The utilization of a building-block-based molecular network is an efficient approach to investigate the unknown chemical space of natural products. However, structure-based automated MS/MS data mining remains challenging. This study introduces building block extractor, a user-friendly MS/MS data mining program that automatically extracts user-defined specified features. In addition to the characteristic product ions and neutral losses, this program integrates the abundance of the product ions and sequential neutral loss features as building blocks for the first time. The discovery of nine undescribed sesquiterpenoid dimers from Artemisia heptapotamica highlights the power of this tool. One of these dimers, artemiheptolide I (9), exhibited in vitro inhibition of influenza A/Hongkong/8/68 (H3N2) with an IC50 of 8.01 ± 6.19 μM. Furthermore, two known guaianolide derivatives (16 and 17) possessed remarkable antiviral activity against influenza A/Puerto Rico/8/1934 H1N1, H3N2, and influenza B/Lee/40 with IC50 values ranging from 3.46 to 11.77 μM. In addition to the efficient discovery of novel natural products, this strategy can be generally applied to grab derivatives with specific fragments and enhance the annotation power of LC-MS/MS analysis.