Multilingualism and BI
Does spoken language matter anymore?
- By Bob Potter
- December 1, 2015
Earlier this year I read an article in the Wall Street Journal and listened to an NPR report about the world becoming more homogenized with fewer spoken languages. Today, more than two billion people speak English. Think of that: every fourth person on Earth speaks or understands English on some level. By the turn of the next century, we may go from 6,000 languages to 600, with English being the default lingua franca. I thought to myself, why translate software user interfaces into multiple languages? Why shouldn't everyone be forced to use their software in English, and what does this mean for business intelligence in the 21st century?
For BI, it is no longer about the experience users have with the application but more about their experience with the data itself. Self-service BI has driven software user interfaces to become more visual with very little emphasis on words. A modern responsive design of a new software package usually has 90 percent fewer words than an application built just a few years ago. The string translations are so small and efficient that developers can translate each release into multiple languages at a fraction of the cost that it took to write it in one language.
With browser-based design and interaction, users can establish their language preference in their profile settings. A single instance of the software can easily support multiple users, each interacting with the product in different languages. That's very cool, but what about the data?
Most business data is organized into multidimensional models that are managed in databases or OLAP cubes. Dimensions, attributes, and measures represent your data and you don't want that data translated or transformed from its language of origin just so it looks good in a report. What if you could have your cake and eat it too?
Modern analytical models allow users to create aliases of the data. For instance, in a budgeting and planning tool, users can create a translation alias of their product definitions -- as well as their attributes -- and from that alias create different cube views for analytics processing. Now my business analysts in China, Russia, or France can gain insight from the data in a localized UI as well as a localized view of the data itself.
Additionally, text analytics and sentiment analysis are emerging as hot trends in business intelligence. These analytics functions must happen in the local language. That's why in addition to programmers, software companies are hiring linguistic engineers to complement their staffs. No longer are we just analyzing numbers in a spreadsheet or database table. We're sifting through words in documents, social media, audio, and video files to either summarize vast amounts of data or to predict what will likely happen and take action. Business intelligence is not very intelligent if you're forcing everyone to use a single language.
Imagine two people from different cultures who speak and understand completely different languages communicating and understanding each other via data. Text analytics and sentiment analysis allows these two people to understand how their products are viewed in their respective markets and how sentiment is trending over time -- all without having to communicate in each other's spoken language.
That's why the BI sector of the software industry does care about multilingualism, and software vendors need to upgrade their data visualization products and data management back-end analytic platforms to be flexible and modern by allowing users to work the way they want to and using a natural language. Forget about your spoken language and let your data do the talking.