Task Definition

Task Design

Task 1: Argument-based Sentiment Analysis (Research Report & Earnings Conference Call)

Research Report

There are three subtasks in the argument-based sentiment analysis task: (1) argument classification, (2) premise sentiment analysis, and (3) claim sentiment analysis. 

In the argument classification subtask, participants are asked to classify the given sentence into claim or premise. 

In the premise sentiment analysis subtask, participants need to classify the given premise into positive, neutral, or negative. 

In the claim sentiment analysis subtask, participants will classify the given claim into bullish, neutral, or bearish.

Earnings Conference Call

There are two subtasks with earnings conference call data: (1) Argument Unit Classification, (2) Argument Relation Detection and Classification

The goal of the Argument Unit Classification task is to classify a given argumentative sentence into one of two classes: claim or premise.

The Argument Relation Detection and Classification is a three-class classification problem. Participants need to identify the relations (support/attack/none) of the given two sentences.

Task 2:  Identifying Attack and Support Argumentative Relations in Social Media Discussion Threads (Social Media)

Participants are asked to identify the argumentative relations (attack, support, or none) between two given social media posts. 

Evaluation

We will report both Micro-F1 and Macro-F1 scores of all tasks, and use Macro-F1 to rank the results.