Natural Language Generation is a field of Machine Learning that deals with Natural Language Processing to generate human-like speech. NLG systems take as input a computer representation such as knowledge base or a logical form for example- a set of rules. Spotfire® integrates with Wordsmith to take structured data and create stories with the tone personality & variability of a human being. These narratives can be embedded within the analytics dashboard. In a world of Big Data, ever increasing automation makes it possible to find Insights real-time. However, manually creating descriptions and reports for the data insights is a major bottleneck. Leveraging Natural Language Generation with Spotfire has the following advantages:
- Reduces reliance on people to (often incorrectly and incompletely) interpret data
- Quickens pace of reporting with the ability to generate hundreds of thousands of distinct unique narratives everyday
- Provides control to standardize the diction and style of your reports
- Provides flexibility to choose the Business questions or Insights to highlight in narrative
Click on link for demo of these capabilities during TIBCO Analytics Meetup of February 28th, 2017 - TIBCO Analytics Meetup page includes link to video
Wordsmith, powered by Automated Insights is a product that uses a rule-based template for Natural Language Generation. In the template, Rules are configured in the form of complex nested branches for each control flow of narrative. Each branches in turn contains more rule Branches, Data Variables (data mapped from Spotfire) and Synonyms used to describe the state of the analysis.
Building Blocks of Wordsmith Template
For ease of use, Wordsmith internally handles several edge cases like varying grammar rules for synonyms (eg. an exceptional quarter vs. a remarkable quarter). It also provides formatting options for display based on type of variable. Once the template is configured it can be accessed with a lightweight, minimal latency Web API.
Format Options for Numerical Variable
Spotfire NLG Solution Architecture
User configures the NLG Template in Wordsmith. In Spotfire, variable information is extracted for markings and visualizations of interest. This variable information is then structed as JSON and passed as an API Request to Wordsmith cloud. Authentication is taken care of with standard API Key convention. The Wordsmith server then returns a response in realtime that contains the narrative displayed in Spotfire text areas.
IronPython Script Snippets
Sample JSON data
Sample UI Walkthrough
There are 2 parts to the UI:
- Wordsmith Template: contains control flow and preview
- Spotfire Dashboard: interactive, real-time narrative generation
Preview facility has a summary tab that allows the user to see statustics related to the generated narrative like a score for Readability and Variability.
Achieve seamless, interactive, real-time Natural Language Generation Capability in your dashboard each time!
For more details please reach out to: