Abstract:
The amount of data is increasing in the financial domain as in every field. It is humanly
impossible to analyse all these data efficiently, so we need optimised computational techniques to achieve this. In this work, We aim to figure out an efficient technique to collect relevant news articles from web sources and extract valuable pieces of information from them. We started by focusing on extracting six important financial events from news articles. We devised a method that could achieve a fair accuracy in these six event types. More work is to be done to integrate additional financial events and enhance accuracy. We believe that this work can serve as a foundation for future developments and research.
Description:
Recent advancements in computational power, Artificial intelligence (AI), Deep Learning
(DL) have increased the efficiency and reliability of Natural Language Processing (NLP)
applications. Nevertheless, the state-of-the-art models require a large amount of labelled data for training. The lack of sufficient labelled data remains a significant challenge in the financial domain. Creating annotated training data sets is one solution, but it is highly labour intensive and not a cost-effective method. Therefore it is essential to figure out a suitable analysis method that requires less data for training. This thesis work is carried out in Algoanalytics Pvt Ltd, Pune. Under the supervision of Dr Anirudha Pant and Prashant Rane.
