This talk addresses an important problem in the OTA (Online Travel Agency) industry, namely property mapping. The problem asks for matching same property entities (hotels, rooms, etc.) across different domains and channels in order to ensure a consistent user experience on the platform. This task is traditionally performed manually by expert users who screen for salient information in the structured data such as property name, address, number of rooms, and other features. With the emerging business model of NHA (Non-hotel Accommodation) and the sharing economy, such as Airbnb and the like, the mapping problem is further convoluted, because as opposed to traditional hotels, most of the property information is user-generated (e.g. images, and natural language textual description) and highly unstructured. Thus, the manual process of mapping is further complicated and requires more time and effort from expert users. On the other hand, the task of automated mapping, which has been traditionally attempted by rule-based algorithms, becomes ineffective and erroneous with low accuracy and coverage. In this talk, I will briefly survey the efficient big data pipeline used at Agoda to automate the process of NHA property mapping, as well as outlining some of the important machine learning problems and solutions used in this pipeline.