Translating transaction descriptions using machine learning

A system, comprising:

• a repository comprising a first source transaction description, a first target entity description, a first transaction comprising the first target entity description, a first translation accuracy metric for the first target entity description, and a plurality of golden entity descriptions;
• a memory coupled to a computer processor;
• a machine translation model executing on the computer processor and using the memory, the machine translation model comprising:
• an encoder model configured to encode the first source transaction description to create a first context vector;
• and
• a decoder model configured to decode the first context vector to predict the first target entity description;
• and
• a translation manager, executing on the computer processor and using the memory, configured to:
• generate the first source transaction description;
• generate the first transaction;
• detect an acceptance, by a user, of a first action performed on the first transaction;
• in response to detecting the acceptance, update the first translation accuracy metric the first translation accuracy metric being determined at least in part from a number of assignments of at least one category to at least one transaction comprising the first target entity description;
• determine that the updated first translation accuracy metric satisfies a translation accuracy criterion;
• and
• in response to determining that the updated first translation accuracy metric satisfies the translation accuracy criterion, add the first target entity description to the plurality of golden entity descriptions.

A non-transitory computer readable medium comprising instructions that, when executed by a processor, perform:

• generating a first source transaction description;
• encoding, by an encoder model of a machine translation model executing on a computer processor, the first source transaction description to create a first context vector;
• decoding, by a decoder model of the machine translation model, the first context vector to predict a first target entity description;
• generating a first transaction comprising the first target entity description;
• detecting an acceptance, by a user, of a first action performed on the first transaction;
• in response to detecting the acceptance, updating a first translation accuracy metric, the first translation accuracy metric being determined at least in part from a number of assignments of at least one category to at least one transaction comprising the first target entity description;
• determining that the updated first translation accuracy metric satisfies a translation accuracy criterion;
• and
• in response to determining that the updated first translation accuracy metric satisfies the translation accuracy criterion, adding the first target entity description to a plurality of golden entity descriptions.