Biomedical events extraction using the hidden vector state model

Objective: Biomedical events extraction concerns about events describing changes on the state of bio-molecules from literature. Comparing to the protein-protein interactions (PPIs) extraction task which often only involves the extraction of binary relations between two proteins, biomedical events extraction is much harder since it needs to deal with complex events consisting of embedded or hierarchical relations among proteins, events, and their textual triggers. In this paper, we propose an information extraction system based on the hidden vector state (HVS) model, called HVS-BioEvent, for biomedical events extraction, and investigate its capability in extracting complex events. Methods and material: HVS has been previously employed for extracting PPIs. In HVS-BioEvent, we propose an automated way to generate abstract annotations for HVS training and further propose novel machine learning approaches for event trigger words identification, and for biomedical events extraction from the HVS parse results. Results: Our proposed system achieves an F-score of 49.57% on the corpus used in the BioNLP'09 shared task, which is only 2.38% lower than the best performing system by UTurku in the BioNLP'09 shared task. Nevertheless, HVS-BioEvent outperforms UTurku's system on complex events extraction with 36.57% vs. 30.52% being achieved for extracting regulation events, and 40.61% vs. 38.99% for negative regulation events. Conclusions: The results suggest that the HVS model with the hierarchical hidden state structure is indeed more suitable for complex event extraction since it could naturally model embedded structural context in sentences.

Extracting protein-protein interactions from MEDLINE using the hidden vector state model

A major challenge in text mining for biomedicine is automatically extracting protein-protein interactions from the vast amount of biomedical literature. We have constructed an information extraction system based on the Hidden Vector State (HVS) model for protein-protein interactions. The HVS model can be trained using only lightly annotated data whilst simultaneously retaining sufficient ability to capture the hierarchical structure. When applied in extracting protein-protein interactions, we found that it performed better than other established statistical methods and achieved 61.5% in F-score with balanced recall and precision values. Moreover, the statistical nature of the pure data-driven HVS model makes it intrinsically robust and it can be easily adapted to other domains.

Discriminative training of the hidden vector state model for semantic parsing

In this paper, we discuss how discriminative training can be applied to the hidden vector state (HVS) model in different task domains. The HVS model is a discrete hidden Markov model (HMM) in which each HMM state represents the state of a push-down automaton with a finite stack size. In previous applications, maximum-likelihood estimation (MLE) is used to derive the parameters of the HVS model. However, MLE makes a number of assumptions and unfortunately some of these assumptions do not hold. Discriminative training, without making such assumptions, can improve the performance of the HVS model by discriminating the correct hypothesis from the competing hypotheses. Experiments have been conducted in two domains: the travel domain for the semantic parsing task using the DARPA Communicator data and the Air Travel Information Services (ATIS) data and the bioinformatics domain for the information extraction task using the GENIA corpus. The results demonstrate modest improvements of the performance of the HVS model using discriminative training. In the travel domain, discriminative training of the HVS model gives a relative error reduction rate of 31 percent in F-measure when compared with MLE on the DARPA Communicator data and 9 percent on the ATIS data. In the bioinformatics domain, a relative error reduction rate of 4 percent in F-measure is achieved on the GENIA corpus.

Ontology-based protein-protein interactions extraction from literature using the hidden vector state model

This paper proposes a novel framework of incorporating protein-protein interactions (PPI) ontology knowledge into PPI extraction from biomedical literature in order to address the emerging challenges of deep natural language understanding. It is built upon the existing work on relation extraction using the Hidden Vector State (HVS) model. The HVS model belongs to the category of statistical learning methods. It can be trained directly from un-annotated data in a constrained way whilst at the same time being able to capture the underlying named entity relationships. However, it is difficult to incorporate background knowledge or non-local information into the HVS model. This paper proposes to represent the HVS model as a conditionally trained undirected graphical model in which non-local features derived from PPI ontology through inference would be easily incorporated. The seamless fusion of ontology inference with statistical learning produces a new paradigm to information extraction.

Modelling financial time series with switching state space models

The deficiencies of stationary models applied to financial time series are well documented. A special form of non-stationarity, where the underlying generator switches between (approximately) stationary regimes, seems particularly appropriate for financial markets. We use a dynamic switching (modelled by a hidden Markov model) combined with a linear dynamical system in a hybrid switching state space model (SSSM) and discuss the practical details of training such models with a variational EM algorithm due to [Ghahramani and Hilton,1998]. The performance of the SSSM is evaluated on several financial data sets and it is shown to improve on a number of existing benchmark methods.

A study of thermodynamic characteristics and predictive computer modelling of an air to water heat pump system

A study on heat pump thermodynamic characteristics has been made in the laboratory on a specially designed and instrumented air to water heat pump system. The design, using refrigerant R12, was based on the requirement to produce domestic hot water at a temperature of about 50 °C and was assembled in the laboratory. All the experimental data were fed to a microcomputer and stored on disk automatically from appropriate transducers via amplifier and 16 channel analogue to digital converters. The measurements taken were R12 pressures and temperatures, water and R12 mass flow rates, air speed, fan and compressor input powers, water and air inlet and outlet temperatures, wet and dry bulb temperatures. The time interval between the observations could be varied. The results showed, as expected, that the COP was higher at higher air inlet temperatures and at lower hot water output temperatures. The optimum air speed was found to be at a speed when the fan input power was about 4% of the condenser heat output. It was also found that the hot water can be produced at a temperature higher than the appropriate R12 condensing temperature corresponding to condensing pressure. This was achieved by condenser design to take advantage of discharge superheat and by further heating the water using heat recovery from the compressor. Of the input power to the compressor, typically about 85% was transferred to the refrigerant, 50 % by the compression work and 35% due to the heating of the refrigerant by the cylinder wall, and the remaining 15% (of the input power) was rejected to the cooling medium. The evaporator effectiveness was found to be about 75% and sensitive to the air speed. Using the data collected, a steady state computer model was developed. For given input conditions s air inlet temperature, air speed, the degree of suction superheat , water inlet and outlet temperatures; the model is capable of predicting the refrigerant cycle, compressor efficiency, evaporator effectiveness, condenser water flow rate and system Cop.

A hybrid generative/discriminative framework to train a semantic parser from an un-annotated corpus

We propose a hybrid generative/discriminative framework for semantic parsing which combines the hidden vector state (HVS) model and the hidden Markov support vector machines (HM-SVMs). The HVS model is an extension of the basic discrete Markov model in which context is encoded as a stack-oriented state vector. The HM-SVMs combine the advantages of the hidden Markov models and the support vector machines. By employing a modified K-means clustering method, a small set of most representative sentences can be automatically selected from an un-annotated corpus. These sentences together with their abstract annotations are used to train an HVS model which could be subsequently applied on the whole corpus to generate semantic parsing results. The most confident semantic parsing results are selected to generate a fully-annotated corpus which is used to train the HM-SVMs. The proposed framework has been tested on the DARPA Communicator Data. Experimental results show that an improvement over the baseline HVS parser has been observed using the hybrid framework. When compared with the HM-SVMs trained from the fully-annotated corpus, the hybrid framework gave a comparable performance with only a small set of lightly annotated sentences. © 2008. Licensed under the Creative Commons.

Modelling and simulation of biomass gasification in a circulating fluidized bed reactor

Computational Fluid Dynamics (CFD) has found great acceptance among the engineering community as a tool for research and design of processes that are practically difficult or expensive to study experimentally. One of these processes is the biomass gasification in a Circulating Fluidized Bed (CFB). Biomass gasification is the thermo-chemical conversion of biomass at a high temperature and a controlled oxygen amount into fuel gas, also sometime referred to as syngas. Circulating fluidized bed is a type of reactor in which it is possible to maintain a stable and continuous circulation of solids in a gas-solid system. The main objectives of this thesis are four folds: (i) Develop a three-dimensional predictive model of biomass gasification in a CFB riser using advanced Computational Fluid Dynamic (CFD) (ii) Experimentally validate the developed hydrodynamic model using conventional and advanced measuring techniques (iii) Study the complex hydrodynamics, heat transfer and reaction kinetics through modelling and simulation (iv) Study the CFB gasifier performance through parametric analysis and identify the optimum operating condition to maximize the product gas quality. Two different and complimentary experimental techniques were used to validate the hydrodynamic model, namely pressure measurement and particle tracking. The pressure measurement is a very common and widely used technique in fluidized bed studies, while, particle tracking using PEPT, which was originally developed for medical imaging, is a relatively new technique in the engineering field. It is relatively expensive and only available at few research centres around the world. This study started with a simple poly-dispersed single solid phase then moved to binary solid phases. The single solid phase was used for primary validations and eliminating unnecessary options and steps in building the hydrodynamic model. Then the outcomes from the primary validations were applied to the secondary validations of the binary mixture to avoid time consuming computations. Studies on binary solid mixture hydrodynamics is rarely reported in the literature. In this study the binary solid mixture was modelled and validated using experimental data from the both techniques mentioned above. Good agreement was achieved with the both techniques. According to the general gasification steps the developed model has been separated into three main gasification stages; drying, devolatilization and tar cracking, and partial combustion and gasification. The drying was modelled as a mass transfer from the solid phase to the gas phase. The devolatilization and tar cracking model consist of two steps; the devolatilization of the biomass which is used as a single reaction to generate the biomass gases from the volatile materials and tar cracking. The latter is also modelled as one reaction to generate gases with fixed mass fractions. The first reaction was classified as a heterogeneous reaction while the second reaction was classified as homogenous reaction. The partial combustion and gasification model consisted of carbon combustion reactions and carbon and gas phase reactions. The partial combustion considered was for C, CO, H2 and CH4. The carbon gasification reactions used in this study is the Boudouard reaction with CO2, the reaction with H2O and Methanation (Methane forming reaction) reaction to generate methane. The other gas phase reactions considered in this study are the water gas shift reaction, which is modelled as a reversible reaction and the methane steam reforming reaction. The developed gasification model was validated using different experimental data from the literature and for a wide range of operating conditions. Good agreement was observed, thus confirming the capability of the model in predicting biomass gasification in a CFB to a great accuracy. The developed model has been successfully used to carry out sensitivity and parametric analysis. The sensitivity analysis included: study of the effect of inclusion of various combustion reaction; and the effect of radiation in the gasification reaction. The developed model was also used to carry out parametric analysis by changing the following gasifier operating conditions: fuel/air ratio; biomass flow rates; sand (heat carrier) temperatures; sand flow rates; sand and biomass particle sizes; gasifying agent (pure air or pure steam); pyrolysis models used; steam/biomass ratio. Finally, based on these parametric and sensitivity analysis a final model was recommended for the simulation of biomass gasification in a CFB riser.

Semi-supervised learning of statistical models for natural language understanding

Natural language understanding is to specify a computational model that maps sentences to their semantic mean representation. In this paper, we propose a novel framework to train the statistical models without using expensive fully annotated data. In particular, the input of our framework is a set of sentences labeled with abstract semantic annotations. These annotations encode the underlying embedded semantic structural relations without explicit word/semantic tag alignment. The proposed framework can automatically induce derivation rules that map sentences to their semantic meaning representations. The learning framework is applied on two statistical models, the conditional random fields (CRFs) and the hidden Markov support vector machines (HM-SVMs). Our experimental results on the DARPA communicator data show that both CRFs and HM-SVMs outperform the baseline approach, previously proposed hidden vector state (HVS) model which is also trained on abstract semantic annotations. In addition, the proposed framework shows superior performance than two other baseline approaches, a hybrid framework combining HVS and HM-SVMs and discriminative training of HVS, with a relative error reduction rate of about 25% and 15% being achieved in F-measure.

A randomised controlled trial of the clinical effectiveness and cost-effectiveness of the levonorgestrel-releasing intrauterine system in primary care against standard treatment for menorrhagia:the ECLIPSE trial

BACKGROUND: Heavy menstrual bleeding (HMB) is a common problem, yet evidence to inform decisions about initial medical treatment is limited. OBJECTIVES: To assess the clinical effectiveness and cost-effectiveness of the levonorgestrel-releasing intrauterine system (LNG-IUS) (Mirena(®), Bayer) compared with usual medical treatment, with exploration of women's perspectives on treatment. DESIGN: A pragmatic, multicentre randomised trial with an economic evaluation and a longitudinal qualitative study. SETTING: Women who presented in primary care. PARTICIPANTS: A total of 571 women with HMB. A purposeful sample of 27 women who were randomised or ineligible owing to treatment preference participated in semistructured face-to-face interviews around 2 and 12 months after commencing treatment. INTERVENTIONS: LNG-IUS or usual medical treatment (tranexamic acid, mefenamic acid, combined oestrogen-progestogen or progesterone alone). Women could subsequently swap or cease their allocated treatment. OUTCOME MEASURES: The primary outcome was the patient-reported score on the Menorrhagia Multi-Attribute Scale (MMAS) assessed over a 2-year period and then again at 5 years. Secondary outcomes included general quality of life (QoL), sexual activity, surgical intervention and safety. Data were analysed using iterative constant comparison. A state transition model-based cost-utility analysis was undertaken alongside the randomised trial. Quality-adjusted life-years (QALYs) were derived from the European Quality of Life-5 Dimensions (EQ-5D) and the Short Form questionnaire-6 Dimensions (SF-6D). The intention-to-treat analyses were reported as cost per QALY gained. Uncertainty was explored by conducting both deterministic and probabilistic sensitivity analyses. RESULTS: The MMAS total scores improved significantly in both groups at all time points, but were significantly greater for the LNG-IUS than for usual treatment [mean difference over 2 years was 13.4 points, 95% confidence interval (CI) 9.9 to 16.9 points; p < 0.001]. However, this difference between groups was reduced and no longer significant by 5 years (mean difference in scores 3.9 points, 95% CI -0.6 to 8.3 points; p = 0.09). By 5 years, only 47% of women had a LNG-IUS in place and 15% were still taking usual medical treatment. Five-year surgery rates were low, at 20%, and were similar, irrespective of initial treatments. There were no significant differences in serious adverse events between groups. Using the EQ-5D, at 2 years, the relative cost-effectiveness of the LNG-IUS compared with usual medical treatment was £1600 per QALY, which by 5 years was reduced to £114 per QALY. Using the SF-6D, usual medical treatment dominates the LNG-IUS. The qualitative findings show that women's experiences and expectations of medical treatments for HMB vary considerably and change over time. Women had high expectations of a prompt effect from medical treatments. CONCLUSIONS: The LNG-IUS, compared with usual medical therapies, resulted in greater improvement over 2 years in women's assessments of the effect of HMB on their daily routine, including work, social and family life, and psychological and physical well-being. At 5 years, the differences were no longer significant. A similar low proportion of women required surgical intervention in both groups. The LNG-IUS is cost-effective in both the short and medium term, using the method generally recommended by the National Institute for Health and Care Excellence. Using the alternative measures to value QoL will have a considerable impact on cost-effectiveness decisions. It will be important to explore the clinical and health-care trajectories of the ECLIPSE (clinical effectiveness and cost-effectiveness of levonorgestrel-releasing intrauterine system in primary care against standard treatment for menorrhagia) trial participants to 10 years, by which time half of the cohort will have reached menopause. TRIAL REGISTRATION: Current Controlled Trials ISRCTN86566246. FUNDING: This project was funded by the NIHR Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 19, No. 88. See the NIHR Journals Library website for further project information.

Large-scale data for multiple-view stereopsis

The seminal multiple-view stereo benchmark evaluations from Middlebury and by Strecha et al. have played a major role in propelling the development of multi-view stereopsis (MVS) methodology. The somewhat small size and variability of these data sets, however, limit their scope and the conclusions that can be derived from them. To facilitate further development within MVS, we here present a new and varied data set consisting of 80 scenes, seen from 49 or 64 accurate camera positions. This is accompanied by accurate structured light scans for reference and evaluation. In addition all images are taken under seven different lighting conditions. As a benchmark and to validate the use of our data set for obtaining reasonable and statistically significant findings about MVS, we have applied the three state-of-the-art MVS algorithms by Campbell et al., Furukawa et al., and Tola et al. to the data set. To do this we have extended the evaluation protocol from the Middlebury evaluation, necessitated by the more complex geometry of some of our scenes. The data set and accompanying evaluation framework are made freely available online. Based on this evaluation, we are able to observe several characteristics of state-of-the-art MVS, e.g. that there is a tradeoff between the quality of the reconstructed 3D points (accuracy) and how much of an object’s surface is captured (completeness). Also, several issues that we hypothesized would challenge MVS, such as specularities and changing lighting conditions did not pose serious problems. Our study finds that the two most pressing issues for MVS are lack of texture and meshing (forming 3D points into closed triangulated surfaces).