Risk Tool

The Risk Tool is a powerful search engine into the text of earnings calls

Overview

The purpose of the Risk Tool is to find discussions about user-specified topics in earnings calls. Users can specify simple keywords or craft complex queries to identify the relevant parts in the text of earnings calls. Based on these text snippets, as well as the surrounding language about risk and sentiment, the Risk Tool then generates several quantitative metrics for each earnings call, which can be downloaded and used offline for further analyses.

The tool is based on the methodology developed in a series of academic papers by the founders of NL Analytics.

Specifically, for each earnings call, the Risk Tool calculates the following quantitative metrics:

• Exposure. The number of sentences that contain at least one keyword from the query.

• Risk. The number of sentences that contain at least one keyword from the query and also a synonym for risk or uncertainty.

• Positive Sentiment. The number of sentences that contain at least one keyword from the query and also a positive keyword.

• Negative Sentiment. The number of sentences that contain at least one keyword from the query and also a negative keyword.

• Sentiment. The difference between Positive and Negative Sentiment.

Please see here for more details about the definition of these metrics as well as the list of risk synonyms and sentiment words.

All metrics are collected in a CSV file that can be downloaded after a query. The CSV file has one row for each earnings call and includes, in addition to the above metrics, a number of useful variables that facilitate linking these data to other data sets. The purpose of this file is to serve as a starting point for systematic downstream analyses, including academic papers and market reports.

In addition, the tool's matched text provides the basis for a sentence level analysis in the Snippet Tool.

Example

After typing in the following query

"AI", LLM, artificial intelligence, chat gpt, deep learning, language processing, learning algorithms, machine intelligence, machine learning, natural language, neural network, reinforcement learning, supervised learning, unsupervised learning, GPT, ChatGPT, Chat GPT

the tool provides the below overview of the search results, several CSV files to download, as well as a time series and across-sector aggregate view of the three metrics.

The results of the search are shown in graphs for the three measures risk, sentiment and exposure.