Is Foreign Language a Barrier for Your Natural Language App?
Last month IBM announced that Watson would learn to think in Japanese. This makes perfect sense. Most people who speak a second language will tell you that the only way to start speaking it fluently is to learn to think in it. But it highlights another issue. Most natural language technologies are developed to work in a single language, typically English, and that adds complications for organizations deploying global applications.
If you’ve ever tried to learn a second language you’ll have realized that the secret to doing it well is not to translate, but to understand the nuances of the words within the language and the context of the conversation. As Nicole Kidman in The Interpreter said, “If I interpreted gone as dead, I would be out of a job.”
The fastest way for customers to lose faith in a natural language application is for it to demonstrate a lack of understanding or an inability to articulate the answers clearly. As humans, we barely make allowances for non-native speakers in a call centre, it’s definitely a no-go area for a digital employee. This is why Teneo already thinks and interprets in over 20 languages.
But it’s not just understanding that is important in dealing with different languages. Being able to react consistently across a global enterprise is important too. However, this requires central control and since most natural language technologies set up foreign languages as separate entities this isn’t easily achievable.
Consequently organizations end up with a solution that is dissolved over separate teams, in different countries, managing the same solution in different ways. Eventually over time, what should be a consistent approach within an organization becomes disparate, both in quality and content.
At Artificial Solutions we take a different approach to developing multi-lingual natural language applications, one that not only provides the central control, but also uses fewer resources, while still allowing for localization.
We use a master structure to develop the NLI application, which enables a core project team within an organization to manage the implementation, steer what the main (common) knowledge should look like and how it should be implemented in the different countries/languages. In the majority of cases around 80% of the knowledge in the master structure can be reused, leaving just 20% to be localized.
This approach also makes it easier to update and maintain the applications and the knowledge, making NLI applications easily scalable across global operations. In the long term a high level of consistency and quality is maintained, raising the success level of the application.
As natural language becomes less of a point solution to an immediate problem and more a long term tool to transform business goals, so its multi-lingual capabilities will become essential. While thinking natively is critical in the interaction with users, without central knowledge control a multi-lingual application can end up costing more than it saves.
How Conversational AI Increases Your Competitive Edge White Paper
This white paper discusses how a conversational AI platform can help improve customer engagement and deliver a vital competitive edge.