Conversational Aspect-based Sentiment Quadruple Analysis

Q: Detailed information about the certificates?

A: The top 3 teams will be awarded certification from NLPCC and CCF-NLP. For more information, please visit the official NLPCC website: NLPCC Website

Q: Detailed information about paper submission?

A: The top 3 teams will be invited to submit papers to the NLPCC conference. The submitted papers will undergo a review process, and if they meet NLPCC's publication acceptance standards, they will be published in the Shared Task Track of NLPCC 2023.

Q: Can you explain the rule "prohibit the use of large models with more than 1 billion parameters for training and prediction?"

A: This means that the parameter count of a single pre-trained language model used in your model cannot exceed 1 billion. There is no limit on the parameter count of your own model.

Q: How many labels are there for the polarity of the quadruple?

A: Our annotated dataset includes five labels for polarity: positive, negative, neutral, doubt, and ambiguous. Since instances with the last three labels are minimal, we merge them into a new category called other for training and evaluation. Overall, we use three label types in our training and evaluation: positive, negative, and other. You can find more details about this process in our released preprocessing and evaluation code.