Professional Services Case Study

Sentiment Analysis of 10-K Filings for Public Companies


Public companies file annual reports, known as Form 10-K, with the Securities and Exchange Commission (SEC), containing both qualitative and quantitative information about their operations and performance. However, the value of the text beyond the numerical data and its impact on the market remain uncertain. Additionally, analyzing these reports poses challenges due to the ambiguity of positive sentiment measures in a business context, where positive words may convey negative information. Simultaneously, when dealing with alternative assets, obtaining accurate and up-to-date data, such as stale values, presents a significant challenge.


First Rate Professional Services tackled both challenges by implementing innovative solutions. To extract valuable information from 10-K filings, we developed the ArtIE (Artificial Intelligence Engine) using recurrent neural networks (RNN). The RNN model utilizes vector representations of words to classify sentiment as negative, positive, or neutral. By employing RNN with Long Short-Term Memory (LSTM) architecture, they addressed issues related to complex dependencies and gradient problems during model training. The sentiment analysis model was trained using 8,000 manually labeled sentences randomly selected from 10-K filings.

Additionally, for alternative asset analysis, First Rate Professional Services implemented an integrated network that combined data from the client's internal systems, Fund Managers' Capital Account Statements, and real-time updates. This approach allowed for the generation of comprehensive reports with infographics, providing insights into portfolio performance.


  • The sentiment analysis model effectively classified sentiments expressed in 10-K filings.
  • With an accuracy of 91%, the model demonstrated its ability to categorize sentiments.
  • Valuable insights were extracted from annual reports beyond numerical data, improving understanding of performance, risk assessment, and decision-making based on sentiment analysis.

Furthermore, the integrated network solution for alternative asset analysis produced notable outcomes:

  • Accurate and up-to-date data from various sources were collected and processed.
  • Complex analytics were generated to enhance portfolio analysis.
  • Comprehensive reports, complemented by infographics, provided visual representations of portfolio performance.


First Rate Professional Services successfully addressed the challenges associated with extracting valuable information from annual reports and enhancing alternative asset analysis. By leveraging advanced technologies, such as recurrent neural networks and LSTM architecture, we accurately classified sentiments in 10-K filings, enabling financial professionals to make informed decisions based on comprehensive insights. Simultaneously, the integrated network solution facilitated the collection of accurate and timely data for portfolio analysis, generating complex analytics and comprehensive reports with visual representations.

Overall, First Rate Professional Services' implementations empowered financial firms to unlock valuable insights, improve risk assessment, and make data-driven decisions. By combining advanced technologies with comprehensive data analysis, they provided a holistic solution for extracting information from annual reports and enhancing portfolio analysis in the dynamic landscape of the financial industry.