Skip to content

Future Work

Panagiotis Antoniadis edited this page Aug 25, 2019 · 2 revisions

Since the program lasts only three months, there are many ideas that have not been implemented and have been opened as issues in Backlog.

  • Minor extensions:
    • Speed up some procedures that can be done in multiple threads, such as email fetching.
    • Add WebSocket support.
    • Improve error detector system using phonetically close words.
    • Replay output email back to the user.
    • Add separate dictation commands to change paragraphs, make lists, etc.
    • Better vector representation for emails. Currently, the vector representation of a sentence is the mean value of its spacy word representations. We can represent a sentence by the mean value weighted by their tf-idf metric.
    • Give the user the opportunity to keep his dictations after log out.
  • Major extensions:
    • Extend the model to support not only Gmail accounts.
    • Extend the model to adapt to messages from other apps, such as Facebook, WhatsApp, etc.
    • Improve the language model's accuracy by making use of the subject of each message.
    • Integrate error corrector system in UI.
You can’t perform that action at this time.