Auto-tagging survey responses
Use the built-in automatic tagging to categorize open-ended survey questions as they come in. This is especially useful for long-running surveys.
See Survey text analysis: tagging for additional information.
Start by tagging the first 100 or so respondents. Then look at which tags are used the most. Are you often applying a tag to the same word or phrase? Then that is a candidate to be auto-tagged.
Setting up auto-tagging
To start auto-tagging, complete the following steps:
- Open your survey.
-
Select Analyze > Text analysis.
-
Select the question you want to analyze.
-
Select Auto-tagging from the cogwheel icon in the top right corner.
- Find the tag you want to auto-tag and click on the plus sign.
- In the keyword textbox, enter the word or phrase that should trigger the tag.
- If sentiment is on, then select the sentiment the keyword should trigger.
- Click Add.
- Click Activate to start the auto-tagging process.
Keyword matching options
Auto-tagging does not only have to look for one word. Often a combination of words is important or maybe if another word is mentioned then you do not want the tag or sentiment applied. By entering your keywords and phrases in a certain syntax, you can change how the system applies the tags and sentiment.
| Match type | Special symbols | Example keyword | Example match | Tag |
|---|---|---|---|---|
| single word | none (case insensitive, accent sensitive) | recommend | I'd recommend them for anyone. | promoter |
| multiple words, in order | plus sign between the different words (any order, other words may appear before, after or in between) | late+delivery | My delivery arrived super late! | late_delivery |
| multiple words, exact match | none (must match exactly, other words may appear in front or behind) | no difference | There was no difference to me. | no_preference |
| entire response | [keyword] (keyword must match the complete answer.) | [smell great] | Smell great | odor |
| do not tag | -keyword (this will overrule other keywords for this tag) | -not recommend | I would not recommend this product. | (tag will not be set) |
| wildcard for 0 or more characters | * (wildcard for any combination of letters or numbers within the same word) | recommend* | I have recommended them before, and would do it again. | promoter |
| wildcard for 0 or 1 character | ? (wildcard for any letter or number within the same word.) | offer? | They have great offers. | sales |
Wildcard matching
To reduce the number of keywords you need to assign to a tag, you can add a asterisk (*) or question mark (?).
- An asterisk stands for zero or more characters.
- A question mark stands for zero or one character.
Look at the single word match example from the above table. The keyword is 'recommend'. If a respondent enters, 'I recommended you several times already', the keyword would not cause the tag to be applied. To catch that response and other suffixes, change the keyword to 'recommend*', without the apostrophes. That would also catch 'recommend', 'recommends' or 'recommending', etc.
You can also place the asterisk (*) or question mark (?) at the beginning of a word to match any prefixes or in the middle of a word.
Automatic process
Once auto-tagging is active, the automatic process will first auto-tag all existing responses and then run every hour and automatically tag new open responses to this question. In the survey activity log, you can see when the last batch was processed and how many open answers were tagged. As a shortcut, click the 'last run' link on the Auto-tagging tab on the text analysis page.
Reviewing auto-tagged open answers
It is recommended that you review the open answers that have been auto-tagged. To do this, select 'to be approved' in the tag status dropdown. Check all items in the list to confirm that they are correct. Click the checkmark to approve the auto tag. If it is incorrect, fix it and then update the auto-tagging.
If there are many responses that are incorrectly tagged, fix the auto-tagging keywords and then have it rerun to process the responses again. That way, you do not have to manually fix them.
