Hyokyeong Lee

A semi-automated wizard of oz interface for modeling tutorial strategies

Human teaching strategies are usually inferred from transcripts of face-to-face conversations or computer-mediated dialogs between learner and tutor. However, during natural interactions there are no constraints on the human tutors behavior and thus tutorial strategies are difficult to analyze and reproduce in a computational model. To overcome this problem, we have realized a Wizard of Oz interface, which by constraining the tutors interaction makes explicit his decisions about why, how, and when to assist the student in a computer-based learning environment. These decisions automatically generate natural language utterances of different types according to two “politeness” strategies. We have successfully used the interface to model tutorial strategies.

A Mutual Information-Based Metric for Identification of Nonlinear Injector Producer Relationships in Waterfloods

In this paper we introduce a new analytical approach for management of waterfloods in heterogeneous reservoirs. The main contribution is the development of a process and metric to evaluate the pair-wise injector-producer (IP) relationships, i.e., to quantify the impact of any injection well on the neighboring producing wells. The proposed metric is particularly designed to consider the non-linearity of the IP relationship between the injection and production rates by using the Mutual Information (MI) data mining tool. Non-linearity of the IP relationship is the main challenge in quantifying this relationship and, to the best of our knowledge, this is the first time that MI is used in the petroleum literature for IP relationship identification. In addition to MI that captures the non-linear correlation in the IP relationship, our metric considers other parameters such as the distance between the IP pair as well as their relative injection and production rates, respectively. Leveraging our proposed metric, we propose a system, for optimal waterflooding with which a field engineer can automatically: 1) Identify the under-performing producers based on their performance characteristics such as wateroil ratio, gas oil ratio, and oil production rate; 2) Rank all injectors based on their impact on the under-performing producers using our proposed IP relationship identification metric; 3) Decide on optimal injection volumes for individual injectors that have the most impact on the under-performing producers and maximize the recovery factor. The proposed technique can significantly reduce the decision-making time for the effective management of complex waterflood.

Inference of kinship using spatial distributions of SNPs for genome-wide association studies

BackgroundGenome-wide association studies (GWASs) are powerful in identifying genetic loci which cause complex traits of common diseases. However, it is well known that inappropriately accounting for pedigree or population structure leads to spurious associations. GWASs have often encountered increased type I error rates due to the correlated genotypes of cryptically related individuals or subgroups. Therefore, accurate pedigree information is crucial for successful GWASs.ResultsWe propose a distance-based method KIND to estimate kinship coefficients among individuals. Our method utilizes the spatial distribution of SNPs in the genome that represents how far each minor-allele variant is located from its neighboring minor-allele variants. The SNP distribution of each individual was presented in a feature vector in Euclidean space, and then the kinship coefficient was inferred from the two vectors of each individual pair. We demonstrate that the distance information can measure the similarity of genetic variants of individuals accurately and efficiently. We applied our method to a synthetic data set and two real data sets (i.e. the HapMap phase III and the 1000 genomes data). We investigated the estimation accuracy of kinship coefficients not only within homogeneous populations but also for a population with extreme stratification.ConclusionsOur method KIND usually produces more accurate and more robust kinship coefficient estimates than existing methods especially for populations with extreme stratification. It can serve as an important and very efficient tool for GWASs.