Revista Romana de Interactiune Om-Calculator

273-288

Authors: Liviu Sebastian Matei, Stefan Trausan Matu

Identificarea entităţilor cu nume într-un text reprezintă un subiect important pentru prelucrarea limbajului natural. În cadrul acestui articol se descrie o metodă nouă de detectare a entităţilor cu nume care încearcă să îmbunătăţească rezultatele obţinute cu detectorul de entităţi cu nume de la Stanford. Modelul prezentat va folosi clasificatorul Bayesian naiv pentru a determina şi clasifica entităţile cu nume.

305-320

Authors: Alexandra-Mihaela Siriteanu, Adrian Iftene

În ultimii ani, odată cu folosirea din ce în ce mai intensă a reţelelor sociale şi a Internetului, utilizatorii au acces la tot mai multe informaţii atât de tip text, cât şi de tip multimedia. Cu toate acestea, găsirea informaţiilor dorite a devenit din ce în ce mai dificil, întrucât majoritatea motoarelor de căutare îşi bazează algoritmii de ordonare a rezultatelor pe criterii precum: popularitate, relevanţă şi mai puţin pe disimilaritate. În cadrul proiectului MUCKE (Multimedia and User Credibility Knowledge Extraction), un proiect CHIST-ERA sponsorizat de Uniunea Europeană, am creat un sistem ce permite căutarea de imagini şi în care abordăm problema diversificării rezultatelor multimedia. Pentru a trata problema diversificării am explorat două metode: (1) folosirea de resurse semantice (precum YAGO) în procesul de prelucrare a interogării iniţiale şi (2) utilizare operaţiilor de clusterizare a rezultatelor, pe baza titlurilor imaginilor sau a conţinutului acestora. În lucrarea de faţă, vom prezenta prima abordare şi vom compara rezultatele obţinute de aplicaţia noastră cu rezultatele oferite de motorul de căutare Google.

333-352

Authors: Teodor Stefanut, Dorian Gorgan

Materialele didactice multimedia (imagini, fişiere video, reprezentări fotorealiste 3D) permit exemplificarea şi reprezentarea intuitivă a unor noţiuni de complexitate ridicată. În acelaşi timp însă, interacţiunea cu aceste resurse nu poate fi realizată în mod eficient folosind analiză şi comunicare bazată pe text (chat, email, forum). Astfel, pentru a permite persoanelor implicate în procesul didactic, instructori sau cursanţi, să se concentreze mai mult asupra conţinutului şi mai puţin asupra comunicării, propunem utilizarea adnotărilor grafice ca fundament al unui nou stil de interacţiune utilizator. Tehnici specifice de interacţiune vor fi prezentate pe parcursul acestui articol, la fiecare etapă a procesului didactic: crearea materialelor, analiza lor, comunicarea între participanţi pe baza informaţiilor prezentate şi chiar evaluarea cunoştinţelor. De asemenea, va fi propus un nou model de codificare şi reprezentare a lecţiei, care să permită integrarea naturală a adnotărilor grafice în structurile de date, atât la nivel de control (ex. generare de materiale) cât şi la nivelul reprezentării cunoştinţelor (ex. codificarea răspunsurilor cursanţilor).

195-208

Authors: Dragos Daniel Iordache

Social Networking Sites (SNSs) allow people to present themselves, join interest groups and establish or maintain connections with others. Most social networks are for teenagers and adults, although some have no lower age limit. These sites can be oriented towards work-related contexts (LinkedIn.com), connecting users with common interests or for university students (original destination of Facebook.com). The goal of this paper is to identify the main factors that motivate students in Romania to use various social networks, the people encouraged to use them as well as the limits and possible risks associated with use of SNSs. Results are based on content analysis of the responses to the open questions of the instrument applied to 172 students from the University of Bucharest and Academy of Economic Studies in Bucharest

225-238

Authors: Oana Balan, Alin Moldoveanu, Florica Moldoveanu, Oana-Maria Ferche

Making computer games accessible to blind people is a very difficult task. As the majority of computer games available on the market today are graphics-based, the visually impaired part of the population is highly deprived of this useful technological means of entertainment and education. This lack of accessibility has led to the development of adaptable and entertaining audio-based games, since the late 1970s. As hearing is a relatively reliable sense that can, to some extent, replace sight, these ludic virtual environments render their complete experience through auditory cues. Even if at the moment audio games do not draw the attention of developers as much as their video counterparts, we believe that they present obvious advantages. They should be studied more intensively in order to lead to the development of useful and accessible applications that would improve the interaction between the computer and the blind users. This paper aims to present a brief classification of the most significant audio games and to discuss their functional aspects, the most used means of interaction and the best design strategies related to the development of alternative game approaches towards facilitating the connection between the visually-disadvantaged players and the computer.

259-272

Authors: Costin Pribeanu, Paul Fogarassy-Neszly

Automatic language identification in a given text belongs to a general category of algorithms of text classifications with many applications. Some recent developments concerns language dependent speech synthesis. For such application, dynamic language identification is required, involving special algorithms, capable to dynamically recognize the language in texts received as a continuous stream of data. This paper presents some preliminary results during the testing of two different versions of the dynamic language identification implementation (alpha and beta), with emphasis on the second one. Two parameters were manipulated during testing: the amount of analyzed text and the inertia in changing the language. The evaluation has been done from a usability perspective with a focus on the effectiveness and efficiency of language identification. Beta testing results are useful to reveal the optimal level of parameters and suggest new directions to improve the method and algorithms.

89-92

Authors: David Faure, Jean Vanderdonckt

The UsiXML ITEA 2 project developed an innovative model-driven engineering method to improve the user interface design for the benefit of both industrial and academic end-users in terms of productivity and reusability. The goal of UsiXML was to define, validate, and standardise an open user interface description language (UIDL), increasing productivity and reusability, and improving usability and accessibility. The development efforts in UsiXML resulted in a draft submitted to W3C and were rewarded with the ITEA Excellence Award 2013 for standardization.

117-138

Authors: Javier R. Escolar, Cristina G.Cachón, Ignacio Marín, Jean Vanderdonckt, Vivian Motti

The development of context-sensitive applications for the mobile web implies significant challenges for developers, mainly due to device diversity and the variability of the Delivery Context. One specific challenge is to create applications able to work not only when connection is available but also when disconnected. The effort required to create such applications is excessively high in spite of the facilities provided by emerging technologies. A model-based approach for the development of this type of applications is presented in this article. We present a navigation model described by means of an extended version of the State Charts eXtensible Markup Language (SCXML) proposed by W3C.

161-194

Authors: Iyad Khaddam, Jean Vanderdonckt

Culture plays a role in the way humans interact with the machine. It implies changes on the UI to fit cultural needs, practices and preferences. Knowledge on these implications are sparsed and insufficient until the moment. Employing culture knowledge as rules in the design and development of software user interfaces looks far to achieve. Even though, user interface designers and developers still need to consider differences in user interfaces. Our approach to support user interface developers and designers is to propose a user interface architecture that enhances their ability to apply changes imposed by culture. Our contribution is to analyse culture studies to identify concerns and requirements for such architecture, propose an architecture that satisfies some of these requirements and discuss an idea towards an enhanced one.

1-20

Authors: Iulia Maria Florea, Traian Rebedea, Costin-Gabriel Chiru

Determining the syntactic dependencies is an important task in natural language processing, as it is useful for improving the results of a wide range of applications, such as machine translation, opinion mining, question-answering systems and others. This paper presents an initial step for semi-automatically building a corpus annotated with syntactic dependencies for Romanian and enriched with information about the type of the words and the relationships between them. As Romanian lacks an open-source syntactic or dependency parser specially trained on Romanian phrases, this corpus is necessary for obtaining better results for linguistic applications that are in need of dependency parse trees. To achieve this, we have started from two types of well-known parsers, the first trained for French syntactic parsing and the second for Spanish dependency parsing, that have been modified for analyzing phrases written in Romanian. The results we have obtained using these two methods are compared with the ones returned by the single existing dependency parser trained for Romanian on a medium-size corpus, which is available at this moment as a web service.

37-52

Authors: Daniela-Anca Sarbu

Telemedicine researches nowadays focus more and more on the complexity of the services that can be delivered and on the development of new paradigms of healthcare delivery. The paper proposes a different approach, a study in one area that has been neglected, but might provide useful insights for optimizing telemedicine services offering and that consists in the analysis of the data generated by a telemedicine application and correlating it with the geographical area where the telemedicine services were accessed. The scope of the study is to identify the most relevant data held by a telecommunication operator and processing it to expose the results in a simple and intuitive manner. Thus, a clear imagine of the usage indicators and the manner in which this usage can be correlated with the geographical position will bring a significant contribution in decision making process of optimizing telemedicine services offering. With this purpose we propose as solution an analytical platform based on a datamart that aliments an OLAP cube for enabling queries on different axes of interest. The paper presents both the data modeling and architecture, and the way of visualizing the results. The results visualization is done in an interactive manner, through personalized dashboards that contain the aggregated information at the highest level, enabling decision making in telemedicine services offering and delivery.

71-88

Authors: Denisa Copândean, Adrian Sabou, Dorian Gorgan

In order to develop flexible, highly-interactive applications we must ensure interaction between the user and modeled objects. Particle-based modeling introduces specific challenges with regard to interaction in the virtual environment, requiring the development of particular interaction techniques in order to simulate the user’s natural operations. This paper presents user interaction techniques with deformable particle-based models, namely selection, manipulation and volume slicing. We also describe the concepts and techniques that form the basis of these interactions, such as bounding volume and the implementation of the slicing instrument. Finally we describe challenges occurring when ensuring interaction in three-dimensional scene using bi-dimensional capable input devices.