01 Deployment Scenarios
CUSTOMER FACING DEPLOYMENT
SenseBased is able to answer customer questions typed on the website by retrieving the relevant information. SensBased is able to access FI documents in tabular format and answer in English.
"What interest can I earn on savings?"
"You can earn different interest in a savings account or in a money market account."
"What interest can I earn in a savings account?"
"You can earn 3% on the first $500, and 0.05% on the remaining balance."
"What is the interest on a money market account?"
"You can earn various interest rates ranging between 0.2% and 0.8% on balances between $1,000 and $100,000 respectively
"What would be the interest on $20,000?"
"The interest on $20,000 would be 0.35%."
Today, financial institutions face a knowledge-transfer challenge. They need to deal with attrition, bring new employees in Contact Centers and branches up to speed quickly, and put the required information at their fingertips. Often, these employees have customers on the line or sitting in front of them and need real-time answers.
This may be challenging, as a FI typically has tens of thousands of documents covering product features, services, procedures, policies, faq’s, competition, regulations and compliance. Internal users can explore document clusters by product/service, user tasks and document types. A visual search enabling drilling down to the right document is available for power-users.
02 Semantic Search
The Semantic Search challenge in financial services
Search of financial document collections is challenging because relatively few concepts tend to occur with high frequency in tens of thousands of documents. This makes returning the relevant document more challenging. North Side offers high-performance semantic search technology able to return the right document to the right user, irrespective of the keywords used for search. This is particularly important for customers, or new employees who are not yet up to speed on the vocabulary used by the FI.
Semantic search, as opposed to keyword search, means that SenseBased will find:
• ‘prepayment’ when user types ‘pay loan before end of term’
• ‘adjudicate’, when user types ‘approve credit’
• ‘delinquency’ when user types ‘skipped a payment’
Validation at a top five Canadian bank confirmed major improvements in:
• relevant results returned (95% of cases versus 50%)
• number of relevant documents returned per query (4X improvement)
• reduction in no documents found (10 X improvement)
• machine learning to associate search queries with documents based on user choices.