The payment messages represent an important share of the overall inter-banking traffic, both in volume and in amount. In such messages, the sender and receiver fields present a form of text that follows mostly, but not necessarily, expected sequences like entity name, street name, city. The standard allows for much freedom for populating those fields.
As a result, the information extraction, both in the MT and XML standards, remains a hard task that is often handled as a manual duty, as well as the related enrichment.
PTparser contributions:
- Entity recognition : extraction and flagging of entity types (leveraging Spacy1)
- Address parsing : extraction and flagging of address components (leveraging Libpostal2)
PTparser is available as an open source Python package3.