March 21, 2023


Moving Forward

How to Use NLP for Smarter Investment Decisions: Two Practical Applications | by Stefano | Feb, 2023

News co-occurrence network and sentiment distribution

Data is everywhere, and in the world of finance, it’s king. As the saying goes, “knowledge is power,” and nowhere is that truer than in the stock market. That’s why the rise of natural language processing (NLP) is such a big deal — it’s giving traders and investors the ability to extract valuable insights from a sea of information.

In this article, we’re going to dive into the applications of NLP and finance. We’ll explore the role of sentiment analysis and news aggregation in finance and show you just how powerful NLP can be. Whether you’re a seasoned trader or just starting out, this article will give you a comprehensive look at the state of NLP in finance and its potential for the future.

Sentiment Analysis

You’ve probably heard the term “sentiment analysis” before, but what exactly is it? Simply put, sentiment analysis is the process of determining the emotional tone of a piece of text, such as a news article or social media post. In finance, sentiment analysis is used to gauge the overall mood of the market, predict stock performance, and analyze the impact of news events on stock prices.

There are two main methods of sentiment analysis in NLP: rule-based approaches and machine-learning approaches. Rule-based approaches rely on a set of predefined rules to classify the sentiment of a text. For example, a rule-based approach might classify words like “good” and “great” as positive, and words like “bad” and “terrible” as negative.

On the other hand, machine learning approaches use algorithms to learn patterns in the data and make predictions based on those patterns. These algorithms are trained on large datasets of annotated text and can be much more accurate than rule-based approaches.

One applicable example of sentiment analysis is FedNLP, which analyses the FED communications and speeches by their members to try to asses whether they will raise, maintain, or drop rates in the next decision. It also provides a summary and sentiment factors. Here is an example of the result provided when analyzing a speech by governor Christopher J. Waller

Results of the Demo version of the FedNLP model

News Aggregation

This is the process of collecting and organizing news articles from multiple sources into a single, unified view. This could be done manually, but if it was efficient, I would not be writing this article. In the world of NLP, news aggregation is done automatically using algorithms. There are two main methods of news aggregation in NLP: web scraping and API-based approaches.

  1. Web Scraping: Web scraping involves using code to extract information from websites. In the case of news aggregation, web scraping can be used to collect news articles from multiple sources and organize them into a single view. This can be done using Python libraries such as BeautifulSoup.
  2. API-based Approaches: API-based approaches use APIs provided by news sources to collect news articles. This is often faster and more efficient than web scraping, as the APIs are specifically designed for data extraction. However, API-based approaches typically require access to the API and may have usage limitations.

The following is a replicable example that analyses the sentiment on The process involves web scraping the headlines from a news website using the BeautifulSoup library in Python and then performing sentiment analysis on each headline using a tool like the TextBlob library. The sentiment values of each headline were categorized as positive, neutral, or negative, and the distribution of these sentiments was represented in a pie chart with green for positive, light blue for neutral, and red for negative. The chart was created using the Matplotlib library in Python and displays the percentage of news headlines that have positive, neutral, or negative sentiments.

Below I present the code and the results (remember to pip install nlt and pip install matplotlib )

import requests
from bs4 import BeautifulSoup
import nltk"vader_lexicon")
import matplotlib.pyplot as plt
from nltk.sentiment import SentimentIntensityAnalyzer

# Define the URL of the news source
url = ""
# Send a request to the URL
response = requests.get(url)
# Check if the request was successful
if response.status_code == 200:
# Parse the HTML content of the response
soup = BeautifulSoup(response.content, "html.parser")
# Find all news articles on the page
articles = soup.find_all("h3")
# Print the headlines of the news articles
for article in articles:
headline = article.text
print(f"Headline: headline")
print("Failed to retrieve news articles. Response status code:", response.status_code)
except requests.exceptions.RequestException as e:
print("Failed to retrieve news articles. Error:", e)

# Create a SentimentIntensityAnalyzer object
sia = SentimentIntensityAnalyzer()
# Analyze the sentiment of each headline
for article in articles:
headline = article.text
sentiment = sia.polarity_scores(headline)["compound"]
print(f"Headline: headline")
print(f"Sentiment: sentiment")

headlines = []
sentiments = []
# Analyze the sentiment of each headline
for article in articles:
headline = article.text
sentiment = sia.polarity_scores(headline)["compound"]
# Count the number of positive, neutral, and negative headlines
positive = sum([1 for s in sentiments if s > 0])
neutral = sum([1 for s in sentiments if s == 0])
negative = sum([1 for s in sentiments if s < 0])
# Create a pie chart of the sentiment distribution
labels = ["Positive", "Neutral", "Negative"]
sizes = [positive, neutral, negative]
colors = ["green", "lightblue", "red"]
plt.pie(sizes, labels=labels, colors=colors, autopct='%1.1f%%')
plt.title("Sentiment Distribution of News Headlines")

Sentiment Distribution of headlines on Seeking Alpha on February 5th, 2023

The use of NLP in finance is not without its hiccups. The quality of the data we use for our analysis can make or break the accuracy of our findings. If we’re not careful, algorithms can also introduce their own biases, potentially leading us down the wrong path. Additionally, let’s not forget the financial lingo — it’s like a whole different language, loaded with technical terms and jargon. So, while NLP in finance has plenty of potentials, it’s important to approach it with a critical eye and a dose of skepticism.

Final Thoughts

NLP has been a game changer in the field of finance, providing new and innovative ways to analyze and interpret financial information. From automating tedious tasks such as news aggregation to providing a more nuanced understanding of public sentiment, NLP has shown that it can be a valuable tool for investors, traders, and analysts alike. By approaching NLP with caution and a critical eye, we can continue to unlock its potential and stay ahead in the ever-evolving world of finance.

It’s also worth noting that NLP is not just a tool for financial professionals, but has the potential to democratize access to financial information. By automating the process of information gathering and analysis, NLP can empower individuals and communities to make informed decisions about their finances. As NLP continues to develop and mature, it will be exciting to see how it will shape and transform the financial industry in the years to come.

A robot (very) carefully analyzes human language