Maritime vessels around the world risk encountering dangerous situations during their voyages such as piracy, kidnappings, hijackings, and smugglings. Such dangers may put crews at risk and potentially cause extensive damage to both the vessel and its cargo. Maritime risk assessment is a process of analyzing such factors to positively impact the safety and timeliness of operations. Analyzing the high-risk areas in seas and oceans can help vessels route around areas with previously recorded high levels of dangerous activities. This information is scraped and extracted from various maritime documents, including newspaper articles, social media, incident blogs, and operational reports. After textual data sources are parsed, it is possible to use this extracted information as training datasets into a Machine Learning (ML) model using Natural Language Processing (NLP) Deep Learning (DL) techniques. With these methods, we can perform document classification, risk extraction, text summarization, topic modeling, and information extraction in an automated manner. Methods that are used include the bag-of words (BOW) approach and, in particular, the use of Named Entity Recognition (NER) to identify specific areas around the world that are considered a risk to a particular voyage. Once this information is processed, the model can determine and aggregate specific areas of high-risk activity. We propose a Maritime Event Log (MEL) pipeline using DL methods. This pipeline includes binary document classification (DC) to extract positive instances of incidents, an incident type classification (IC,) and an information extraction module (IE) including date and location and ship type extraction. With such information available, it is possible to avoid dangerous situations, which may have a significant level of impact on the safety and success rate of maritime operations.