In this article
Overview
Forsta Visualizations allows for open-ended comment data to be analyzed in the Forsta Plus text analytics tool, providing additional insights. By analyzing the text responses, you can identify categories and sentiment scores, which can be reported on in StoryTeller dashboards or explored further in for example the Crosstab tool.
The text analytics process can be initiated manually or automated through a job that triggers when new data is loaded into Forsta Visualizations.
Note: Applying text analytics on data in Forsta Visualizations requires a user account/license in Forsta Plus.
Prerequisites
To apply text analytics to any open-ended variable in a Forsta Visualization project, you first need the following:
A hub
A custom data table
A text analytics dataflow in Forsta Plus.
After finalizing the setup in Forsta Plus, you create a text analytics job within the project in Forsta Visualization. This job will send the data to the custom data table in the hub for analysis, either on demand or automatically when new data is imported.
Once the data has been analyzed in Forsta Plus, it will be sent back to Forsta Visualizations and will be ready for reporting.
Setup in Forsta Plus
1: Create a Hub
Create a Hub to be used for storing the data and running the text analytics dataflow.
2: Create custom data table
In Custom data, create a new table. No specific table name is required but the following columns must be used:
RespondentID
Date
[Question code of question to be analyzed]
For example, if you want to analyze an open ended variable with the code "Q22", the column names should be RespondentID, Date, Q22.
To create the table structure for the above example import an Excel file with the following format during creation of the table in Custom data.
Note: The default text length for open text in a custom table is 255 characters. You should change this to the appropriate value to meet your specific needs. Please refer to the example below.
3: Create Text Analytics Dataflow
In the Automation page create a new Text analytic dataflow.
As source in the dataflow select the custom table created in step 2.
Add the open ended variable field to “Selected fields”, see image 1 below.
Select model and set processing to “Continues”, see image 2 below
Image 1:
Image 2:
Note: If multiple open ended variables should be analyzed you need to set up a custom data table and text analytics dataflow for each variable.
For more information, please read the documentation for Forsta Plus on how to setup Text analytics (Genius) and how the Dataflows works.
Setup in Forsta Visualizaitons
Create External Data Service Jobs
In your Forsta Visualizations project go to External data service page and create a job for each open ended variable to be analyzed in Forsta Plus.
To create a new job, click the Create new button and follow the steps below.
Name the job
Set the job status to Active by ticking the check box to be able to run the job later on (only active jobs can be run). A job that should be paused/not run can be set to inactive.
Select Forsta Plus Text analytics as provider.
Select the variable to be analyzed.
Filter date- Apply a filter to send a subset of the data for analysis only if necessary. This can be useful when different text analytics models need to be used for different touchpoints in a CX survey. (optional)
Enter the base url for the Forsta Plus instance you use, for example https://ws.euro.confirmit.com..
Enter the ClientId and ClientSecret.
Select the Hub, custom table and text analytic data flow created in Forsta Plus
-
Save.
Note: If you do not have your ClientId and ClientSecret contact Forsta Plus helpdesk.
Note: If you do not have your ClientId and ClientSecret contact Forsta plus helpdek.
Manual Job Execution
To run the job, manually enter the job and the Run/Re-run tab shown in the image below.
In the Run tab you select which data batches you want to send to for analysis.
Select the desired data batches and click the Run button to start the process.
When starting the job the status is shown in the bottom of the screen.
The job is completed when the new data batches with analyzed data has been imported and activated in Forsta Visualizations, a successful message is shown when the process is
completed.
The time the process takes is dependent on the data volumes: the more data the longer time the analyzes and data activation takes.
The below image shows an example of the status of a running process:
You can also always check the status of currently running jobs and previous run jobs in the Log tab.
The below image shows the Log tab, which can be used to check the status and any potential error messages if the process fails.
Automation of Job Execution
The jobs can be run automatically on data batches imported via the Import scheduler. All the Text Analytics jobs set to active can be selected when creating an Import scheduler job.
The selected jobs will be run automatically as the last step in the import process after the survey data has been imported and activated.
The below image shows the option for selecting the Text Analytics job to be run automatically when importing data via the Import scheduler.
Note: If the same respondent is sent to your Custom Table more than once from Forsta Visualization using the same RespID, the existing record is overwritten. This may cause any related analysis in Plus to run again for that respondent, depending on your flow configuration. Updated results are then reflected in Forsta Visualization.
Data Batches and Formats
The analyzed data imported to Forsta Visualization is imported as two different batches:
A “flat” batch with overall sentiment scores for the whole comment
A looped batch with categories and sentiment score per category
These data batches are displayed in the Data page in same way as all other data sets, see example below.
The flat data batch contains two questions:
an open ended question showing the text that was analyzed
a scale question with data for the overall sentiment score.
The below image shows these variables in Questions page and also the data of these in a Respondent table in StoryTeller.
The looped data batch contains five questions:
a single choice question showing the code of the analyzed variable
a single choice question showing the category
a scale question showing the sentiment score of the category
An open ended question showing sentiment label (optional)
An open ended question showing is deleted (optional)
Below you can see the flat questions marked in green and the looped questions marked in blue, both in the question page and in a respondent table.
Note: The data batches containing the analyzed text analytics data always include data for all respondents that have been sent to Forsta Plus for analysis. After each import, all previously imported data batches with text analytics data are deactivated and only the latest one is activated. This process is automated, as well as the deletion of deactivated Text Analytics data sets.