Using nearly 8000 10-K documents published in 2016 and 2017, we generate contextual vectors through artificial neural networks and test whether the language of 10-K documents, without any detailed numeric indicators of financial performance, correlate with earnings per share and other financials of the S&P 500. We find significant correlation between earnings per share and contextual vectors, concluding that semantic analysis is a valuable tool that has great potential in financial analysis.
The author has given permission for this work to be deposited in the Digital Archive of Colorado College.
Colorado College Honor Code upheld.
Includes bibliographical references.
The author has given permission for this work to be deposited in the Digital Archive of Colorado College.
Colorado College Honor Code upheld.
