本研究致力於對動態權重的對話網路研發新的自動化抽取方式。我們首先提出一種關於此類型網絡的通用辦法，並將其應用於文學小說。基於該應用，本研究提出了一種新的自動話語歸因方法。最後，我們開發一款簡易的網路視覺化工具，藉由此工具，我們可以針對所抽取的特徵網路，進行更深入的分析。;Due to their importance in humanity culture, literary works have been extensively studied during the course of history. In those works, characters and their relationships often play a central role. The study of the structure of those relationships is the study of character networks : that is, a special kind of graph that can be used to represent these structures.
Due to the importance of dialogue between characters, one can extract a specialised kind of network : a conversational network, extracted using only dialogues between characters. Using tools from graph theory or other fields of computer science, those networks can be studied to reveal original insights unattainable fromtraditional literary analysis.
This work is dedicated to the automatic extraction of dynamic signed conversational networks. We propose a general method to extract those kinds of networks, that can be used in any type of work. We then show an example where we apply this method on novels in particular, which makes us propose a new technique for automatic utterance attribution. Lastly, we create a simple example of software allowing the visualization of extracted networks to analyze them.
