Natural Language Generation (NLG) is arguably the hub of natural language technologies. It uses natural language processing (NLP), is a requirement for conversational AI, and largely requires natural language comprehension (NLU) for meaningful answers to questions or commands.
However, when incorporated into a holistic platform, proper implementation of NLG does more than generate natural language communication from cognitive computing systems. Likewise, its capabilities transcend the mere understanding of natural language to actually produce relevant answers to questions and user dictations.
Credible solutions can couple this technology with a variety of analytical approaches to self-analyze datasets and then present the results in natural language. The business value of this widespread functionality is discussed in a recent analyst report from Forrester“The Overall Economic Impact of Arria Natural Language Generation”.
The findings show that this solution delivers a 209% ROI over three years, with a net present value of more than $3 million. It also indicates that users need to do 80 percent less labor to manually generate reports.
The ability to synthesize natural language searches, analyze a plethora of data sources, and present the results in natural language is inextricably linked to the organizational value this technology delivers. Due to its inherent linguistic appeal, it also supports a range of use cases, from increasing reporting scale to democratizing analytics across the enterprise.
“What people want to do is articulate what they want in a quick, natural way, instead of applying filters and drill-downs and swapping sliders and stuff.” Arria NLG CTO Neil Burnett explained. “For the most part, people are very good at putting themselves in words.”
The triad of capabilities that enable holistic NLG solutions like Arria NLGs to leverage the enterprise through natural language technologies starts with NLU† This subset of NLP allows users to interact with the NLG platform in natural language. This application of NLU connects to Natural Language Query, which allows users to ask natural language queries that are converted into queries of the underlying data sources.
With these natural language capabilities, users aren’t necessarily limited in the nature of their questions or the way they phrase them — both of which are inhibitors of traditional BI tools, which don’t support natural language commands or questions at all. “It’s not like a predefined set of questions, like a chatbot system,” Burnett explains.
The backend analytics that Arria NLG provides are regularly updated so that users can expand their range of questions to get answers that are useful. This distinction is essential for the value derived from NLG systems. Other implementations of this technology dependent on BI or artificial intelligence tools for data analysis, which can also integrate Arria’s solution. However, the company’s solution can also run its own analytics on data, such as that in a CRM system like Salesforce, to provide increasingly tailored answers to the kind of information users are seeking.
With this approach, “the first and most important thing you do is analyze the data,” Burnett noted. “You bring your data to us again, or your data will be updated below. There is no need to retrain our system. Our system is ready to analyze the data and respond.” The CTO characterized the type of analysis involved in this solution as statistical analysis, trend analysis, and anomaly analysis among others.
Generate Natural Language
The final part of this approach is the proper formulation of natural language answers, which is vital to provide results to users in a way they understand. Burnett identified a use case where “your new monthly sales figures come in; there is no work on the system. It’s going to treat them as completely new because it knows how to analyze those things, what’s going to be important to you, and how to articulate what it found there.”
The delicacy of using proper language, diction and semantics to generate natural language is at the heart of NLG. The whole point of this technology for analytic applications is to take what may be complicated statistical findings and make them understandable to tech novices in the know. “It’s important to pick and use the right words that match what the data wants to say,” Burnett revealed. “It could be as simple as thresholds or ways of naming something important, which will vary depending on what aspect you’re looking at.”
Arria NLG’s solution — and the recent Forrester study quantifying its value to business users — illustrates that comprehensive NLG solutions offer three principal benefits. They support natural language searches, natural language answers and backend analytics, devoid of external tooling (which can be used if desired), to get the right answers users need from their data.
About the author
Jelani Harper is an editorial consultant serving the information technology market. He specializes in data-driven applications focused on semantic technologies, data governance and analytics.
Follow us on Twitter: @InsideBigData1 – https://twitter.com/InsideBigData1
Sign up for the free insideBIGDATA newsletter†