Star Trek The Next Generation Quote Identification

SI 630: Natural Language Processing, January-May 2024

In this project, I tackled the task of quote identification, a hybrid task of author identification and speaker identification, of quotes from Star Trek: The Next Generation using the natural language processing techniques of multi-label classification and sentence similarity. I found that the Hugging Face multi-label classifier was the most successful method for identifying the character who spoke a quote, with an F-1 score of 0.4516 compared to the baseline of 0.182. Additionally, I concluded that RoBERTa was a more effective tokenizer for the Star Trek: The Next Generation quotes than BERT. Though this project does not provide a reliable method for identifying the character who spoke a quote, it offers insights into which Star Trek: The Next Generation have the most unique and distinguishable voices.