The Singapore Department of Statistics has developed the Advanced Coding Environment (ACE) software jointly with Kent Ridge Digital Labs (KRDL), a software R&D laboratory. The ACE, which will be deployed for the Census 2000’s coding operations, is a sophisticated, windows-based coding software that uses artificial intelligence (AI) techniques for coding statistical information collected in censuses and surveys.

Online coding is carried out in 2 stages under the ACE, namely, Automatic coding (via Auto-Coder module) and Intelligent coding (via Code Wizard module). From the Census or survey data collected, the Auto-Coder performs a direct string-match of code descriptions with numeric codes and assigns codes to the clear-cut cases. At this automated stage, some 40 per cent and 10 per cent of industry and occupation codes are expected to be coded respectively.

In the second stage, records that are not coded automatically are sent for Intelligent coding. ACE retrieves each record and uses AI to suggest possible codes. The AI uses in-built logic such as synonymous words and past relational data between education, occupation, income and industry to recommend a list of possible codes. Human coders make the final judgement and select the correct code. The panel above illustrates how a record gets coded in 3 simple steps.

With ACE, the coding operations for Census 2000 will be done speedily and efficiently. Coders need not refer to coding dictionaries as they did in past Censuses. The manual operations in coding will be reduced without any compromise to data quality. Significant cost savings will be achieved as less manpower is required. Being a generic tool capable of large volume coding, ACE could serve as an efficient coding tool for complex statistical information.

Source :  Statistics Singapore Newsletter, July 1999