Sequence-to-Sequence Modelling and Learned Embeddings

Many financial NLP tasks map one ordered input to another ordered output: normalise a ticker alias to a canonical symbol, translate a short research headline into a risk tag sequence, or align two time-indexed feature streams of different lengths. Sequence-to-sequence (seq2seq) models pair an encoder RNN that reads the input with a decoder RNN that generates the output step by step. Embeddings replace sparse one-hot identifiers with dense vectors learned jointly with the task, so semantically related tokens sit near one another in representation space.