ESCORTER : Guide towards correct entries. Question Ordering in form. Greedy Information Gain Dynamically Reorder Questions Predict Errors to Re-ask. Contextualized Error Likelihood Principle.
Concept : An unscrupulous door-to-door surveyor Shirks Work, ask only important questions. Greedy Information Gain Uniform Prior : Equal likely inputs Training Set Context – specific Model Required Bayesian Learning
The patient dataset collected at a rural HIV/AIDS clinic at Tanzania. Survey dataset, responses from 1986 poll about race and politics
Approximates Double Entry Uncertainty : High Entropy Outliers
Due to digital divide between the developing and developed areas, it is very difficult to collect and use data in the developing regions. The main problems being : Lack of reliable infrastructure, Proper connectivity, and, Inadequate expertise. Currently available tools for data collection like Pedragon Forms, Nokia Data Gathering, Java-Rosa, RapidSMS etc. are difficult to deploy, hard to use, complicated to scale and rarely customizable.
The Open Data Kit or simply ODK is a suite of tools for data collection that uses Google’s Android platform. The main objectives of the technology are : Modularising and customising tools Use of open interfaces and standards Long time survival of tools. The three components of ODK are: 1. ODK Collect : collects data using Forms. 2. ODK Aggregate : ready to deploy online repository to store, view and export collected data. 3. ODK Build : enables users to generate forms. 4. ODK Voice : maps Forms to sound snippets. 5. ODK Clinic : mobile medical record system. 6. ODK Manage : maintains database of all phones for remote device management 7. ODK Validate : validates Form. Other tools being ODK Dropbox, ODK Rangefinder, ODK Tasks, ODK Listen and ODK Visualise.