181

Segmentation and Disposition Classification through Text MiningBFSI

Imagine a debt aging like cheese. Early on, it’s fresh and manageable. But as time passes, it hardens, becoming more challenging to recover. Late stage collections deal with these “aged” debts, typically exceeding 90 days past due. It’s the final act in the debt recovery play, where traditional tactics might have failed, and more assertive strategies are employed. Just like a skilled cheese connoisseur, effective late-stage collections require strategies and expertise, to implement them to achieve success.

Colored crowd people isometric pie chart composition with people who organized flash mob vector illustration

The client, one of the largest banks in UAE, has initiated a Proof of Concept (POC) on the usage of text mining for customer segmentation and disposition classification in late-stage collections.

Challenges
  • Information Extraction from Agent Call Transcripts

Extract relevant information from agent call transcripts, including customer interactions, sentiments, and reasons for late payments.

  • Topics and Aspect Modeling

Topics and aspect modeling on the call transcripts to identify key themes and topics related to late-stage collections.

  • Combining Features for Disposition Classification

Combine the extracted features of topics and aspects to classify dispositions, such as payment promises, disputes, or financial difficulties.

Custom Models & Strategies to Fill the Gap in Achieving Targets