Showing papers by "INESC-ID published in 2015"

Finding Function in Form: Compositional Character Models for Open Vocabulary Word Representation

Abstract: We introduce a model for constructing vector representations of words by composing characters using bidirectional LSTMs. Relative to traditional word representation models that have independent vectors for each word type, our model requires only a single vector per character type and a fixed set of parameters for the compositional model. Despite the compactness of this model and, more importantly, the arbitrary nature of the form‐function relationship in language, our “composed” word representations yield state-of-the-art results in language modeling and part-of-speech tagging. Benefits over traditional baselines are particularly pronounced in morphologically rich languages (e.g., Turkish).

Two/Too Simple Adaptations of Word2Vec for Syntax Problems

Abstract: We present two simple modifications to the models in the popular Word2Vec tool, in order to generate embeddings more suited to tasks involving syntax. The main issue with the original models is the fact that they are insensitive to word order. While order independence is useful for inducing semantic representations, this leads to suboptimal results when they are used to solve syntax-based problems. We show improvements in part-ofspeech tagging and dependency parsing using our proposed models.

Improved Transition-Based Parsing by Modeling Characters instead of Words with LSTMs

Abstract: We present extensions to a continuous-state dependency parsing method that makes it applicable to morphologically rich languages. Starting with a high-performance transition-based parser that uses long short-term memory (LSTM) recurrent neural networks to learn representations of the parser state, we replace lookup-based word representations with representations constructed from the orthographic representations of the words, also using LSTMs. This allows statistical sharing across word forms that are similar on the surface. Experiments for morphologically rich languages show that the parsing model benefits from incorporating the character-based encodings of words.

Improved Transition-based Parsing by Modeling Characters instead of Words with LSTMs

Abstract: We present extensions to a continuousstate dependency parsing method that makes it applicable to morphologically rich languages. Starting with a highperformance transition-based parser that uses long short-term memory (LSTM) recurrent neural networks to learn representations of the parser state, we replace lookup-based word representations with representations constructed from the orthographic representations of the words, also using LSTMs. This allows statistical sharing across word forms that are similar on the surface. Experiments for morphologically rich languages show that the parsing model benefits from incorporating the character-based encodings of words.

Not All Contexts Are Created Equal: Better Word Representations with Variable Attention

Abstract: We introduce an extension to the bag-ofwords model for learning words representations that take into account both syntactic and semantic properties within language. This is done by employing an attention model that finds within the contextual words, the words that are relevant for each prediction. The general intuition of our model is that some words are only relevant for predicting local context (e.g. function words), while other words are more suited for determining global context, such as the topic of the document. Experiments performed on both semantically and syntactically oriented tasks show gains using our model over the existing bag of words model. Furthermore, compared to other more sophisticated models, our model scales better as we increase the size of the context of the model.

Overview of insular power systems under increasing penetration of renewable energy sources: Opportunities and challenges

Abstract: Insular electricity grids are considered to have a more fragile structure than the mainland ones due to several factors such as the lower inertia because of lower number of generation facilities connected to the system, absence or insufficient interconnection with the main grid, etc. The recent trend of integrating large portions of environmentally sustainable power generation units that have a significantly volatile nature in the generation mix (such as wind and solar energy conversion systems) within this fragile structure, poses profound challenges that need deeper and specific analysis. This study aims to provide an overview of insular power system structures and operational requirements, especially under increasing penetration of renewable energy sources. Firstly, a general evaluation of insular power systems is presented. Then, potential challenges are thoroughly discussed together with opportunities to tackle them. Future technological developments, as well as innovative applications are also given special attention. Hence, this paper contributes to the scarce literature regarding insular power systems by providing a critical overview of issues regarding their operation and possible solutions.

Physio@Home: Exploring Visual Guidance and Feedback Techniques for Physiotherapy Exercises

Abstract: Physiotherapy patients exercising at home alone are at risk of re-injury since they do not have corrective guidance from a therapist. To explore solutions to this problem, we designed Physio@Home, a prototype that guides people through prerecorded physiotherapy exercises using real-time visual guides and multi-camera views. Our design addresses several aspects of corrective guidance, including: plane and range of movement, joint positions and angles, and extent of movement. We evaluated our design, com-paring how closely people could follow exercise movements under various feedback conditions. Participants were most accurate when using our visual guide and multi-views. We provide suggestions for exercise guidance systems drawn from qualitative findings on visual feedback complexity.

Cooperative development of logical modelling standards and tools with CoLoMoTo.

Abstract: The identification of large regulatory and signalling networks involved in the control of crucial cellular processes calls for proper modelling approaches. Indeed, models can help elucidate properties of these networks, understand their behaviour and provide (testable) predictions by performing in silico experiments. In this context, qualitative, logical frameworks have emerged as relevant approaches, as demonstrated by a growing number of published models, along with new methodologies and software tools. This productive activity now requires a concerted effort to ensure model reusability and interoperability between tools. Following an outline of the logical modelling framework, we present the most important achievements of the Consortium for Logical Models and Tools, along with future objectives. Our aim is to advertise this open community, which welcomes contributions from all researchers interested in logical modelling or in related mathematical and computational developments.

Solving QBF by clause selection

Abstract: Algorithms based on the enumeration of implicit hitting sets find a growing number of applications, which include maximum satisfiability and model based diagnosis, among others. This paper exploits enumeration of implicit hitting sets in the context of Quantified Boolean Formulas (QBF). The paper starts by developing a simple algorithm for QBF with two levels of quantification, which is shown to relate with existing work on enumeration of implicit hitting sets, but also with recent work on QBF based on abstraction refinement. The paper then extends these ideas and develops a novel QBF algorithm, which generalizes the concept of enumeration of implicit hitting sets. Experimental results, obtained on representative problem instances, show that the novel algorithm is competitive with, and often outperforms, the state of the art in QBF solving.

Model Checking to Assess T-Helper Cell Plasticity

Abstract: Computational modeling constitutes a crucial step toward the functional understanding of complex cellular networks. In particular, logical modeling has proven suitable for the dynamical analysis of large signaling and transcriptional regulatory networks. In this context, signaling input components are generally meant to convey external stimuli, or environmental cues. In response to such external signals, cells acquire specific gene expression patterns modeled in terms of attractors (e.g., stable states). The capacity for cells to alter or reprogram their differentiated states upon changes in environmental conditions is referred to as cell plasticity. In this article, we present a multivalued logical framework along with computational methods recently developed to efficiently analyze large models. We mainly focus on a symbolic model checking approach to investigate switches between attractors subsequent to changes of input conditions. As a case study, we consider the cellular network regulating the differentiation of T-helper (Th) cells, which orchestrate many physiological and pathological immune responses. To account for novel cellular subtypes, we present an extended version of a published model of Th cell differentiation. We then use symbolic model checking to analyze reachability properties between Th subtypes upon changes of environmental cues. This allows for the construction of a synthetic view of Th cell plasticity in terms of a graph connecting subtypes with arcs labeled by input conditions. Finally, we explore novel strategies enabling specific Th cell polarizing or reprograming events.

Automatic Keyword Extraction on Twitter

Abstract: In this paper, we build a corpus of tweets from Twitter annotated with keywords using crowdsourcing methods. We identify key differences between this domain and the work performed on other domains, such as news, which makes existing approaches for automatic keyword extraction not generalize well on Twitter datasets. These datasets include the small amount of content in each tweet, the frequent usage of lexical variants and the high variance of the cardinality of keywords present in each tweet. We propose methods for addressing these issues, which leads to solid improvements on this dataset for this task.

Mobility Network Evaluation in the User Perspective: Real-Time Sensing of Traffic Information in Twitter Messages

Abstract: We discuss a new type of mobility study focusing on subjective user opinions of mobility networks. Taking advantage of the growth in popularity of opinion mining in social media, and social media itself, we present the architecture of a system capable of automatically capturing user perspective toward a mobility network, based on web user-generated content. We discuss the value of acquiring this subjective information, especially for urban planners, and its contribution to an overall understanding of human mobility. We look at users as sensors of mobility dynamics, capable of providing an insight into the flaws of mobility networks. To achieve this, this chapter presents an initial attempt to use microblogging messages posted on Twitter (by users in transit) to perform real-time sensing of traffic-related information. We propose a text classification approach to the problem: we wish to automatically identify traffic-related messages posted on Twitter, among the millions of unrelated messages posted by users.

A Survey on Implicit Surface Polygonization

Abstract: Implicit surfaces (IS) are commonly used in image creation, modeling environments, modeling objects, and scientific data visualization. In this article, we present a survey of different techniques for fast visualization of IS. The main classes of visualization algorithms are identified along with the advantages of each in the context of the different types of IS commonly used in computer graphics. We focus closely on polygonization methods, as they are the most suited to fast visualization. Classification and comparison of existing approaches are presented using criteria extracted from current research. This enables the identification of the best strategies according to the number of specific requirements, such as speed, accuracy, quality, or stylization.

Proof Complexity of Resolution-based QBF Calculi

Abstract: Proof systems for quantified Boolean formulas (QBFs) provide a theoretical underpinning for the performance of important QBF solvers. However, the proof complexity of these proof systems is currently not well understood and in particular lower bound techniques are missing. In this paper we exhibit a new and elegant proof technique for showing lower bounds in QBF proof systems based on strategy extraction. This technique provides a direct transfer of circuit lower bounds to lengths of proofs lower bounds. We use our method to show the hardness of a natural class of parity formulas for Q-resolution and universal Q-resolution. Variants of the formulas are hard for even stronger systems as long-distance Q-resolution and extensions. With a completely different lower bound argument we show the hardness of the prominent formulas of Kleine Buning et al. [34] for the strong expansion-based calculus IR-calc. Our lower bounds imply new exponential separations between two different types of resolution-based QBF calculi: proof systems for CDCL-based solvers (Q-resolution, long-distance Q-resolution) and proof systems for expansion-based solvers (forallExp+Res and its generalizations IR-calc and IRM-calc). The relations between proof systems from the two different classes were not known before.

INESC-ID: A Regression Model for Large Scale Twitter Sentiment Lexicon Induction

Abstract: We present the approach followed by INESCID in the SemEval 2015 Twitter Sentiment Analysis challenge, subtask E. The goal was to determine the strength of the association of Twitter terms with positive sentiment. Using two labeled lexicons, we trained a regression model to predict the sentiment polarity and intensity of words and phrases. Terms were represented as word embeddings induced in an unsupervised fashion from a corpus of tweets. Our system attained the top ranking submission, attesting the general adequacy of the proposed approach.