Magic Tags
complete
C
CJ
Please create a feature where users create manual tags (like they do now), but AI goes through the transcript and automatically tags based on manual tags (similar to magic highlight)
Log In
Jazmin Taheri
complete
Magic highlight now uses your existing project tags to automatically classify and group suggested highlights. You can read more about the feature here: https://dovetail.com/help/highlight-and-tag-project-content/#automatically-surface-highlights-with-ai
A
Alisa Y
I am experiencing a bug with this feature. When I add any new tags to the existing ones, all of the previously automatically selected tags are cleared.
Pat Barlow
in progress
O
Olivia Harold
I just want to add, that I don't see this being accurate or trustworthy unless the magic tags can be trained on my own tags from the past. The magic highlights right now seem random, the transcript doesn't know what's a user speaking versus a facilitator, the clustering is random and shallow, so I'm not feeling confident that this would be functional soon.
Jazmin Taheri
Merged in a post:
Automated Tags & Fields
A
Ashley Churchill
One of the few things i miss from enjoyhq is the automated tags and properties. They would let us define some regular expressions with advanced logic and, if met, they would update the properties (aka fields) and tags accordingly.
There are certain instances where this would work really well, for example, tagging which industry or job role a participant belongs to based on the transcript content.
Having this feature goes hand-in-hand with offering more robust search logic and would help to speed up the work of classifying a high volume of data. As I work with primarily non-researchers who do their own research, getting folks to consistently tag or apply our workspace fields is a struggle and this feature would help address that issue.
"Automated properties - https://enjoyhq.helpdocs.io/article/0r8hje74tk-property-management
Properties automatically assigned by rules cannot be removed or renamed - if you want to make any changes to them, you'll have to edit the rule which assigns the property."
Pat Barlow
Thank you for posting, Ashley Churchill! I have a few more questions for you:
- Can you provide examples of the regular expressions with advanced logic that you used in enjoyhq for automated tags and properties?
- Could you elaborate on the 'robust search logic' you mentioned? What specific functionalities are you looking for?
- Can you share more about the struggles your non-researcher colleagues face when tagging or applying workspace fields? Are there specific areas where they find difficulty?
A
Ashley Churchill
Pat Barlow
- yes! So specifically i was using "rules" which you can see all the details for in their documentation here. I think it is distinct from "Magic tags" as classification is user-defined & rule-based vs more flexible & AI-based.
If someone mentioned a specific phrase in a transcript, came from a specific data source, had a particular note title, or some combination of these things, i could tag a highlight accordingly OR classify the fields associated with the note. A common one we would use is to classify highlights about particular product areas or audiences. So if someone mentioned payroll or a collection of phrases associated with payroll, it would auto-tag those highlights. Or if a data source came from NPS responses and the NPS value was below 6, it set the NPS field to “detractor”.
- In terms of robust search, I'm looking for multiple and and or statements, nested logic, the ability to use a query language even. It seems the more data we add to dovetail, the more complicated it becomes to query it effectively.
- Non-researchers using the platform have limited time and lack strong incentives to classify their data with tags and fields. Especially after introducing AI analysis elements.
To address this, my existing taxonomy is pretty simplistic and I often end up picking up the labor of tagging/classifying fields when my stakeholders do not. Automatically doing this for certain fields/tags would take that labor off of me and my stakeholders, while also opening up more complex/robust taxonomy designs.
Jazmin Taheri
on our radar
A
Ashley Churchill
+1!
I'd love if it pulled in definitions and details we supply for tags to help inform which ones to use in different contexts.
That would make for a good incentive to better document our taxonomies and analysis frameworks with definitions for key tags, but also help fuel more accurate magic tags.
A
Adriano D'Alo
This would be a great improvement!
We do record feedback received from many sources and when we import it into Dovetail we need to go through it and tag what piece of our Product the feedback is about.
In our data we record the page the users were when they submitted the feedback, so when they say "I have an issue with the table view because [blah-blah]" we know they were referring to the view of a specific page/action in our Product.
We use tags to keep track of this information and having to manually read all the feedback notes is very time consuming and prone to error. It would be great if it was automated.