The following was created with TensorFlow Text and Google Colab to generate text using a character-based RNN. TensorFlow and other libraries were used to download a dataset of Sherlock Holmes radio dramas from the 1930s and 1940s, process the text, build and train a model, and generate the new text below based on the model’s predictions.
Title: Sherlock Holmes and the Subsiding Scuffle
WATSON: I've searched and examined everywhere, but I still haven't found Santa Claus.
HOLMES: What about his hat, Watson?
WATSON: No I have not found his hat.
HOLMES: Come on Watson, we need his hat. If you rub it 5 times, your foot turns to stone and he appears under the table.
WATSON: You're a scoundrel Holmes, and a rat, a silly boy and a bell-ringing slippery pencil manager.
HOLMES: Please sit my dear goose.
WATSON: What? On this canoe?
HOLMES: Yes, thank you Mrs. Rush, I just want to ask you a couple of questions. Did you steal the diamond and ruby chicken?
WATSON: Holmes, of course not... do you want a biff on the nose?
HOLMES: No, I want to sing Good King Wenceslas.
WATSON: You mean Good King Wenthettle of the Octagonal Square?
HOLMES: No, I mean Mrs Crabapple of the Vinegar Police.
WATSON: Holmes, do you have the Christmas measles by chance?
HOLMES: Of course, how ever did you know?
WATSON: Well you've soiled your father's hand, broken your wife and have a French accent.
HOLMES: Well done my dear sir.
Comments