Shohreh Haddadan

I graduated in Computational Linguistics at Sharif University of Technology, Tehran, Iran, where I defended my thesis titled: “Persian Abstractive Summarization using graph-based Abstract Meaning Representation” in January 2017. My Bachelor degree I gained at Ferdowsi University of Mashhad in Software Computer Engineering.

Before I started my Master’s degree, I developed professional programming skills while working in an artificial intelligence based software company in Tehran for five years. I am now looking into “Argumentation Mining” in my PhD project.

Applying NLP Techniques Towards a Complete Argumentation Mining System in Political Debates

How are political arguments structured? How are we automatically able to detect, extract and analyze these arguments? How are argument structures evolved through time or how they vary across different topics? In my research I focus on argument mining of political debates between the presidential candidates in the United States elections since 1960.


From top to bottom: Pragmatics, Semantics, Syntax, Morphology. From top to bottom: Pragmatics, Semantics, Syntax, Morphology.


The first challenging step I encountered was choosing an annotation scheme and annotating the data for further use of machine learning algorithms. The annotation scheme needs to be chosen carefully in order to be able to answer the research questions on the argument structures. A micro-level argument annotation has been applied to the dataset, extracting argumentative discourse units and relations between them. The next challenges are trying to find methods and features which can best capture the various argument structures in the political discourse based on the assumed annotation scheme. Are the machine learning methods capable of capturing and generalizing the structure of argument made by the politicians and apply them on new debates? Is the data actually representative of the broader domain so that we can make assumptions on the extracted argument structures in the case of a high accuracy machine learning model? How can extracted argument structures help political and social scholars in detecting logical concepts in arguments such as fallacies?

Main Supervisor: Prof. Dr Leon van der Torre

Other supervisors: Dr Serena Villata (I3S research centre in Sophia Antipolis)


INCEpTION: A Semantic Annotation Platform

Shohreh Haddadan battles her way succesfully through Argument mining in the political domain

Deep Learning for Language Analysis Summer School

Annotations, Motivations and Methods

Creating Concordance Tables in Python