Claims billing and coding is non-trivial for health care providers. Accurate coding can help medical providers get reimbursements that they deserve for their professional services. Meanwhile, incorrect coding (e.g. up-coding) is considered by authorities to be one of the most important frauds with severe penalties. Therefore, accurate coding is of great importance to medical professionals. However, claims coding is challenging. Besides the knowledge of the E/M coding system, accurate coding requires an adequate depiction of patient health conditions and treatments, part of which are contained in unstructured clinical notes, e.g. discharge summaries and physician notes. We aim to develop a coding decision support system by leveraging state-of-the-art natural language processing (NLP) techniques and algorithms. The expected result of the project is to build an effective system that can extract essential information for claims coding from real clinical narratives. This phase-1 study compared five popular existing NLP software in named entity recognition based on 108 public available transcribed medical discharge summary notes from MTsamples.com. Qualitative comparison finds that CLAMP, Amazon Comprehend Medical, and cTAKES are more powerful. Quantitative analysis shows that CLAMP is more accurate and efficient than Amazon Comprehend Medical. Future work includes integrating a section segmentation tool before NER recognition as well as testing and implementation of the system in a clinical scenario.