Analyze earnings call transcripts with Watson Tone Analyzer and Natural Language Understanding

This lab uses the Tone Analyzer service to extract the most positive sentences from earnings call transcripts and then runs those remarks through the Natural Language Understanding service to extract the most relevant keywords and Semantic Roles from those sentences .

After instances of Tone Analyzer and Natural Languaga Understanding, you'll test drive yhe scenario with a test application.

1. Setup

1.1 Sign up for IBM Cloud

If you are not already signed up for the IBM Cloud, sign up here

1.2 Create an instance of the Watson Tone Analyzer Service

1.2.1 From the IBM Cloud Dashboard click on Create resource

1.2.2 Select the AI category project type and then click on Tone Analyzer

1.2.3 Make sure the Lite plan is selected and then click Create

1.2.4 Select Service credentials at the left and then click on New credentials

1.2.5 Accept the default values and click Add

1.2.6 Click on View Credentials

1.2.7 Click on the icon to copy the credentials to the clipboard and then save them in a text file on your Desktop (or some other convenient location). You'll need the apikey value later in the lab.

1.3 Create an instance of the Watson Natural Language Understanding Service

1.3.1 From the IBM Cloud Dashboard click on Create resource

1.3.2 Select the AI category project type and then click on Natural Language Understanding

1.3.3 Make sure the Lite plan is selected and then click Create

1.3.4 Select Service credentials at the left and then click on New credentials next to the credentials generated for your service instance

1.3.5 Accept the default values and click Add

1.3.6 Click on View Credentials

1.3.7 Click on the icon to copy the credentials to the clipboard and then save them in a text file on your Desktop (or some other convenient location). You'll need the apikey value later in the lab.

2 Test scenario

Test the scenario using 3 earnings report transcripts using a standalone app. Both Python and Java are provided. Choose the one that you feel most comfortable with.

Requirements

You can install Python/Java locally or use Docker

Docker

If you use Docker you can run either app without additional software installs

Docker requirements Docker for your platform

Local Java or Python

If you don't use Docker then you'll need one of the following installed locally

Python:

• Python 3.5 or later.

Java:

• Java 1.7 or later JVM
• Gradle Build Tool Note: Version 5.0 was used to verify this example

2.1 Test with a Python app

2.1.1 Edit the file settings.py in the tone-analyzer-nlu-python sub folder of this project. Put in the values of your Tone Analyzer apikey and your Natural Language Understanding apikey that you saved earlier.

2.1.2 In a command prompt or terminal navigate to the tone-analyzer-nlu-python sub folder of this project.

Docker

Run the following commands to get a bash shell in a Python environment to run the subsequent commands

docker run -it --rm -v "$(pwd):/repo" python:3.5-slim-jessie bash

cd /repo

Locally installed Python

Run the subsequent commands from your command prompt

2.1.3 Run the following command to install the dependencies for this project

pip install -r requirements.txt

2.1.4 Run the following command to run the tester application

python tone-analyzer-nlu.py

2.1.5 Verify the app runs without errors and the output looks something like the following.

Analyzing  3 earnings call transcripts ...
Analyzing transcript file name ../test_data/JPMC1Q2018EarningsCall.txt

Most positive statements from earnings call:

1) Client investment assets were up 13% year-on-year with half of the growth from net new money flows and with record flows this quarter.

NLU Analysis:
keywords: Client investment assets, record flows, net new money
semantic_roles:
subject: Client investment assets action: were object: 13% year-on-year
...

2.2 Test with a Java app

2.2.1 Edit the file settings.properties in the tone-analyzer-nlu-java/src/main/resources sub folder of this project.Put in the values of your Tone Analyzer apikey and your Natural Language Understanding apikey that you saved earlier.

2.2.2 In a command prompt or terminal navigate to the tone-analyzer-nlu-java sub folder of this project.

Docker

Run the following commands to get a bash shell in a Java+Gradle environment to run the subsequent commands

docker run -it --rm -v "$(cd ../ && pwd):/repo" gradle:slim bash

cd /repo/tone-analyzer-nlu-java

Locally installed Java

Run the subsequent commands from your command prompt

2.2.3 Run the following command to build the app

Linux/Mac

./gradlew build

Windows

gradle.bat build

2.2.4 Run the following command to run the tester application

Linux/Mac

./gradlew run

Windows

gradlew.bat run

2.2.5 (Optional) Run the following command to generate Eclipse artifacts so the project can be imported into Eclipse

Linux/Mac

./gradlew eclipse

Windows

gradlew.bat eclipse

Note: after running the command import this folder as an existing project into Eclipse

2.2.6 Verify the app runs without errors and the output looks something like the following.

Analyzing 3 earnings call transcripts

Analyzing transcript filename JPMC1Q2018EarningsCall.txt


Most positive statements from earnings call:

1) Client investment assets were up 13% year-on-year with half of the growth from net new money flows and with record flows this quarter.


NLU Analysis:
keywords: Client investment assets,record flows,net new money
semantic roles:
subject: Client investment assets action: were object: 13% year-on-year

...

Conclusion

Congratulations ! You successfully worked with both Tone Analyzer and NLU to start creating a framework to extract highlights from earnings call transcripts.