How many times a certain word is mentioned during a company's earnings call is often discussed on financial news such as Robinhood Snacks Daily. For example, "Mirrors" appeared 40 times while "Lululemon" was only mentioned 25 times by the executive team, indicating Mirrors is the focus of investors and long-term strategy for customer engagement.
These earnings call transcripts provide invaluable insights. However, it is quite challenging (tedious) to extract valuable information from these transcripts, conduct analysis, compare transcripts from multiple companies and multiple years, and observe future trends.
The goal of this project is to make earnings call analysis automated.
Function 1: Get Words Frequency by Choosing A Company and Year
Function 2: Followed by interesting keyword counts in the previous step, now we will pull sentences from the earnings call transcript to understand more context.
Function 3: The Word Trends.
Due to Covid-19, the hospitality industry was hit hard. The keyword in 2020 is "recovery". China was mentioned 86 times by the Marriott executive team and analysts in 2020's earnings call.
Function 4: The Sentiment
Based on previous analyses, recovery is the focus of 2020. Next, we are going to explore the sentiment related to recovery. Two possible scenarios: 1) people may think the recovery will be slow and the situation will be unstable, or 2) people have a positive view of the future due to the fast recovery in China once Covid-19 is under control.
Screenshot of some sentences' sentiment scores:
The overall sentiment score distribution:
Overall, the company executives have a very positive view of recovery and potential performance rebounce.
Firstly, we have collected the earnings call transcript pdf links on companies' official websites.
Then, we downloaded and saved each file as PDF, and consolidate all PDFs into a .pkl file.
The final data structure looks like this... Each row will have one quarter's earnings call transcript from one company.
Python (request, NLTK, pandas, ipywidgets, pdfplumber)
For more information, please check out the Project Portfolio page.