Multi-Agent Based Peer-to-Peer
Workflow Management System
Multi-Agent
Based Peer-to-Peer Workflow Management System (Download
Poster .doc)
Fuzzy
Regression Trees (Download
Poster .pdf) Decision trees are constructed by learning and reasoning from feature-based examples. Decision tree algorithms induce decision trees in which each branch node represents a choice between a number of alternatives, and each leaf node represents a decision. A tree is constructed to classify instances by traversing them through the tree from the root to some of the leaf nodes. Classification and regression trees are two models of decision trees. Classification trees are used to predict classification outcomes while regression trees predict outcomes which are continuous. There are numerous algorithms which describe the approach of building decision trees based on information theory, such as ID3[3], C4.5[2] and C5[8] which can be used for classification whilst AID [1] and CHAID[5] are well suited to use for prediction with regression tree. Classification and regression tree (CART) is used to refer to both above models [9,7]. Decision tree algorithms induce trees with sharp decision boundaries, which clearly impacts decisions made when the attribute has continuous value. This phenomenon is the most important weakness of decision tree algorithms such as ID3[6]. Improvement in performance using decision trees have been made by applying fuzzy logic [4]. Fuzzification is applied to traditional classification decision trees in order to soften sharp decision boundaries and fuzzy inference is used to calculate a final decision outcome. This research will concentrate on investigating and developing a framework for fuzzy regression trees, and then extend the concept to include fuzzy regression tree forests using time series data. The fuzzy decision forests will then be optimised using a suitable approach. This poster describes initial research on the investigation and development of a novel approach for combining the CHAID regression decision tree algorithm and the principles of fuzzy theory to produce fuzzy regression trees. References [1] J. N. Morgan and J. A. Sonquist,(1963), Problems in the Analysis of Survey Data and a Proposal, American Statistical Association Journal 58, pp 415-434. [2] J. Ross Quinlan, (1993), C4.5 program for machine learning, Morgan Kaufmann Publishrs [3] J. Ross Quinlan, (1986). Induction of Decision Trees, Machine Learning, 1, pp. 81-106 [4] Jay Fowdar, Keeley Crockett, Zuhair Bandar, James O’Shea, (2005), On the Use of Fuzzy Trees for Solving Classification Problems with Numeric Outcomes, IEEE International Conference on Fuzzy Systems, pp. 436-441. [5] V. G. Kass, (1979), An Exploratory Technique For Investigating Large Quantities Of Categorical Data, Applied Statistics, 29(2) pp 119-127. [6] Keeley A. Crockett, Zuhair Bandar, David McLean, James O'Shea (2006), On constructing a fuzzy inference framework using crisp decision trees, Fuzzy Sets and Systems 157(21),pp 2809-2832. [7] Leland Wilkinson, (1992), Tree Structured Data Analysis: AID, CHAID and CART. Sun Valley, ID, Sawtooth/SYSTAT Joint Software Conference. [8] J. Ross Quinlan, (2002) Is C5.0 better than C4.5; available at ttp://www.rulequest.com/see5-comparison html. [9] Servane Gey and Elodie Nedelec, (2005), Model Selection for CART Regression Trees, IEEE Transactions on Information Theory, 51, pp. 658-670.
An FTIR study of Reaction Kinetics and Structure Development in
Flexible Polyurethane Foam-Layered Silicate Nanocomposites (Download
Poster .ppt) Polymer-layered silicate nanocomposites (PLSNs) offer the potential for significant enhancement in mechanical properties and thermal stabilities, and reductions in permeability and flammability, resulting from incorporation of only a few weight percent (wt%) of a nanoscale lamellar filler into a polymer. There have been a number of studies on PLSN foams based on solid polyurethane (PU), but relatively few on cellular PU/layered silicate nanocomposites. Adiabatic temperature rise (ATR) measurement and forced-adiabatic Fourier transform-infrared (FTIR) spectroscopy were used to study the reaction kinetics and structure development in PLSN foams. The MMTs used in this study were Cloisite® Na+ (CNa+, sodium montmorillonite) and Cloisite® 30B (C30B, montmorillonite modified with methyl tallow bis-2-hydroxyethyl ammonium chloride). The polyurethane foams studied were formed via the reaction of toluene diisocyanate (TDI) with a mixture of a polyether polyol (triol type) and de-ionized water acting as a chemical blowing agent. Seven foams were produced; an unfilled foam, PLSN foams based on (1, 3, 5 wt%) CNa+ and PLSN foams based on (1, 3, 5 wt%) C30B. Whilst consumption of TDI during the initial stage (t < 240 s) of the copolymerization was accelerated significantly by the addition of CNa+ and C30B, CNa+ has better effect than C30B in terms of increasing the rate of reaction. As the overall heat rise in these foaming system is expected to be dominate by the reaction of TDI with water (with a molar ratio of reactive groups, water/polyol=6.4), the water molecules associated with hydrated Na+ ions in CNa+ are more acidic and hence more readily accessible to the TDI. In contrast, hydroxyl groups present at the platelet edges of C30B may catalyze the reaction between TDI and water. The formation of intermediate (monodentate) urea within the urea groups of the hard-segment phase in PLSN foams was also accelerated as the initial rate of formation of urea groups increased upon addition of CNa+ and C30B. However, when the MMT content exceeds 3 wt%, the formation of hydrogen-bonded (bidentate) urea was hindered in PLSN foams based on CNa+ and was limited in PLSN foams based on C30B. This result may be reflected in the stronger interactions of urea groups with hydrated Na+ ions in CNa+ relative to those with C30B.
Can Automated
Tutors be Socially Intelligent? (Download
Poster .pdf) Intelligent tutoring systems (ITS) attempt to model and mimic the behaviour of human tutors in a computer-assisted learning environment in order to provide more personalised learning than previous content delivery systems. ITS have recently been extended to create more personalised learning by adapting the content shown to a student’s existing knowledge and learning style (determined by questionnaires and quizzes). Conversational agents (CAs) are computer programs which allow people to communicate with computers using natural language, and have been employed within ITS to help students learn to apply knowledge when discussing problems and formulating answers. CAs are in their infancy, and it is difficult for a computer program to understand the complexities of natural language, or to pick up on social signals in order to adapt their response to the conversant. CAs rely on manually coded scripts to direct their conversation, which are time consuming and require expertise to create. This research aims to investigate the detection of social cues (specifically a student’s learning style) during tutoring conversations with a CA, and to automatically adapt the tutoring style to suit the student and aid learning. A CA tutor will be developed which automatically estimates and adapts to a student’s learning style. Machine learning techniques will be employed to automate the adaptation, which will then be tested for a different subject.
An Approach to Conversational Agent Design using Sentence Similarity
Measures A ‘Conversational Agent’ (CA) is a program, which enables communication with a user through natural language. Traditionally, text-based CA’s operate using scripts consisting of a number of ‘rules’ organised into ‘contexts’ [1], [2]. Each context contains a number of hierarchically organised rules with each rule possessing a list of structural patterns of sentences and an associated response. A user’s utterance is then matched to a pattern and the associated response is ‘fired’ and sent as output. Writing scripts for such systems, however, possess many difficulties such as anticipating the inordinate number of ways to say the same thing [1]. The new proposed CA employs a sentence similarity measure [4] to interpret scripts consisting of natural language sentences. Through the use of the sentence similarity measure a match is determined between the user’s utterance and the natural language sentences. Considering semantics rather than structural patterns of sentences meant that scripting was reduced to a couple of natural language sentences per rule as opposed to potentially 100s of patterns. Furthermore, results indicate good levels of accuracy between user input and the scripted sentences [3]. Further work will entail expansion of the CA’s capabilities, such as developing adaptability and self-awareness. System robustness testing and an evaluation using human subjects will also be conducted. References [1] Sammut C, Managing Context in a Conversational Agent. Linkoping Electronic Articles in Computer and Information Science. Vol. 3 (7), 2001, pp.1-7. [2] Michie D. and Sammut C., Infochat Scripter’s Manual. Manchester: Convagent Ltd, 2001. [3] O’Shea K., Bandar Z., and Crockett K., A Novel Approach for Constructing Conversational Agents using Sentence Similarity Measures. World Congress on Engineering, International Conference on Data Mining and Knowledge Engineering, London, July 2008, pp. 321-326. [4] Li Y., McLean D., Bandar Z. A., O’Shea J. D., Crockett K., 2006. Sentence Similarity Based on Semantic Nets and Corpus Statistics. IEEE Transactions on Knowledge and Data Engineering. Vol. 18 (8) pp. 1138-1149.
Semantic similarity of a pair of Sentences using verbs and adjuncts.
(Download
Poster .doc) This poster will detail the common approaches to the task of comparing a pair of sentences using a computer. It details some of the important applications such as automatic e-mail response and chat room monitoring. It gives an overview of the semantic database (WordNet) and of the code to perform the comparison of a pair of sentences using nouns, verbs, adjectives and adverbs, (combining many of the rules of linguistics whilst avoiding the complications of Natural Language processing. It refers to how a human interprets written language dependent upon their experience as well as assumptions about the definitions of the language. It mentions some of the difficulties of benchmarking results in a non-standardised field, referencing the need to survey English speakers. A preview of initial results is also referenced. The poster contains proposals of future work where the model is refined to include more complex linguistic terms and ideas such as adjunct clauses and negatives. The final evaluation will include a domain testing using the algorithm to enhance automated e-mail response.
Inhibition of microorganisms associated with denture plaque by Streptococcus
salivarius strains K12, MIA and T32. Objectives: This study aimed to determine the activity of bacteriocin-like inhibitory substances (BLIS) produced by the probiotic Streptococcus salivarius strains K12, MIA and T32 on microorganisms associated with denture plaque and the oral cavity. These strains have proposed use in the prevention and treatment of pharyngitis and oral malodour, thus their potential role in denture-associated infection is being explored. Methods: The BLIS production of S. salivarius K12, MIA and T32 was established by simultaneous (spot & well diffusion) and deferred (pre-growth & physical separation) antagonism techniques against 15 streptococcal strains, 4 Candida albicans strains, and 4 obligate anaerobic bacteria. Results: S. salivarius K12 inhibited growth of 11/15 streptococcal strains including S. mutans, S. oralis and S. mitis; whilst S. salivarius MIA and T32 inhibited only 5/11 strains. No inhibitory activity against Candida albicans was observed. Inhibition of 3/4 anaerobes was observed in the well-diffusion simultaneous antagonism assay only. Conclusion: The S. salivarius strains K12, MIA and T32 produce detectable BLIS activity in vitro against bacteria associated with denture plaque and the oral cavity. The detectable inhibitory activity of S. salivarius was affected by the assay method. Results indicate an increased concentration of bacteriocin is needed to inhibit Gram-negative obligate anaerobes.
Factors affecting the colonisation of food processing surfaces by
Listeria monocytogenes. A major concern in the food industry is the contamination of preparation surfaces by pathogenic microorganisms, and those involved in spoilage. Listeria monocytogenes is a human pathogenic bacterium which may contaminate food and food preparation surfaces. The ability of L. monocytogenes to attach to, and be retained on, food contact surfaces is important for survival in food processing plants, where surfaces are continually being cleaned and disinfected. In addition, residual organic soil (ie food material) on the surface may reduce cleanability and the effectiveness of disinfection, as well as affecting the retention and survival of the attached microorganisms. It is therefore important to control surface contamination and cross-infection by implementing effective cleaning and disinfecting protocols, to ensure optimum product safety and plant hygiene. The aim of this work was to differentially assess (using staining and culture techniques) the retention of L.monocytogenes and surface soil (whey) on stainless steel following a number (0 - 30) of cleaning cycles in the presence and absence of industrial cleaning products. In general, organic soil was less easy to remove, and some indication of accumulation was evident. The presence of residual soil may interfere with the subsequent hygienic status of the surface due to interactions with contaminating microorganisms and/or cleaning products/components. It is therefore valuable to assess the presence of both soil and microorganisms on surface before and after cleaning.
Synthesis of Continuous Silicalite-1 Films by a Steam-assisted
Crystallization Method
Abstract A synthesis solution was prepared from the hydrolysis of a solution containing tetraethoxysilane, tetrapropylammonium hydroxide and distilled water. Uniform coatings were prepared from alcohol-diluted synthesis solutions using a G3P-8 spin-coater. The prepared films were steamed on a crucible stand in a PTFE-lined autoclave at 150 °C for 5 days with 0.2 g of H2O. The crystallized films were characterized by Powder X-ray Diffraction (XRD) and Scanning Electron Microscopy (SEM). The films were smooth and homogeneous with a thickness in the range 900 – 1800 nm. The film thickness increased in the order of 2-propanol < ethanol < methanol< 1-butanol. The Silicalite-1 crystals building up the films show the MFI-type structure and typical coffin shape morphology, confirmed by XRD and SEM respectively. Continuous Silicalite-1 films were synthesized through alcohol/water vapour steaming. This method employed limited amounts of structure directing agent. The films were easily recovered after the synthesis and negligible amounts of waste products were formed.
The effect of a commercial probiotic drink on the oral health of
healthy volunteers Abstract: Objectives: The beneficial effect of probiotic drink Yakult®, containing viable Lactobacillus casei Shirota (LcS), on gastrointestinal disorders is well documented. The aim of this study was to determine whether ingestion of Yakult affects oral malodour, and saliva and tongue microbiota in healthy dentate individuals. Materials and Methods: Twenty one healthy volunteers (8 males and 13 females, 25-45 years) undertook a 10 week trial comprising of 3 phases: three week baselines; 4 week intervention phase during which subjects drank Yakult Light one per day; followed by 3 week washout period. Samples of mouth air were taken using a portable sulphide monitor (Halimeter®). The microbial viability and composition of saliva and tongue dorsum coating were assessed using a range of solid selective and indicator media. Results: A total of 19 participants completed the trial. There was no significant change in the mean concentrations VSC in sampled mouth air between the 3 study phases. The levels of lactobacilli, streptococci, acidogenic microorganisms and Gram-negative anaerobic species did not vary significantly between baseline and Yakult intervention phases. However, a significant difference was observed in some individual cases. Out of the study group, fourteen (74%) were found to carry yeast in the saliva, five on which were constant carriers. The majority of isolates were identified as C. albicans. Conclusions:
Consumption of Yakult did not have a detrimental effect on the natural
fluctuations of oral populations.
Improvement of Thin Film Gas Detectors by Incorporation of Novel
Nanoparticles (Download
Poster .ppt) ABSTRACT The determination of SO2 levels in the past has been limited to aqueous phase and we are concerned with determining SO2 in the atmosphere. These nanosensors will allow the atmospheric SO2 levels to be measured. The objective of this work is to produce an enhanced sensor that works simultaneously. The fluorophore itself is by no means sufficient in giving the sensor its characteristic; the matrix it exists in also plays an important role. These nanosensors will be encapsulated in an organically modified sol-gels (ormosil) matrix. The organic ormosil group will anchor the dye molecules and therefore significantly improve gas uptake. This work aims to produce various gaseous sensors using different dyes. These gaseous detectors give a response by quenching luminescence. For the SO2-sensitive fluorescent nanosensor rhodamine B isothiocyanate (RITC) will be used, while ruthenium-tris(4,7-diphenyl-1,10-phenanthroline) dichloride (Ru(dpp)) is used as O2 sensor. The Ru(dpp) and RITC nanosensor produced a significantly high response to gases along with response recovery. The monodispersed nanoparticles sizes ranged from 200nm-500nm. These nanosensors were prepared by self-assembly technique using carboxylic acid functionalized dyed silica nanoparticles by ring opening linker elongation reaction of the amine function with succinic anhydride. Therefore this will allow for the synthesis of an enhanced duel sensor that works simultaneously. This ormosil encapsulated duel sensor should detect both sulphur dioxide (SO2) and oxygen (O2) gases using luminescence spectroscopy.
Parrot clay lick distribution in South America: Do patterns of where
help answer the question why? (Download
Poster .ppt)
|
