The objective of this research project was to monitor Twitter conversations in Indonesia to understand how conversation volume trends relate to world events. In particular, this project monitored food-price related tweets between January 2011 and December 2012 to see if variations in their volumes could be connected with food and fuel price inflation.
The first step in the project was to create a taxonomy of relevant keywords and phrases in Bahasa Indonesia to extract relevant tweets, which included phrases related to the prices of food and fuel. As a second step, the researchers define categories in which to classify the relevant tweets, depending on the sentiment they express (i. e. “positive, ” “negative, ” “confused. ”). A representative, hand-labeled sample of tweets was then used to train a monitor to classify the tweets in the correct category and detect the sentiment of new tweets being published in real time.
Finally, all relevant data was analyzed to provide a proportion of tweets related to each theme to determine the statistical pattern of conversation for each category. The general volume of relevant tweets, independent from the conversation, was analyzed as well, and three spikes in volume of tweets were observed in 2012, corresponding to three real-world events:
- July 2012: a global soybean price rise, which affected the prices of tempeh and tofu, two dishes made of soybeans consumed by many Indonesians as affordable protein-rich options.
- March 2012: an proposal by the Indonesian government that its fuel subsidy would be cut by 33%, which caused violent protests and raised concerns that food prices may eventually be affected.
- November 2012: approval of a law establishing a new food agency with policymaking authority with the aim of helping Indonesia reach self-sufficiency in staple foods, including rice and soybeans.
The initial research results showed that around the same time when these real-world events occurred, conversations related to food prices also spiked dramatically among Indonesian Twitter users, illustrating the potential value of employing regular social media analysis for early warning and impact monitoring.
