Extraction de réseaux dynamiques conversationnels par lissage narratif

Abstract : Modern TV series tend to develop complex, multiple, parallel plots, spanning several seasons. However, they are typically watched on a very short time period. Viewers therefore sometimes feel the need to consult a summary before watching a new season. However, such summaries are generally produced manually, which is expensive. This work is the first step in the proposition of an automatic summarization method. We model the plot of a modern TV series through the dynamic conversational network of its characters. Instead of aggregating these relations over the whole considered period or over a sequence of time windows, we propose narrative smoothing, a new approach more appropriate to the complex plots of such series. It automatically determines the temporal scale of the network based on the dynamics of the character interactions. We present the preliminary results obtained on three popular current series: Breaking Bad, Game of Thrones and House of Cards.
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01385215
Contributor : Vincent Labatut <>
Submitted on : Monday, October 24, 2016 - 5:10:56 PM
Last modification on : Friday, March 22, 2019 - 11:34:07 AM

Files

liss_narr.pdf
Files produced by the author(s)

Licence


Distributed under a Creative Commons Attribution - NonCommercial - ShareAlike 4.0 International License

Identifiers

  • HAL Id : hal-01385215, version 1

Collections

Citation

Xavier Bost, Vincent Labatut, Serigne Gueye, Georges Linarès. Extraction de réseaux dynamiques conversationnels par lissage narratif. 7ème Conférence sur les modèles et l'analyse de réseaux : approches mathématiques et informatiques (MARAMI), Oct 2016, Cergy-Pontoise, France. ⟨hal-01385215⟩

Share

Metrics

Record views

191

Files downloads

583