Birds are important indicator species and obtaining better data about their distribution and activities can assist conservation efforts, and improve our understanding of their interactions with the environment and other organisms. However, traditional observation methods are labor intensive, non-scalable and error-prone. Machine learning based bioacoustic monitoring offers an appealing solution that can potentially address these issues. Most prior work on machine learning for bird song is not applicable to real-world acoustic monitoring, because it assumes recordings contain only a single species of bird, while in-field recordings typically contain multiple simultaneously vocalizing birds. This talk will describe our approaches based on the multi-instance multi-label (MIML) learning framework to address a variety of bioacoustic species monitoring tasks.