expressed in them. When several users react to an article, a distribution of these emotions is accumulated. Writers and publishers would benefit to have an estimation of how users will react to an article. This work proposes a method to predict the distribution of emotions that users would express in Twitter after reading a news article. More than one emotion can be expressed in responses, so that an approach of modeling this distribution as a supervised multi-target classification problem is followed. For this purpose, it was necessary to collect a corpus of Spanish news articles and their associated responses and a group of annotators tagged the emotions expressed in them. The use of this strategy allows to naturally model instances (news articles) that have more than one associated class (emotions expressed in responses). The predicted values are expressed in terms of the percentage of responses that triggered each specific emotion. The proposed method is evaluated by measuring the deviation of the predicted emotion distribution with regard to the annotated set of emotions, obtaining a precision above 90%. In addition to that, the proposed method was used in a foreign corpus in order to compare it with 10 state of the art methods. Results show that the proposed method performs better than 9 of these methods on this corpus. [ABSTRACT FROM AUTHOR]