Built surname matching and review CSV workflow

• Day: 2026-06-04
• Time: 11:35 to 11:45
• Project: Dev
• Workspace: WP 2: Operational
• Status: Completed
• Priority: MEDIUM
• Assignee: Matías Nehuen Iglesias
• Tags: Python, Pandas, Record-Linkage, Csv, Data-Cleaning, Normalization

Description

Session Goal

Develop a robust Python workflow to match contact sheets against a canonical roster using exact surname-based logic, while preserving original sheet order and producing reviewable outputs for manual adjudication.

Key Activities

• Drafted a pandas-based script to cross normalized contact sheets against a canonical roster by surname.
• Adjusted the matching logic from separate name columns to a single canonical Apellido Nombre field.
• Added text normalization and phrase matching with word boundaries to reduce false positives.
• Designed output splits for review, too-many, and no-match cases so ambiguous records can be manually resolved.
• Specified processing rules to preserve original row/sheet order and prevent rare duplicates from reappearing across sheets.
• Proposed a downstream decision CSV format to support manual validation and later reconciliation.

Achievements

• Defined a clearer entity-resolution approach centered on surname containment and normalized text matching.
• Established a repeatable review workflow that surfaces ambiguous matches instead of forcing automatic assignment.
• Clarified operational requirements for traceability: order preservation, duplicate avoidance, and console-visible review blocks.

Pending Tasks

• Run the script against the real CSV inputs and validate match quality.
• Review the generated review, too-many, and no-match files to tune normalization and matching thresholds.
• Confirm the final decision CSV schema for manual adjudication and downstream ingestion.

Evidence

• source_file=2026-06-04.sessions.jsonl, line_number=3, event_count=0, session_id=86b59157c0b2f6e2008e5e4a542baec72ebe5b83009fbbcfd6a5f40aef16d4aa
• event_ids: []