Astock - Leveraging NLP for Automated Stock Trading

Before diving into the details of Astock and its automated trading strategy, it's essential to understand the context and motivations behind its development. The background section provides insight into the challenges and opportunities in the financial industry that led to the creation of Astock, setting the stage for understanding its significance and impact.


Astock is a tool that helps trade stocks automatically by understanding news articles about specific companies. We made it because we wanted to create a better way to make investment decisions using technology. Unlike other tools, Astock focuses on three important things:

  • It gives you news articles about each company, so you can make informed decisions.
  • It provides different details about each stock to help you understand them better.
  • It measures how well the tool works using important financial measures.

We use a smart method called Semantic Role Labeling Pooling (SRLP) to make sense of news articles quickly. This helps us predict how a stock might behave. We also use a special learning strategy to make our predictions even better.

What We Did

We created a system where the tool automatically decides to buy or sell stocks based on news articles. If the news suggests a stock will go up, it might decide to buy (long). If the news suggests a stock will go down, it might decide to sell (short). If the news isn’t clear, it might choose to do nothing (preserve). We make sure to close the deals the day after the next trading day to stay flexible.

To control how much we buy or sell, we use a special formula based on recent returns and how likely the news is to be accurate. We also calculate a pretend return rate for when we choose to do nothing.

Technologies Used

Challenges Faced During Model Training

Training our model came with its own set of challenges, including:

Data Complexity: Understanding and processing financial news data can be tricky due to its complexity and varying formats.

Algorithm Design: Designing an algorithm that accurately predicts stock movements based on news required careful experimentation and optimization.

Overfitting: Preventing our model from becoming too specialized to the training data and not performing well on new data was a challenge.

Computational Resources: Training a sophisticated model like ours required significant computational resources, which posed logistical challenges.

Deciphering the Financial Puzzle: Imagine putting together a puzzle where each piece represents a different part of financial data. It’s like trying to solve a mystery using scattered clues. We had to figure out how to make sense of all this data, which wasn’t always easy.

Cracking the Stock Prediction Code: Predicting how stocks will move is a bit like solving a puzzle with a hidden code. We had to come up with a way to crack this code and make accurate predictions. Sometimes, we hit dead ends and had to get creative to find the right solution.

Avoiding Overcomplication: Our model needed to be just right—not too simple, but not too complex either. It’s like walking on a tightrope; we had to find the perfect balance. If our model got too complicated, it would get confused and make mistakes. But if it was too simple, it wouldn’t work well either.

Using Enough Power: Training our model required a lot of computing power, like having enough fuel to power a rocket. We had to make sure we had the right tools and resources to handle all the data and calculations involved.

Following the Rules: Just like in any game, there are rules to follow in the world of finance. We had to make sure our model followed these rules and regulations. This meant working closely with experts to make sure everything was done correctly and ethically.

Key Features

Astock offers several key features that set it apart from traditional stock trading platforms:

Tailored Financial News: Astock provides users with up-to-date financial news specifically related to individual stocks, enabling informed decision-making based on current market trends and events.

Comprehensive Stock Factors: The platform offers a wide range of stock-specific factors, allowing users to gain a deeper understanding of each stock’s performance and potential.

Advanced Predictive Modeling: Leveraging advanced Natural Language Processing (NLP) techniques, including Semantic Role Labeling Pooling (SRLP), Astock predicts stock movements based on analyzed news articles, enhancing the accuracy of trading decisions.

Dynamic Trading Strategy: Astock’s dynamic trading strategy automatically triggers buy (long), sell (short), or hold (preserve) actions based on predicted outcomes from the NLP model, optimizing trading performance in response to market fluctuations.

Risk Management Mechanisms: The platform incorporates risk management mechanisms such as position closure to prevent significant losses and ensure the safety of trading operations.

Transparent Evaluation Metrics: Astock evaluates its performance using comprehensive financial metrics, providing users with transparent insights into the effectiveness of the automated trading algorithms.

Featured Images


Our system performed really well in tests. It beat other methods in making money from trading stocks. We also made sure our system was safe by closing deals if we started losing too much money.


Astock shows how technology can help us make better decisions when trading stocks. By understanding news articles and using smart algorithms, we can make better choices and hopefully make more money. Our tool is available for anyone to use, and we hope it helps others make smarter investments too.