Skip to content

Latest commit

 

History

History
69 lines (42 loc) · 3.98 KB

README.md

File metadata and controls

69 lines (42 loc) · 3.98 KB

AI Story Completion - Do we get a happily ever after?

In this repository, we try to use the recently published 117M model of GPT-2 to continue a story. We fine-tuned the model for three particular use-cases:

  • Short Stories
  • Game of Thrones (books)
  • Essays

Generated Samples

Short Stories

Sample Generated Short Stories

(The first sentence is the input to the model.)

Essays

Sample Generated Essays

(The text in gold is the input to the model.)

Game of Thrones

Sample Generated GOT Text

(The text in gold is the input to the model.)

Datasets Used

Model Training Dataset Evaluation Dataset
Short Stories ROCStories (80%) ROCStories (20%)
Essays Paul Graham's Essays ASAP AES
Game of Thrones First 5 books Sample chapters from 6th book

Performance

We used 3 criteria to evaluate the trained models:

  • Perplexity: We created a unigram model for each of the datasets and calculated the perplexity of the generated text
  • Cosine Similarity: We calculated the cosine similarities between the average vectors (Word2Vec) of the generated text for a seed and the actual continuation in the corpus. For Game of Thrones, we trained our own Word2Vec on the books.
  • Discriminative Classifier: We trained a Logistic Regression classifier to classify between the generated text and the human-written text. Note that 50% accuracy of the classifier is the best case, as it indicates that the classifier is unable to distinguish between the generated text and the human-written one.
Model Perplexity Cosine Similarity Accuracy (of Discriminative Classifier) F-1 (of Discriminative Classifier)
Short Stories 807.0305 0.5393 0.4901 0.4999
Essays 606.3497 0.4760 0.6666 0.6382
Game of Thrones 312.3814 0.4787 0.4901 0.5517

All of the modified code is present here

If you want to see a nice poster which summarizes our findings, you can find it here

Guide to Important Source Files

Final Report

You can read a more comprehesive report on our methodology and findings here