NLP based Text Summarization of Fintech RFPs

In today’s day and age, where huge quantities of textual data are generated every second, it has become difficult to keep ourselves abreast with new information. Documents in the financial sector tell a quantitative story. The qualitative or language content that accompanies financial statements is an essential component of the information set used by financial market participants for monitoring and stewardship. This necessitates the development of efficient technological methods for leveraging the existence of these massive amounts of textual data. Reading financial documents such as annual reports is extremely time-consuming and thus corporations have to allocate precious human capital to understand, analyze and derive insight from these reports. As a result, automatic summarization techniques can make this task simpler by enabling access to a smaller but informative portion of a given document. In this work, a system utilizing NLP techniques for summarizing financial documents based on queries provided by the user, is presented. This system overcomes the challenges posed by existing section-based summarization models. As machine learning methods have proven to be effective on downstream tasks such as text summarization, the intent of this work is to take advantage of these methods for effective extractive summarization and generate human-like summaries. The proposed model categorizes phrases according to their relevance by leveraging the strength of unsupervised clustering approaches, and it gets a ROUGE-1 score of 46%.