In response to the coronavirus pandemic, governments have enacted a host of policies. The effectiveness of these efforts in slowing down the spread of COVID-19 has been widely discussed. However, to study this problem, the very first thing is to find the data. Great efforts have been put into the curation of databases, resulting in hundreds of policy trackers. Most efforts have used crowdsourcing on news reports. This requires a relatively large effort to find relevant sources and extract policy from the source, which makes it hard to maintain the database.

This work proposes to use Twitter data from county officials, county departments of public health, and county offices of emergency services as a source for detailed county level nonpharmaceutical interventions. With the 140 word limit on tweets, these tweets are oftentimes a concise summary of current county policy, which with further processing, could possibly serve to reduce the workload in COVID-19 policy collection. Here, we built a pipeline for extracting and tagging county policies in California. With a proper selection of Twitter accounts and tagging schemes, this pipeline can be easily extended to other states.