Page 173 - Contributed Paper Session (CPS) - Volume 3

CPS1983 Chong N. et al.

Diagram 2: Process flow of Sentiment Analysis

Sentiment analysis is also utilized when assigning cases to interviewers
performing fieldwork. Cases can be accurately matched to interviewers who
possess the necessary skillset to complete the case. For example, a new
interviewer will be assigned cases which previously had positive or neutral
sentiments while a more experienced interviewer will handle cases that had
history of negative sentiments. This ensures that each interviewer will be able
to conduct effective interviews with the respondents, hence increasing the
survey response rate.
Sentiment analysis can also help to measure the performance of
interviewers. An interviewer who continuously has negative connotations in
their conversations with survey respondents will be flagged out for retraining
on customer service skills. Similarly, interviewers with good customer service
skills can also be identified. As such, interviewers are not only monitored
quantitively on their output, but also on the qualitative side such as soft skills.
This ensures that all survey respondents go through an optimized survey
journey experience.

c) Automated Classification System
As MRSD compiles statistics on the labour market, information on
occupation and industry are key data items that will aid policymakers have an
accurate sensing of the labour market. It is also the most tedious data item to
collect, requiring large amounts of time and resources to classify the textual
information. In the past, respondents would just provide some details of their
occupation and interviewers have to classify each of them into one of the 1,202
codes of the Singapore Standard Occupational Classification (SSOC).
The Automated Classification System (ACS) was developed to cope with
the increasing textual data being collected. It utilizes text analytics algorithms
to convert unstructured data into meaningful structured data that can be used
for analysis. The data is cleaned through a process of word tokenisation,
customised stemming, removal of stop words and punctuations (Diagram 3).

162 | I S I W S C 2 0 1 9