Technology - Cognitive Architecture

Cognitive Architecture

Inspired by the Human mind, BankBuddy architecture is based on the cognitive mechanisms of Knowledge & skill representation as domain ontology, short-term memory in the form of context, Multi step reasoning as recommendation systems , perception based on reinforcement learning, & sensory input/output as text, touch, speech, vision from computer vision, speech (ASR,TTS), multi lingual text NLP & NLG.

GenAI Natural Language Processing

Most comprehensive banking domain specific Natural processing engine, powered by a knowledge graph ontology with pre-defined products, attributes, slang, abbreviations and language packs.

Bidirectional Context

Product – Attribute - Product mapping for human like conversations & context specific streaming recommendations.

Polyglot Language Model

Understand Slang, mixed language, Abbreviations across multiple languages in the same model.

Sentiment Detection

Superlative sentiment detection with subjectivity analysis, intensifiers & negation handling

Multi-intent processing

Context driven Multi intent analysis and joint slot filling.

Get a personalized customer experience automation assessment

Discover

Machine vision

Advanced Machine vision models trained and optimized for Banking specific user journeys viz. eKYC - ID OCR, photo extraction, face match, document forgery detection, deep fake check, liveness detection.

Face Verification

Photo extraction from ID’s and matching with photo and live selfie/streaming video.

Liveness detection

70+ parameters to detect liveness accurately.
.

Forgery detection

Video and document analysis for forgery and deepfake detection.
.

Text Extraction/OCR

Text extraction from ID’s and documents with processing pipeline

Speech

BankBuddy speech engine is phonetically trained specifically for banking speech / voice recognition across languages, dialects, slang and abbreviations for highest accuracy.

Speech Verification

Bank grade speech verification for MFA and frictionless banking.

Multi lingual ASR

Domain specific ASR with language models for lower WER and higher accuracy.

Speech Analytics

Realtime emotion detection and analysis for tailored actions.

Standards based

Leveraging standards like SSML, Voice XML , SISR, SRGS, WebRTC

Recommendations

BankBuddy offers the most advanced banking recommendation model with Ensemble learning that combines multiple algorithms to deliver the personalised recommendations for accurate targeting domain specific cognitive model with dynamic persona generation based on Behavioural data analytics and scoring based on past interactions (Demographics, Past behaviour patterns, social profile, interaction pattern) & Reinforcement learning for continuous improvement

Dynamic Persona Generation

BFSI specific model trained on your product and customer attributes

Nearest neighbor

Identify closest persona in the financial Journey

Context pairing

Dynamic persona paired with current context for higher prediction accuracy

Streaming recommendations

Live active stream of recommendations with reinforcement learning