Popular Quoting Tweet Generation via Auto-Response Augmentation

Anonymous

Abstract: A quoting tweet allows users to share others' content while adding their comments. To help users write a quoting tweet with better public engagement, we study the task of popular quoting tweet generation. The focus is to generate quoting tweets with higher popularity reflected by more likes, replies, and retweets. While large language models (LLMs) showed exceptional language generation capabilities, limited work has examined how LLMs can learn the popularity of text to engage the public better. Consequently, we propose a novel Response-augmented Popularity-Aligned Language Model (RaPALM) to align language generation to popularity by incorporating insights from augmented automatic responses. Here, we employ the Proximal Policy Optimization (PPO) framework with a dual-reward mechanism to jointly explore popularity in quoting tweet generation. The experiments on two newly gathered datasets of quoting tweets for external links or others’ tweets show that RaPALM exhibits state-of-the-art results.

Paper Type: long

Research Area: Computational Social Science and Cultural Analytics

Contribution Types: NLP engineering experiment

Languages Studied: English