We start by showing that in an active learning setting, the Perceptron algorithm needs $\Omega(\frac{1}{\epsilon^2})$ labels to learn linear separators within generalization error $\epsilon$. We then present a simple selective sampling algorithm for this problem, which combines a modification of the perceptron update with an adaptive filtering rule for deciding which points to query. For data distributed uniformly over the unit sphere, we show that our algorithm reaches generalization error $\epsilon$ after asking for just $\tilde{O}(d \log \frac{1}{\epsilon})$ labels. This exponential improvement over the usual sample complexity of supervised learning has previously been demonstrated only for the computationally more complex query-by-committee algorithm.

Active learning reduces annotation costs for supervised learning by concentrating labelling efforts on the most informative data. Most active learning methods assume that the model structure is fixed in advance and focus upon improving parameters within that structure. However, this is not appropriate for natural language processing where the model structure and associated parameters are determined using labelled data. Applying traditional active learning methods to natural language processing can fail to produce expected reductions in annotation cost. We show that one of the reasons for this problem is that active learning can only select examples which are already covered by the model. In this thesis, we better tailor active learning to the need of natural language processing as follows. We formulate the Unreliable Parameter Principle: Active learning should explicitly and additionally address unreliably trained model parameters in order to optimally reduce classification error. In order to do so, we should target both missing events and infrequent events. We demonstrate the effectiveness of such an approach for a range of natural language processing tasks: prepositional phrase attachment, sequence labelling, and syntactic parsing. For prepositional phrase attachment, the explicit selection of unknown prepositions significantly improves coverage and classification performance for all examined active learning methods. For sequence labelling, we introduce a novel active learning method which explicitly targets unreliable parameters by selecting sentences with many unknown words and a large number of unobserved transition probabilities. For parsing, targeting unparseable sentences significantly improves coverage and f-measure in active learning.

Spray Polyurethane Foam insulation (SPF) kits are currently being marketed and sold to do-it-yourselfers to meet various insulating needs. Like commercial SPF systems, the primary health concern with SPF kits is user overexposure to the isocyanates during product application. The potential health risk associated with SPF applications is driven by several factors including (but not limited to): the toxicity of isocyanates; the potentially high exposure intensity; the quantity of isocyanates used in the process; the enclosed nature of the environment in which the product could be applied; the potentially high exposure duration/frequency; and the limited availability of control measures to reduce agent intensity (e.g., personal protective equipment, dilution ventilation). To better understand the potential hazards associated with the use of SPF kits, the current study was designed to provide an initial characterization of user exposure to airborne particulate during the application process. Specifically, the study would aim to answer the following: * What is the particle size distribution of the aerosol a SPF kit user is exposed to during application? * What is the airborne particle mass concentration a SPF kit user is exposed to during application? To answer these questions, a single commercially available SPF kit was selected for use and a mock residential environment was constructed to support repeated applications of SPF. Size-selective and total dust air sampling were conducted during the applications to determine the particle size distribution and mass concentration of aerosols generated by the selected kit. The particle size distributions developed from the size selective sampling results showed the presence of airborne particulate capable of penetration to the gas exchange regions of the respiratory tract. The average mass median diameter and geometric standard deviation of the particle size distributions were 4.6 µm and 2.7 respectively. The total dust sampling results showed mean airborne concentrations of 10.40 mg/m3. Based on the sampling results the study, personal air monitoring is needed to assess the degree of user exposure to methylene diphenyl diisocyanate (MDI) and to provide information for the selection of exposure control methods.