DataTales: A Benchmark for Real-World Intelligent Data Narration

We introduce DataTales, a novel benchmark designed to assess the proficiency of language models in data narration, a task crucial for transforming complex tabular data into accessible narratives. Existing benchmarks often fall short in capturing the requisite analytical complexity for practical applications. DataTales addresses this gap by offering 4.9k financial reports paired with corresponding market data, showcasing the demand for models to create clear narratives and analyze large datasets while understanding specialized terminology in the field. Our findings highlights the significant challenge that language models face in achieving the necessary precision and analytical depth for proficient data narration, suggesting promising avenues for future model development and evaluation methodologies.

DEKGCI: A double-sided recommendation model for integrating knowledge graph and user-item interaction graph

Both knowledge graphs and user-item interaction graphs are frequently used in recommender systems due to their ability to provide rich information for modeling users and items. However, existing studies often focused on one of these sources (either the knowledge graph or the user-item interaction graph), resulting in underutilization of the benefits that can be obtained by integrating both sources of information. In this paper, we propose DEKGCI, a novel double-sided recommendation model. In DEKGCI, we use the high-order collaborative signals from the user-item interaction graph to enrich the user representations on the user side. Additionally, we utilize the high-order structural and semantic information from the knowledge graph to enrich the item representations on the item side. DEKGCI simultaneously learns the user and item representations to effectively capture the joint interactions between users and items. Three real-world datasets are adopted in the experiments to evaluate DEKGCI's performance, and experimental results demonstrate its high effectiveness compared to seven state-of-the-art baselines in terms of AUC and ACC.