​How good is IBM Watson at foreign languages? One Spanish bank is about to find out

Catalan financial group CaixaBank is the first company to try training the IBM Watson cognitive computing system in Spanish.
Written by Anna Solana, Contributor

IBM Watson has been trained by CaixaBank's foreign-trade specialists to answer questions in Spanish.

Image: IBM

Your cognitive computing future is here, apparently. Last year, a study by IBM argued that intelligent machines simulating the capabilities of the human brain will fundamentally change a range of industries, including banking.

Now, in Spain, IBM Watson, the AI and analytics technology that can interact with humans using natural language and which made a name for itself by winning Jeopardy in 2011, has been trained by CaixaBank's foreign-trade specialists to answer questions and resolve technical problems in Spanish.

IBM and CaixaBank describe the new system as the first Watson application in Spanish and the first in Europe's financial sector.

Admittedly, it can't deal with tough general questions, such as, 'What's the formula for a company to succeed internationally?'.

But it is good at responding to specific issues involving huge amounts of structured and unstructured data. For example: 'What documents are needed for a documentary credit? What is a confirmed export documentary letter of credit? What is the safest payment mode for an exporter?'.

IBM Watson returns responses in areas where the system has more confidence and provides the information that enabled it to reach its conclusions. It is designed to learn from previous analyses and from its interactions with users.

The IBM Watson application also enables the user to view its whole decision-making process. Those insights help staff at CaixaBank, a large financial group in the Spanish market with 14 million customers, to answer clients' questions much more quickly.

Yet, it took some time to get there. First, a team of IBM and CaixaBank engineers and researchers from Spain, Ireland, and the US collaboratively transformed the machine-learning algorithms and internal components of IBM Watson's recognition engine to understand Spanish.

Then, Watson needed nearly five months to learn precise foreign-trade terminology in Spanish and to be able to respond according to the specific needs of CaixaBank, headquartered in Barcelona.

The bank's team worked with IBM's Digital Transformation Lab, IBM Studio in Madrid, on improving the user experience and interaction with the system, and simplifying navigation.

Now, the resulting system is designed to complement CaixaBank's service for helping companies succeed in international markets, which currently manages operations in 127 countries. IBM Watson should also help in relation to competing with other banks and public agencies, such as the Catalan ACCIÓ and Spain's ICEX.

"IBM Watson cognitive computing allows us to reinforce the assessment to our business clients while providing agents with the most innovative technology to respond to clients in real time and thus improve our customer service," CaixaBank managing director, Juan Alcaraz said in a statement.

CaixaBank has not revealed whether it has plans for training the system to improve other business areas such as risk-management advice, an area where IBM's Watson already displayed its potential a few years ago.

Read more about IBM Watson

Editorial standards