NOAA's PORTS® (Physical Oceanographic Real-Time System) currently provides historical and real-time observations and predictions of water levels, coastal currents and other meteorological and oceanographic data for a dozen US estuaries. PORTS (http://co-ops.nos.noaa.gov/d_ports.html) displays its real-time data in graphical and tabular form. The tabular form can require lengthy inspection to retrieve the relevant details, particularly in a large PORTS locale with many sensors.

ARNS (Automated Real-time Narrative Summaries) is a prototype system being developed by CoGenTex for NOAA's National Ocean Service (NOS) [1]. ARNS will complement the tabular display by providing automatically generated natural-language summaries of estuary conditions (see Appendix for example), in order to give users a more comprehensive overview of the data. ARNS is built using Exemplars [2] (http://www.cogentex.com/technology/exemplars/index.shtml), a rule-based, object-oriented framework for dynamic generation of text and hypertext. Unlike earlier rule-based systems used to generate weather forecasts via multi-layered linguistic representations [3,4], ARNS relies on Exemplars' provision for linguistically marked “text snippets” which can be maintained and extended by user organizations for a given application without recourse to special linguistic training. Revision rules [5] provide “aggregation” and other types of final smoothing needed for concise, fluent text.

NOAA and CoGenTex have designed an XML format for input of sensor data to ARNS. NOAA is concentrating on building the appropriate data feed, while CoGenTex is building the generator and an ARNS-specific template editor for use by NOAA in maintaining the system. Pending completion and evaluation of the prototype, NOAA hopes to deploy ARNS to all PORTS web pages and voice response systems.