The 2017 Coyote Creek Flood caused an estimated $100M in damage and forced the emergency evacuation of 14,000 people, most of them residents of neighborhoods in and around downtown San Jose (Mercury News, 2019). During these types of events, emergency services (fire fighters, police, and volunteers) go door-to-door to inform residents of evacuation orders. At the same time, emergency management staff use social media channels (Facebook and Twitter) and SMS (via Santa Clara County’s Everbridge system) to notify residents. Disproportionately impacted by these methods of communication are non-English speaking residents (predominately Vietnamese and Spanish), low-income, and elderly individuals, for whom these emergency notifications are not efficiently conveyed.
Additionally, resident action (or in-action) is not fully captured by reporting methods. The communications department reports media coverage metrics (impressions, clicks, etc) from social channels and outreach numbers from SMS channels, but metrics are not available from door-to-door efforts. For instance, in the Sonoma County fires, 100,000 people were notified to evacuate, leading to only 200 people showing up to the shelter; yet, this does not necessarily represent the total number of individuals taking action, as they may go elsewhere (family-and-friends, hotels, etc).
The above results in:
· Decreased clarity of evacuation instructions
· Inconsistent notifications
· Frustrated residents
· Increased time to action
· Lack of visibility into resident action
The Office or Emergency Management seeks a solution that would have the following characteristics:
· Real-time transcription and translation tools for use in the field and in the office
· Speech-to-text and text-to-text transcription and translation for English, Vietnamese, and Spanish via mobile and desktop
· Integration with social media and SMS communication channels
· Reporting and analytics tools
· High transcription and translation accuracy
· Language auto-detection
· Conversation database