Episode Summary: Machine learning (ML) can be used to identify objects and pictures or help steer vehicles, but is not best suited for text-based AI applications says Robbie Allen, founder of Automated Insights.
In this episode of AI in Industry, we speak with Robbie about what is possible in generating text with AI and why rules based processes are a big part of natural language generation (NLG). We also explore which industries are likely to adopt such NLG techniques and in what ways can NLG help in business intelligence applications in the near future.
Subscribe to our AI in Industry Podcast with your favorite podcast service:
Guest: Robbie Allen, Exec Chairman at Automated Insights
Expertise: Machine learning, Artificial intelligence, Natural language generation
Brief Recognition: Before founding Automated Insights in 2007, Robbie earned two Masters degrees from the Massachusetts Institute of Technology (MIT) in civil and environment engineering and engineering and management. He has also had previous experience as a software engineer at Cisco and IBM.
Current Affiliations: Robbie is currently the CEO of at Infinia ML and is pursuing a Ph.D. from University of North Carolina in artificial intelligence
Big Idea
According to Robbie, although sports and finance applications for NLG are common, they are the low hanging fruit use-cases. He goes on to add that ‘democratizing insight in business intelligence’ (BI) is one of the largest growing applications for NLG today.
A typical use case for such an application would involve a business user with access to a labelled database of business intelligence variables. For such users, NLG platforms can enable data summarization and personalization at a scale that is not possible with humans.
For example, an organization could use data tracked from around 10-20 BI variables to define what would represent a ‘good’ financial quarter. NLG platforms can represent the same data in plain-language (output – You did/did not have a good quarter), making insights from large datasets more accessible to non-technical staff.
Business-use patterns for companies looking to apply NLG is would be similar to Salesforce, the CRM (Customer Relationship Management) platform. In general, most companies don’t use Salesforce just ‘out of the box’, but rather customize it as the complexity of the business increases.
Current NLG platforms have basic summarizing capabilities (from generic input data of BI variables common to most companies). Adding user-specific BI variables to improve accuracy of the platform requires more expertise and domain understanding, resulting in reconfiguration with more input data sets.
Robbie believes that internal business use-cases for NLG will rapidly grow in the future mainly because this is one area where there already exists a large amount of well structured and labelled data.
Interested readers can learn more about how Automated Insights’s Wordsmith platform works through our case-studies here.
Interview Highlights with Robbie from Automated Insights
- (1.32) What’s possible today with natural language generation?
- (5.44) What are some of the prevalent use-cases for NLG technology today in Business Intelligence applications?
- (19.38) What does the initial set up of such a platform involve in terms of configuration and training of the AI?
- (24.15) Where do you see NLG becoming ubiquitous in the future?
Subscribe to our AI in Industry Podcast with your favorite podcast service:
Header image credit: Adobe Stock
