PtyParser — the open-source parser behind TxFlow
An open-source parser for entity and address data in payment messages — and the work that preceded TxFlow.
This is not a product. There is no licence, support or roadmap attached to it. The commercial answer to address structuring is TxFlow Workbench, which adds confidence scoring, manual review of uncertain results and a field-level audit trail.
What it does
Payment messages carry sender and receiver details as loosely formatted free text across SWIFT MT, ISO 15022 and ISO 20022. PtyParser converts those party fields into structured, machine-readable data.
- Entity recognition — detects and labels persons, organisations and locations, using spaCy.
- Address parsing — breaks free text into house number, street, city, region, postcode and country, using Libpostal.
- Standard-agnostic — works across MT and ISO 20022, and tolerates format variation.
- Plain interface — raw strings in, structured JSON out.
Status, stated plainly
Published as
payment-text-parser
on PyPI, with source at
GitLab.
- 114,000 downloads since 2020.
- One release, v0.0.9, published October 2020. It has not been updated since, and current usage is in the single digits per month.
Both figures are stated because the first without the second would overstate it. The package is useful and real; it is not actively maintained.
Where it led
The problem PtyParser addresses is the one TxFlow Workbench now solves commercially — structuring and remediating postal address data in ISO 20022 and legacy MT. The difference is everything a bank needs around the parsing itself: confidence-based processing, manual review where the result is uncertain, a field-level audit trail, and full-fidelity conversion that does not invent information the source does not contain.