Being able to analyze sentiment at the entity level in textual data presents a significant opportunity for businesses.

Github repository of this project: Link text

What does it mean to analyze sentiment by entity?

It involves extracting the sentiment associated with each entity mentioned in textual data, going beyond a general “Positive” or “Negative” label — although this overall sentiment is also important.

Visual Example of Entity Sentiment Analysis (by Author)

Below, I explain more about Sentiment Score, Sentiment Magnitude, and Salience.

The business value with that?

1. Customer Feedback Analysis

Analyzing customer reviews or surveys to identify sentiment about specific entities like products, services, or customer support teams.

Example: Determining if customers are happy with the delivery process (entity: “delivery”) versus the product quality (entity: “product”).

2. Brand and Reputation Analysis

Monitoring social media or news articles to understand public sentiment about a company or its competitors.

Example: Identifying if negative sentiment is associated with a specific entity like a CEO, a product, or a recent campaign.

3. Competitor Analysis

It can be done with the data available on social media, for example.

Analyzing sentiment toward competitors’ products or services to gain competitive insights.

Example: Extracting and comparing customer satisfaction between your company’s smartphone brand and a competitor’s.

4. Financial Sentiment Analysis

Analyzing news or reports for sentiment on entities like companies, stocks, or sectors.

Example: Tracking how investors feel about “Tesla” or “renewable energy” in financial discussions.

5. Human Resources

Analyzing employee feedback to assess sentiment toward specific organizational entities, such as teams, policies, or leadership.

Example: Identifying dissatisfaction related to “remote work policies.”

These cases contribute to decision-making, customer satisfaction and retention, as well as reputation management.

An Easy approach to do that:

An easy approach is to use the Google Cloud Natural Language API to perform this analysis.

I completed a project where I created an API key on GCP to call the endpoint using a Python script.

You can check out my project on GitHub, which includes all the necessary dependencies.

Additionally, below you can test the application of this project on Streamlit Cloud:

Link to streamlit Link text

(Wait 1 minute to load the application)

Once you click, you will receive three values for each captured entity:

Sentiment Score: Indicates the positivity or negativity of the sentiment (-1 to 1).
Sentiment Magnitude: Reflects the intensity of the sentiment (0 and above).
Salience: Represents the importance of the entity in the text (0 to 1).

For future analysis, the data can be stored in a CSV file locally, Google Sheets, Firestore, or other storage solutions.
See below:

Data extracted and saved into a CSV file.

Conclusion

This application provides a straightforward way to extract analytical insights from customer feedback using Natural Language API. This approach becomes even more powerful when dealing with large volumes of data that require detailed analysis.