/DSER_Movie_Subtitles

Dialog Segmentation, Speaker ID Mapping and Emotion Recognition in Movie Subtitles

Primary LanguageJupyter NotebookMIT LicenseMIT

Dialog Segmentation, Speaker ID Mapping and Emotion Recognition in Movie and TV Subtitles

HCI Project (E1 slot) : Prof. Joshva Devadas T
Prishita Ray- 17BCE2405
Arushi Das- 17BCE0087

Link to paper: https://www.aclweb.org/anthology/W18-5021.pdf
Link to dataset used for testing the models: https://www.cs.cornell.edu/~cristian/Cornell_Movie-Dialogs_Corpus.html

Our Aim:

  1. To find situations where turn of dialog can occur based on a set of defined rules for a pair of dialog acts
  2. When turn of dialog occurs, map speaker id to it, if not obvious keep it as unknown
  3. Recognize emotion from the entire dialog sequence in each movie being tested and predict

Steps involved in Natural Language Processing:

  1. Sentence Segmentation
  2. Word Tokenization
  3. Predicting Parts of Speech for Each Token
  4. Text Lemmatization
  5. Identifying and Removing Stop words
  6. Dependency Parsing
  7. Finding Noun Phrases
  8. Named Entity Recognition
  9. Coreference Resolution (resolving ambiguity with usage of pronouns)