Expressiveness improvements of OutSystems DSL query primitives

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

Analysable software language translations

Dissertação para obtenção do Grau de Doutor em Engenharia Informática

Learning Preferences with Kernel-Based Methods

Learning of preference relations has recently received significant attention in machine learning community. It is closely related to the classification and regression analysis and can be reduced to these tasks. However, preference learning involves prediction of ordering of the data points rather than prediction of a single numerical value as in case of regression or a class label as in case of classification. Therefore, studying preference relations within a separate framework facilitates not only better theoretical understanding of the problem, but also motivates development of the efficient algorithms for the task. Preference learning has many applications in domains such as information retrieval, bioinformatics, natural language processing, etc. For example, algorithms that learn to rank are frequently used in search engines for ordering documents retrieved by the query. Preference learning methods have been also applied to collaborative filtering problems for predicting individual customer choices from the vast amount of user generated feedback. In this thesis we propose several algorithms for learning preference relations. These algorithms stem from well founded and robust class of regularized least-squares methods and have many attractive computational properties. In order to improve the performance of our methods, we introduce several non-linear kernel functions. Thus, contribution of this thesis is twofold: kernel functions for structured data that are used to take advantage of various non-vectorial data representations and the preference learning algorithms that are suitable for different tasks, namely efficient learning of preference relations, learning with large amount of training data, and semi-supervised preference learning. Proposed kernel-based algorithms and kernels are applied to the parse ranking task in natural language processing, document ranking in information retrieval, and remote homology detection in bioinformatics domain. Training of kernel-based ranking algorithms can be infeasible when the size of the training set is large. This problem is addressed by proposing a preference learning algorithm whose computation complexity scales linearly with the number of training data points. We also introduce sparse approximation of the algorithm that can be efficiently trained with large amount of data. For situations when small amount of labeled data but a large amount of unlabeled data is available, we propose a co-regularized preference learning algorithm. To conclude, the methods presented in this thesis address not only the problem of the efficient training of the algorithms but also fast regularization parameter selection, multiple output prediction, and cross-validation. Furthermore, proposed algorithms lead to notably better performance in many preference learning tasks considered.

Identifying language needs of ESL students in a Canadian university based intensive English language program

This study investigated the needs of adult ESL learners intending to pursue higher education in Canada. Its chief purpose was to enable educators and administrators to design ESL programs that would prepare students to function at optimal levels in academic and social settings during their university studies. The study adopted a mixed research method that was predominantly qualitative in its orientation and narrative in its implementation. It focused on an Intensive English Language Program (IELP) offered at an Ontario university. Using a holistic approach, the study sought to represent the various perspectives of all the participants in the program: the students, the instructors, and the administrators. Analysis of both qualitative and quantitative data gathered from 17 students, 6 instructors, and 1 administrator in the IELP showed that to a large extent the academic needs ofESL learners in the IELP were generally not being met. Most notably, the study found that learners were not receiving sufficient training in speaking and listening skills, a factor that contributed to their sense of insecurity and lack of confidence in their ability to communicate successfully in academic and social settings. The study also revealed that the solutions to many of the problems it identified lay not in the classroom but in the way the ESL program was structured administratively. One major recommendation to come out of the study is that programs like the IELP should be restructured so as to give them greater flexibility in meeting individual needs. While the study labored under certain limitations and did not achieve all of its goals, it did succeed in creating awareness ofthe problems and in establishing a methodological approach that can serve as a framework within which future research may be conducted in this somewhat neglected area.

Design And Development Of A Named Entity Based Question Answering System For Malayalam Language

This is a Named Entity Based Question Answering System for Malayalam Language. Although a vast amount of information is available today in digital form, no effective information access mechanism exists to provide humans with convenient information access. Information Retrieval and Question Answering systems are the two mechanisms available now for information access. Information systems typically return a long list of documents in response to a user’s query which are to be skimmed by the user to determine whether they contain an answer. But a Question Answering System allows the user to state his/her information need as a natural language question and receives most appropriate answer in a word or a sentence or a paragraph. This system is based on Named Entity Tagging and Question Classification. Document tagging extracts useful information from the documents which will be used in finding the answer to the question. Question Classification extracts useful information from the question to determine the type of the question and the way in which the question is to be answered. Various Machine Learning methods are used to tag the documents. Rule-Based Approach is used for Question Classification. Malayalam belongs to the Dravidian family of languages and is one of the four major languages of this family. It is one of the 22 Scheduled Languages of India with official language status in the state of Kerala. It is spoken by 40 million people. Malayalam is a morphologically rich agglutinative language and relatively of free word order. Also Malayalam has a productive morphology that allows the creation of complex words which are often highly ambiguous. Document tagging tools such as Parts-of-Speech Tagger, Phrase Chunker, Named Entity Tagger, and Compound Word Splitter are developed as a part of this research work. No such tools were available for Malayalam language. Finite State Transducer, High Order Conditional Random Field, Artificial Immunity System Principles, and Support Vector Machines are the techniques used for the design of these document preprocessing tools. This research work describes how the Named Entity is used to represent the documents. Single sentence questions are used to test the system. Overall Precision and Recall obtained are 88.5% and 85.9% respectively. This work can be extended in several directions. The coverage of non-factoid questions can be increased and also it can be extended to include open domain applications. Reference Resolution and Word Sense Disambiguation techniques are suggested as the future enhancements

Intelligent Agent-based Multilingual Information Retrieval System

The goal of this work is to develop an Open Agent Architecture for Multilingual information retrieval from Relational Database. The query for information retrieval can be given in plain Hindi or Malayalam; two prominent regional languages of India. The system supports distributed processing of user requests through collaborating agents. Natural language processing techniques are used for meaning extraction from the plain query and information is given back to the user in his/ her native language. The system architecture is designed in a structured way so that it can be adapted to other regional languages of India

The Developmental Sentence Analysis and the Grammatical Analysis of Elicited Language - Complex Level: a correlative study

This paper studies the Developmental Sentence Analysis (DSA), which measures syntactic maturity of spontaneous utterances, and the Grammatical Analysis of Elicited Language (GAEL), which uses a highly structured set of games and activities designed to elicit specific target sentences.

Motivation in second language acquisition in Chinese research students in the UK

The present study sets out to examine motivation to learn English by Chinese research students in an informal setting. Data were collected, using semi-structured interviews, from four research students at two points in time during their first year in the UK. The main findings are: they believed that learning English was important; their main goal orientations were instrumental and extrinsic; they set learning goals and persisted to attain them; they valued their current learning environment in general and saw it as supportive of their goals; they held both positive and negative attitudes towards the British, which had differential effects on their motivation; their self-perceived support seemed to have a positive impact on their motivation and the development of self-confidence; they tended to attribute their success to stable causes such as the environment and failure to unstable but controllable causes such as effort. It is concluded that qualitative data of this kind may complement insights from quantitative research. Implications for target country institutions in the provision of support are discussed.

Task design and second language performance: the effect of narrative type on learner output

This article reports on a detailed empirical study of the way narrative task design influences the oral performance of second-language (L2) learners. Building on previous research findings, two dimensions of narrative design were chosen for investigation: narrative complexity and inherent narrative structure. Narrative complexity refers to the presence of simultaneous storylines; in this case, we compared single-story narratives with dual-story narratives. Inherent narrative structure refers to the order of events in a narrative; we compared narratives where this was fixed to others where the events could be reordered without loss of coherence. Additionally, we explored the influence of learning context on performance by gathering data from two comparable groups of participants: 60 learners in a foreign language context in Teheran and 40 in an L2 context in London. All participants recounted two of four narratives from cartoon pictures prompts, giving a between-subjects design for narrative complexity and a within-subjects design for inherent narrative structure. The results show clearly that for both groups, L2 performance was affected by the design of the task: Syntactic complexity was supported by narrative storyline complexity and grammatical accuracy was supported by an inherently fixed narrative structure. We reason that the task of recounting simultaneous events leads learners into attempting more hypotactic language, such as subordinate clauses that follow, for example, while, although, at the same time as, etc. We reason also that a tight narrative structure allows learners to achieve greater accuracy in the L2 (within minutes of performing less accurately on a loosely structured narrative) because the tight ordering of events releases attentional resources that would otherwise be spent on finding connections between the pictures. The learning context was shown to have no effect on either accuracy or fluency but an unexpectedly clear effect on syntactic complexity and lexical diversity. The learners in London seem to have benefited from being in the target language environment by developing not more accurate grammar but a more diverse resource of English words and syntactic choices. In a companion article (Foster & Tavakoli, 2009) we compared their performance with native-speaker baseline data and see that, in terms of nativelike selection of vocabulary and phrasing, the learners in London are closing in on native-speaker norms. The study provides empirical evidence that L2 performance is affected by task design in predictable ways. It also shows that living within the target language environment, and presumably using the L2 in a host of everyday tasks outside the classroom, confers a distinct lexical advantage, not a grammatical one.

The Whorfian mind : electrophysiological evidence that language shapes perception

Color perception has been a traditional test-case of the idea that the language we speak affects our perception of the world.1 It is now established that categorical perception of color is verbally mediated and varies with culture and language.2 However, it is unknown whether the well-demonstrated language effects on color discrimination really reach down to the level of visual perception, or whether they only reflect post-perceptual cognitive processes. Using brain potentials in a color oddball detection task with Greek and English speakers, we demonstrate that language effects may exist at a level that is literally perceptual, suggesting that speakers of different languages have differently structured minds.

Writing in French in secondary schools in England and Germany: are the British really 'bad language learners’?

It is widely assumed that the British are poorer modern foreign language (MFL) learners than their fellow Europeans. Motivation has often been seen as the main cause of this perceived disparity in language learning success. However, there have also been suggestions that curricular and pedagogical factors may play a part. This article reports a research project investigating how German and English 14- to 16-year-old learners of French as a first foreign language compare to one another in their vocabulary knowledge and in the lexical diversity, accuracy and syntactic complexity of their writing. Students from comparable schools in Germany and England were set two writing tasks which were marked by three French native speakers using standardised criteria aligned to the Common European Framework of Reference (CEF). Receptive vocabulary size and lexical diversity were established by the X_lex test and a verb types measure respectively. Syntactic complexity and formal accuracy were respectively assessed using the mean length of T-units (MLTU) and words/error metrics. Students' and teachers' questionnaires and semi-structured interviews were used to provide information and participants' views on classroom practices, while typical textbooks and feedback samples were analysed to establish differences in materials-related input and feedback in the two countries. The German groups were found to be superior in vocabulary size, and in the accuracy, lexical diversity and overall quality – but not the syntactic complexity – of their writing. The differences in performance outcomes are analysed and discussed with regard to variables related to the educational contexts (e.g. curriculum design and methodology).

A game theoretical approach to the evolution of structured communication codes

Structured meaning-signal mappings, i.e., mappings that preserve neighborhood relationships by associating similar signals with similar meanings, are advantageous in an environment where signals are corrupted by noise and sub-optimal meaning inferences are rewarded as well. The evolution of these mappings, however, cannot be explained within a traditional language evolutionary game scenario in which individuals meet randomly because the evolutionary dynamics is trapped in local maxima that do not reflect the structure of the meaning and signal spaces. Here we use a simple game theoretical model to show analytically that when individuals adopting the same communication code meet more frequently than individuals using different codes-a result of the spatial organization of the population-then advantageous linguistic innovations can spread and take over the population. In addition, we report results of simulations in which an individual can communicate only with its K nearest neighbors and show that the probability that the lineage of a mutant that uses a more efficient communication code becomes fixed decreases exponentially with increasing K. These findings support the mother tongue hypothesis that human language evolved as a communication system used among kin, especially between mothers and offspring.

Extraction of structured programs from specification proofs

We present a method using an extended logical system for obtaining programs from specifications written in a sublanguage of CASL. These programs are “correct” in the sense that they satisfy their specifications. The technique we use is to extract programs from proofs in formal logic by techniques due to Curry and Howard. The logical calculus, however, is novel because it adds structural rules corresponding to the standard ways of modifying specifications: translating (renaming), taking unions, and hiding signatures. Although programs extracted by the Curry-Howard process can be very cumbersome, we use a number of simplifications that ensure that the programs extracted are in a language close to a standard high-level programming language. We use this to produce an executable refinement of a given specification and we then provide a method for producing a program module that maximally respects the original structure of the specification. Throughout the paper we demonstrate the technique with a simple example.

A model and an algebra for semi-structured and full-text queries

The need for a convergence between semi-structured data management and Information Retrieval techniques is manifest to the scientific community. In order to fulfil this growing request, W3C has recently proposed XQuery Full Text, an IR-oriented extension of XQuery. However, the issue of query optimization requires the study of important properties like query equivalence and containment; to this aim, a formal representation of document and queries is needed. The goal of this thesis is to establish such formal background. We define a data model for XML documents and propose an algebra able to represent most of XQuery Full-Text expressions. We show how an XQuery Full-Text expression can be translated into an algebraic expression and how an algebraic expression can be optimized.

Integrating structured and unstructured data in a business-intelligence-system

The recent liberalization of the German energy market has forced the energy industry to develop and install new information systems to support agents on the energy trading floors in their analytical tasks. Besides classical approaches of building a data warehouse giving insight into the time series to understand market and pricing mechanisms, it is crucial to provide a variety of external data from the web. Weather information as well as political news or market rumors are relevant to give the appropriate interpretation to the variables of a volatile energy market. Starting from a multidimensional data model and a collection of buy and sell transactions a data warehouse is built that gives analytical support to the agents. Following the idea of web farming we harvest the web, match the external information sources after a filtering and evaluation process to the data warehouse objects, and present this qualified information on a user interface where market values are correlated with those external sources over the time axis.