Using domain-specific knowledge to improve information retrieval performance

Mémoire numérisé par la Direction des bibliothèques de l'Université de Montréal.

A Generic Model of Execution for Synthesizing Domain-Specific Models

Software engineering researchers are challenged to provide increasingly more pow- erful levels of abstractions to address the rising complexity inherent in software solu- tions. One new development paradigm that places models as abstraction at the fore- front of the development process is Model-Driven Software Development (MDSD). MDSD considers models as first class artifacts, extending the capability for engineers to use concepts from the problem domain of discourse to specify apropos solutions. A key component in MDSD is domain-specific modeling languages (DSMLs) which are languages with focused expressiveness, targeting a specific taxonomy of problems. The de facto approach used is to first transform DSML models to an intermediate artifact in a HLL e.g., Java or C++, then execute that resulting code. Our research group has developed a class of DSMLs, referred to as interpreted DSMLs (i-DSMLs), where models are directly interpreted by a specialized execution engine with semantics based on model changes at runtime. This execution engine uses a layered architecture and is referred to as a domain-specific virtual machine (DSVM). As the domain-specific model being executed descends the layers of the DSVM the semantic gap between the user-defined model and the services being provided by the underlying infrastructure is closed. The focus of this research is the synthesis engine, the layer in the DSVM which transforms i-DSML models into executable scripts for the next lower layer to process. The appeal of an i-DSML is constrained as it possesses unique semantics contained within the DSVM. Existing DSVMs for i-DSMLs exhibit tight coupling between the implicit model of execution and the semantics of the domain, making it difficult to develop DSVMs for new i-DSMLs without a significant investment in resources. At the onset of this research only one i-DSML had been created for the user- centric communication domain using the aforementioned approach. This i-DSML is the Communication Modeling Language (CML) and its DSVM is the Communication Virtual machine (CVM). A major problem with the CVM’s synthesis engine is that the domain-specific knowledge (DSK) and the model of execution (MoE) are tightly interwoven consequently subsequent DSVMs would need to be developed from inception with no reuse of expertise. This dissertation investigates how to decouple the DSK from the MoE and sub- sequently producing a generic model of execution (GMoE) from the remaining appli- cation logic. This GMoE can be reused to instantiate synthesis engines for DSVMs in other domains. The generalized approach to developing the model synthesis com- ponent of i-DSML interpreters utilizes a reusable framework loosely coupled to DSK as swappable framework extensions. This approach involves first creating an i-DSML and its DSVM for a second do- main, demand-side smartgrid, or microgrid energy management, and designing the synthesis engine so that the DSK and MoE are easily decoupled. To validate the utility of the approach, the SEs are instantiated using the GMoE and DSKs of the two aforementioned domains and an empirical study to support our claim of reduced developmental effort is performed.

Independent component analysis for domain independent watermarking

A new principled domain independent watermarking framework is presented. The new approach is based on embedding the message in statistically independent sources of the covertext to mimimise covertext distortion, maximise the information embedding rate and improve the method's robustness against various attacks. Experiments comparing the performance of the new approach, on several standard attacks show the current proposed approach to be competitive with other state of the art domain-specific methods.

Modular average case analysis: Language implementation and extension

Motivated by accurate average-case analysis, MOdular Quantitative Analysis (MOQA) is developed at the Centre for Efficiency Oriented Languages (CEOL). In essence, MOQA allows the programmer to determine the average running time of a broad class of programmes directly from the code in a (semi-)automated way. The MOQA approach has the property of randomness preservation which means that applying any operation to a random structure, results in an output isomorphic to one or more random structures, which is key to systematic timing. Based on original MOQA research, we discuss the design and implementation of a new domain specific scripting language based on randomness preserving operations and random structures. It is designed to facilitate compositional timing by systematically tracking the distributions of inputs and outputs. The notion of a labelled partial order (LPO) is the basic data type in the language. The programmer uses built-in MOQA operations together with restricted control flow statements to design MOQA programs. This MOQA language is formally specified both syntactically and semantically in this thesis. A practical language interpreter implementation is provided and discussed. By analysing new algorithms and data restructuring operations, we demonstrate the wide applicability of the MOQA approach. Also we extend MOQA theory to a number of other domains besides average-case analysis. We show the strong connection between MOQA and parallel computing, reversible computing and data entropy analysis.

Comparison of Wavelet Transform and Time-Domain Analysis of Second Trimester Uterine Artery Doppler Waveforms in Screening for Pre-Eclampsia

The aim of this study was to compare time-domain waveform analysis of second-trimester uterine artery Doppler using the resistance index (RI) with waveform analysis using a mathematical tool known as wavelet transform for the prediction of pre-eclampsia (PE). This was a retrospective, nested case-cohort study of 336 women, 37 of whom subsequently developed PE. Uterine artery Doppler waveforms were analysed using both RI and waveform analysis. The utility of these indices in screening for PE was then evaluated using receiver operating characteristic curves. There were significant differences in uterine artery RI between the PE women and those with normal pregnancy outcome. After wavelet analysis, significant difference in the mean amplitude in wavelet frequency band 4 was noted between the 2 groups. The sensitivity for both Doppler RI and frequency band 4 for the detection of PE at a 10% false-positive rate was 45%. This small study demonstrates the application of wavelet transform analysis of uterine artery Doppler waveforms in screening for PE. Further prospective studies are needed in order to clearly define if this analytical approach to waveform analysis may have the potential to improve the detection of PE by uterine artery Doppler screening.

Association Study of 167 Candidate Genes for Schizophrenia Selected by a Multi-Domain Evidence-Based Prioritization Algorithm and Neurodevelopmental Hypothesis

Integrating evidence from multiple domains is useful in prioritizing disease candidate genes for subsequent testing. We ranked all known human genes (n = 3819) under linkage peaks in the Irish Study of High-Density Schizophrenia Families using three different evidence domains: 1) a meta-analysis of microarray gene expression results using the Stanley Brain collection, 2) a schizophrenia protein-protein interaction network, and 3) a systematic literature search. Each gene was assigned a domain-specific p-value and ranked after evaluating the evidence within each domain. For comparison to this
ranking process, a large-scale candidate gene hypothesis was also tested by including genes with Gene Ontology terms related to neurodevelopment. Subsequently, genotypes of 3725 SNPs in 167 genes from a custom Illumina iSelect array were used to evaluate the top ranked vs. hypothesis selected genes. Seventy-three genes were both highly ranked and involved in neurodevelopment (category 1) while 42 and 52 genes were exclusive to neurodevelopment (category 2) or highly ranked (category 3), respectively. The most significant associations were observed in genes PRKG1, PRKCE, and CNTN4 but no individual SNPs were significant after correction for multiple testing. Comparison of the approaches showed an excess of significant tests using the hypothesis-driven neurodevelopment category. Random selection of similar sized genes from two independent genome-wide association studies (GWAS) of schizophrenia showed the excess was unlikely by chance. In a further meta-analysis of three GWAS datasets, four candidate SNPs reached nominal significance. Although gene ranking using integrated sources of prior information did not enrich for significant results in the current experiment, gene selection using an a priori hypothesis (neurodevelopment) was superior to random selection. As such, further development of gene ranking strategies using more carefully selected sources of information is warranted.

Proposal of a simplified process to correct the phase decoupling using modal analysis

Modal analysis is widely approached in the classic theory of transmission line modeling. This technique is applied to model the three-phase representation of conventional electric systems taking into account their self and mutual electrical parameters. However the methodology has some particularities and inaccuracies for specific applications which are not clearly described in the basic references of this topic. This paper provides a thorough review of modal analysis theory applied to line models followed by an original and simple procedure to overcome the possible errors embedded in the modal decoupling through the three-phase system modeling. © 2012 IEEE.

Domains of physical activity and all-cause mortality: systematic review and dose-response meta-analysis of cohort studies

Background The dose–response relation between physical activity and all-cause mortality is not well defined at present. We conducted a systematic review and meta-analysis to determine the association with all-cause mortality of different domains of physical activity and of defined increases in physical activity and energy expenditure. Methods MEDLINE, Embase and the Cochrane Library were searched up to September 2010 for cohort studies examining all-cause mortality across different domains and levels of physical activity in adult general populations. We estimated combined risk ratios (RRs) associated with defined increments and recommended levels, using random-effects meta-analysis and dose–response meta-regression models. Results Data from 80 studies with 1 338 143 participants (118 121 deaths) were included. Combined RRs comparing highest with lowest activity levels were 0.65 [95% confidence interval (95% CI) 0.60–0.71] for total activity, 0.74 (95% CI 0.70–0.77) for leisure activity, 0.64 (95% CI 0.55–0.75) for activities of daily living and 0.83 (95% CI 0.71–0.97) for occupational activity. RRs per 1-h increment per week were 0.91 (95% CI 0.87–0.94) for vigorous exercise and 0.96 (95% CI 0.93–0.98) for moderate-intensity activities of daily living. RRs corresponding to 150 and 300 min/week of moderate to vigorous activity were 0.86 (95% CI 0.80–0.92) and 0.74 (95% CI 0.65–0.85), respectively. Mortality reductions were more pronounced in women. Conclusion Higher levels of total and domain-specific physical activity were associated with reduced all-cause mortality. Risk reduction per unit of time increase was largest for vigorous exercise. Moderate-intensity activities of daily living were to a lesser extent beneficial in reducing mortality.

Taxon-specific δ13C analysis of chitinous invertebrate remains in sediments from Strandsjön, Sweden

Taxon-specific stable carbon isotope (δ13C) analysis of chitinous remains of invertebrates can provide valuable information about the carbon sources used by invertebrates living in specific habitats of lake ecosystems (for example, sediments, water column, or aquatic vegetation). This is complementary to δ13C of sedimentary organic matter (SOM), which provides an integrated signal of organic matter produced in a lake and its catchment, and of diagenetic processes within sediments. In a sediment record from Strandsjön (Sweden) covering the past circa 140 years, we analyzed SOM geochemistry (δ13C, C:Natomic, organic carbon content) and δ13C of chitinous invertebrate remains in order to examine whether taxon-specific δ13C records could be developed for different invertebrate groups and whether these analyses provide insights into past changes of organic carbon sources for lacustrine invertebrates available in benthic and planktonic compartments of the lake. Invertebrate taxa included benthic chironomids (Chironomus, Chironomini excluding Chironomus, Tanytarsini, and Tanypodinae), filter-feeders on suspended particulate organic matter (Daphnia, Plumatella and Cristatella mucedo), and Rhabdocoela. δ13C of chironomid remains indicated periodic availability of 13C-depleted carbon sources in the benthic environment of the lake as δ13C values of the different chironomid taxa fluctuated simultaneously between -34.7 and -30.5‰ (VPDB). Daphnia and Bryozoa showed parallel changes in their δ13C values which did not coincide with variations in δ13C of chironomids, though, and a 2-3‰ decrease since circa AD 1960. The decrease in δ13C of Daphnia and Bryozoa could indicate a decrease in phytoplankton δ13C as a result of lower lake productivity, which is in accordance with historical information about the lake that suggests a shift to less eutrophic conditions after AD 1960. In contrast, Rhabdocoela cocoons were characterized by relatively high δ13C values (-30.4 to -28.2‰) that did not show a strong temporal trend, which could be related to the predatory feeding mode and wide prey spectrum of this organism group. The taxon-specific δ13C analyses of invertebrate remains indicated that different carbon sources were available for the benthic chironomid larvae than for the filter-feeding Daphnia and bryozoans. Our results therefore demonstrate that taxon-specific analysis of δ13C of organic invertebrate remains can provide complementary information to measurements on bulk SOM and that δ13C of invertebrate remains may allow the reconstruction of past changes in carbon sources and their δ13C in different habitats of lake ecosystems.

GENETIC ANALYSIS OF THE HIPPO PATHWAY IN MOUSE LIVER

Cancer therapy and tumor treatment remain unsolved puzzles. Genetic screening for tumor suppressor genes in Drosophila revealed the Hippo-signaling pathway as a kinase cascade consisting of five core components. Disrupting the pathway by deleting the main component genes breaks the balance of cell proliferation and apoptosis and results in epithelial tissue tumorigenesis. The pathway is therefore believed to be a tumor suppressor pathway. However, a corresponding role in mammals is yet to be determined. Our lab began to investigate the tumor suppression function of the potent mammalian Hippo pathway by putting floxed alleles into the mouse genome flanking the functional-domain-expressing exons in each component (Mst1, Mst2, Sav1, Lats1 and Lats2). These mice were then crossed with different cre-mouse lines to generate conditional knockout mice. Results indicate a ubiquitous tumor suppression function of these components, predominantly in the liver. A further liver specific analysis of the deletion mutation of these components, as well as the Yap/Taz double deletion mutation, reveals essential roles of the Hippo pathway in regulating hepatic quiescence and embryonic liver development. One of the key cellular mechanisms for the Hippo pathway’s involvement in these liver biological events is likely its cell cycle regulation function. Our work will help to develop potential therapeutic approaches for liver cancer.

Automated attribute inference in complex service workflows based on sharing analysis

The properties of data and activities in business processes can be used to greatly facilítate several relevant tasks performed at design- and run-time, such as fragmentation, compliance checking, or top-down design. Business processes are often described using workflows. We present an approach for mechanically inferring business domain-specific attributes of workflow components (including data Ítems, activities, and elements of sub-workflows), taking as starting point known attributes of workflow inputs and the structure of the workflow. We achieve this by modeling these components as concepts and applying sharing analysis to a Horn clause-based representation of the workflow. The analysis is applicable to workflows featuring complex control and data dependencies, embedded control constructs, such as loops and branches, and embedded component services.

Semantic neology in the domain of videogames in Spanish

The aim of this paper is to discuss the meaning of five neologisms in the domain of videogames in Spanish: título, aventura, personaje, plataforma, and rol. Our study focuses on a special type of neologism since the Spanish terms we deal with here are not strictly new words; they are what have been called sense neologisms or neosemanticisms, that is, old words taking a new sense in a different domain. These words were identified as new concepts after a process of analysis based on contextual evidence. This study of neology is based on the analysis of a corpus of press articles evaluating videogames published by the Spanish newspaper El País from 1998 to 2008. The analysis of the instances of use of domain specific terms in the corpus revealed that they acquired new senses different to those they have in other domains where they are also used. The paper explains the process of discovering the specialized meaning these words have developed in the domain of videogames and how the analysis of collocational behavior helps in the process of discovering the new sense and in the design of the definition provided.

Semantic Neology in the Domain of Videogames in Spanish.

The aim of this paper is to discuss the meaning of five neologisms in the domain of videogames in Spanish: título, aventura, personaje, plataforma, and rol. Our study focuses on a special type of neologism since the Spanish terms we deal with here are not strictly new words; they are what have been called sense neologisms or neosemanticisms, that is, old words taking a new sense in a different domain. These words were identified as new concepts after a process of analysis based on contextual evidence. This study of neology is based on the analysis of a corpus of press articles evaluating videogames published by the Spanish newspaper El País from 1998 to 2008. The analysis of the instances of use of domain specific terms in the corpus revealed that they acquired new senses different to those they have in other domains where they are also used. The paper explains the process of discovering the specialized meaning these words have developed in the domain of videogames and how the analysis of collocational behavior helps in the process of discovering the new sense and in the design of the definition provided. RESUMEN: En este trabajo se presentan cinco neologismos del ámbito del videojuego en español: “título”, “aventura”, “personaje”, “plataforma” y “rol”. Se trata de un tipo especial de neologismo, conocido también como “neologismo semántico” o “neosemanticismo”, ya que son palabras ya existentes en la lengua que adquieren un nuevo significado. Los nuevos significados que adquieren estos términos en el ámbito del videojuego se establecieron tras el análisis del contexto de uso en un corpus periodístico de críticas de videojuegos. Este corpus recoge las críticas de videojuegos publicadas por el periódico El País entre 1998 y 2008. El análisis de los casos de uso de los términos en el corpus de videojuegos reveló que adquirían un nuevo significado diferente al de su uso en otros ámbitos o en el lenguaje general. El artículo describe cada uno de los neologismos y el proceso de análisis contextual que conduce a descubrir el nuevo significado y elaborar su definición.

Self-training from labeled features for sentiment analysis

Sentiment analysis concerns about automatically identifying sentiment or opinion expressed in a given piece of text. Most prior work either use prior lexical knowledge defined as sentiment polarity of words or view the task as a text classification problem and rely on labeled corpora to train a sentiment classifier. While lexicon-based approaches do not adapt well to different domains, corpus-based approaches require expensive manual annotation effort. In this paper, we propose a novel framework where an initial classifier is learned by incorporating prior information extracted from an existing sentiment lexicon with preferences on expectations of sentiment labels of those lexicon words being expressed using generalized expectation criteria. Documents classified with high confidence are then used as pseudo-labeled examples for automatical domain-specific feature acquisition. The word-class distributions of such self-learned features are estimated from the pseudo-labeled examples and are used to train another classifier by constraining the model's predictions on unlabeled instances. Experiments on both the movie-review data and the multi-domain sentiment dataset show that our approach attains comparable or better performance than existing weakly-supervised sentiment classification methods despite using no labeled documents.