Decision Criteria For Using Automated Coding In Survey Processing

• Activity Number: 74
• Type: Contributed
• Date/Time: Monday, August 12, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods*
• Author(s): Daniel Gillman*+
• Affiliation(s): U.S. Bureau of Labor Statistics
• Keywords: classification ; cost model ; error control

Abstract

Some survey data are classified into pre-specified categories during a process known as coding. If a computer assigns codes without human interaction, then this is called automated coding. Manual coding, computer-assisted manual coding, and interactive coding all require some level of human interaction.

The decision to employ automated coding in survey processing is not simple. There are many options and expenses to consider. Some of these are as follows: 1.) who develops the software; 2.) what is an acceptable error rate; 3.) how will errors be controlled; 4.) what percent of the cases must the automated coder classify; 5.) how much maintenance will a production system require; and 6.) what new resources must be developed to build an automated coder.

These criteria and others are described in this paper. A cost model is developed along with a description of the interactions between the criteria. Finally, some examples are given to show how the model might be employed by a survey organization.