Setting Up Text Analytics Variables With One Loop For Each Variable

For each question in the survey that is to be analyzed using Genius, a loop variable containing a set of single, multi and grid variables, and a metadata folder containing a PushedDate and a ProcessedDate variable, must be created.

Note: In the survey on which you wish to run text analytics, ensure the following:

- English (a non-country-specific version) must be set as one of the languages for the survey, otherwise when reporting, the categories might be displayed blank.

- You cannot have an underscore (_) in the name of a Qid that is to be used for Text Analytics. If you must use Text Analytics on a Qid that includes an underscore in its name, you must set up a Data Processing rule to pull these comments into a new Qid that does not have an underscore.

1. If you do not have access to the Text Analysis functionality, contact Forsta Support and ask them to enable the Text Analytics permissions.
2. Once you have access to the functionality, log in to Authoring and select the survey to which you want to add the Text Analytics module.

1. Create the reusable scale for sentiment (go to Creating the Sentiment Scale for more information).
2. Add the categorization model list (the Database Designer table).
3. Add a folder for the metadata (go to Creating MetaData for more information).
4. Create the variables in the survey.
5. When you have added the loop and metadata folder as described, launch the survey.
6. Run the Genius task .

If everything is set up correctly, the task will run. The data will be transferred to Genius, it will be processed, and the results will be transferred back to the Forsta survey.

Tip: If you have a previous example of the Text Analytics folder in another project, go to that project in the secondary tab and copy the full folder (i.e. loopreference and the meta data folder) into the current project. Make sure that you rename all variables and set up the database designer tables correctly and in full before launching the survey.

For each open text question that is to be analyzed, there must exist in your survey a specific set of folders and variables to receive the processed data. Note that these variables must be set as Hidden so they are not displayed to the respondents. The setup procedure below describes only the open text question and the folders and variables required for the analysis functionality.

Important!
The variables must be set as Hidden Variables.

Note: The qid must not include the underscore character (_) because multis using loop reference answer lists cannot use qids with that character.

For each of the open text questions you want to run text analytics on, you have to do the following:

1. Create a folder for the model and name the folder <qid> - <TextAnalytics> - <Model no.>.

Note that a new folder with the required loop object and variables will be required for each model.

2. In the folder you have just created, create a single variable and name it <qid><model no>HierarchyReference. In its properties, set Answers to Hierarchy lookup and Variable type to Hidden - .



Figure 1 - Example of the Hierarchy question

3. In the Answers tab, set the Schema name and Table to be the schema name and table of your categorization model that you have set up in Database Designer. Note that you should use the 2nd tab ‘Categories and Attributes’ model for the HierarchyReference variable, as this will enable you to use the Discovery Analytics widget for Text Analytics.
4. Create a loop object and name the loop <qid><model nr.>CategoriesLoopReference. The Loop Object is used as the answer list in the multi variables, allowing the table lookup to be used in the multis. Open the Properties for the object and set the type to Table Lookup, and set the schema and table you want to use - .



Figure 2 - Setting the Loop properties

5. Create a Single variable and name it <qid><model nr.>OverallSentiment. Set it to Hidden, and specify the SentimentScale reusable answerlist as its answer list (go to Creating the Sentiment Scale for more information).
6. If reporting in Reportal or Discovery Analytics:
• Create a Multi variable and name it <qid><model nr.>Categories, and in this variable's Answers tab click Add Loop Reference and specify the loop as its answer list.
• Create a Multi variable and name it <qid><model nr.>PositiveMentions, and in this variable's Answers tab click Add Loop Reference and specify the loop as its answer list.
• Create a Multi variable and name it <qid><model nr.>NegativeMentions, and in this variable's Answers tab click Add Loop Reference and specify the loop as its answer list.
• Create a Grid variable and name it <qid><model nr.>CategorySentiment. In this variable's Answers tab click Add Loop Reference and specify the loop as its answer list, and in its Scale tab, specify the 11-point SentimentScale reusable scale/list (go to Creating the Sentiment Scale for more information).
7. If reporting is to be done in Studio, create three single, Boolean variables <qid><model nr.>CategoriesSec, <qid><model nr.>PositiveMentionsSec and <qid><model nr.>NegativeMentionsSec with answers true and false, and a single variable <qid><model nr.>CategorySentimentSec with the SentimentScale reusable scale/list (go to Creating the Sentiment Scale for more information).

If setting up both for Reportal/Discovery Analytics and Studio, the Survey Tree should look as in the example below - , with the Open text question why and the folder why - TextAnalytics - Model 1042 with its loop object and variables. You can drop the variables specific to Reportal/Discovery Analytics if reporting in Studio only, and vice versa.

Figure 3 - Example of the Survey set up for processing

Note: You can also copy a folder from an existing project that has the Text Analytics Loopreference already setup. Be aware that if you do this you cannot launch the survey until after you have changed the loopreference name; problems will arise if you attempt to launch before you have made the changes.

Only launch the survey after you have set everything up properly. Ensure that the name of the loopreference has the name of the open end comment that you’re analyzing, spelled and written EXACTLY the same. Note that it is case sensitive – it must be identical, then the model ID number and the rest of the name CategoriesLoopReference.

If you have copied from another survey, remember to ensure that you have referenced the correct Table Lookup and that you have renamed each of the variables correctly, including the Meta data folders.