Narrative Science Gives a Voice to Data


Narrative Science provides a natural language generation (NLG) platform (cloud or on-premises) that transforms raw data into a narrative dialogue. So we might provide sales data for the last year, and the platform will pick out salient points – that sales rose by 10% year on year, and that the fastest growth was in product category X.

The platform offered by Narrative Science is called Quill, and it represents one of the more advanced solutions in this space, with its AI driven technology. Other NLG platforms might require users to set up ‘templates’, but the technology within Quill automatically understands the nature of the data and how to extract the most meaningful dialogue.

narrative sceince

Quill has found broad application in financial services, insurance, government, retail and health. However one of the more exciting applications is that in business intelligence (BI). Instead of total reliance on charts and graphs, users of several BI platforms can now hook in to Quill to generate easily understood narratives. The BI platforms currently supported include Qlik, Tableau, Power BI, Sisense, Lumira and Microstrategy, with Qlik being the deepest integration, but Sisense possibly being the most advanced. The Sisense integration is worthy of special mention, simply because it points to some interesting futures. Sisense can deliver the dialogues created by Quill through voice and messaging media. In effect the narratives can be spoken using Alexa, or sent as a text message using any of the popular messaging platforms (e.g. Slack).

That Quill does not need to be configured in any way before conducting analysis of data, is due to the inbuilt intelligence. It understands that time series data for example, will display trends, outliers and other features typical of this type of data. So when the data from a chart or graph is sent to Quill, it knows exactly what to look for.

The user interfaces are very easy to use, and dragging a narratives icon onto a chart or dashboard is all that need be done to create a narrative. As users change focus in a particular chart or graph, so the narrative automatically changes too, and various settings allow the narratives to be finely tuned for purpose. A senior executive for example, might only want to see a subset of the items a narrative generates. In fact the settings options are quite sophisticated, and give users plenty of opportunity to tailor Quill narratives to their particular needs. Thresholds can be specified at which a trend becomes of interest (it is unreported until this threshold is met), and if data contains more than 30 periods then Quill can generate forecasts for ten per cent of the data sample period count, into the future.

narrative sceince

Narrative Science provides its Narratives for BI API so that any BI or analytics platform can generate narratives. More generally Quill can be used as a cloud service, or installed on-premises. This latter option is seen as absolutely necessary for businesses where data security is paramount. Any application where data narratives add value can pass data to Quill and have a meaningful dialogue passed back to them. Many applications require that the narratives are highly structured, and Quill provides an environment where users can specify communication goals and other factors that will affect narrative generation.

NLG will become widely adopted, and particularly by users of BI platforms. We estimate that within three years all BI platforms will offer it as an option, and that natural language will become the dominant form of data analysis communication within five years. It reduces the skill set necessary to analyze data, and also complements visual data analysis. Inevitably NLG will facilitate voice communication on mobile devices, making it unnecessary to fiddle around with charts and graphs on a small screen. The future of NLG looks very promising, and as one of the leading suppliers in this space, so does that of Narrative Science.