Labels and Tags

Question

One method of finding information within the community is through labels and tags. A project is underway to reorganize the Employee Spaces category. This project provided an opportunity for us to review the use of labels and tags within the community as a whole (see Jira story, COM-187)

A few notes about labels and tags:

* Labels are used to categorize the first content message and can only be added by the author of the content.
* Tags are used to categorize any post and can be added/removed by any community member.
* Khoros recommends using pre-defined labels. Pre-defined labels cannot be shared across nodes. Each node would have to have its own set of pre-defined labels.
* Labels and tags within the Employee Spaces is being reviewed by @JonesieAtJMP  and @fiona_kelly 
* In the Discussion category, of the 10,000+ topics, 3 topics have labels and 690 topics have tags.
* @ChrisCostin analyzed the text within discussions to determine potential labels that can be used.
* Tags are used by Lingotek for translated content.

My recommendations for moving foward with labels and tags:

* Use pre-defined labels only
* Use the results of Chris' project as the basis for the labels in customer categories (e.g. Discussions, Wish List, etc.)
* Review flow of adding labels to a topic via the UI
* Review use of filtering content by labels in the UI
* Investigate programmatically labelling topics
* Only allow staff and super users to tag content (initially, with the potential for expanding to others)
* Review current tags used within the Discussions forum and adjust accordingly (i.e. remove, consolidate, etc.)

Ryan_Gilmore · Answer

For 2020, there were close to 1300 tags used in messages across all boards within the Community. Discussions and JMP Blogs were the boards with the most tags.

Since 2016, there are just over 4400 tags used in messages across all boards within the Community. Almost 3000 of the tags are used only once.

I view tagging as a more granular method of subsetting the content than labels. The advantage of tags is that it spans all boards. However, too many tags can be overwhelming potentially rendering minor ones useless. It is also helpful to have consistency among the labels, e.g. JMP 16 Pro vs. JMP Pro 16, scripting vs. JSL.

The tag word cloud is the main component for viewing information related to a tag.

Initial proposal in moving forward with tags:

* Remove tags that are only used once
* Consolidate similar tags
* Restrict adding of tags to staff and super users with the intent of expanding to higher level ranking members
* Review remaining tags to determine potential list. Verify list with super users.

Ryan_Gilmore · Answer

@LandraRobertson , @Wendy_Murphrey , @Duane_Hayes , and @Ryan_Gilmore meet to discuss the text to be used for the labels. Technical Support was invited to get their perspective on the wording since they interact frequently with customers.

There was general agreement on the following:

* Data Table
* Analytics
* Scripting
* DOE
* Graphics
* Other

There were concerns about Other becoming too populated. When we have enough data, we will analyze the Other category and expand if needed.

A user has the option to choose more than one.

Posts regarding Dashboard/Application Builder, Query Builder, file importing, etc. may not be easy to categorize.

We read a few existing posts to categorize based on the labels selected. While some questions are not necessarily scripting related, a script was given as a solution.

Only the owner of the topic is allowed to label. We may want to give certain staff/super users the ability to modify the labels.

Ryan_Gilmore · Answer

An attempt was made to determine what labels to use within the Discussions area (and most likely Blogs and Wish List areas too) (JMPCOM-11).

The following data sources were used:

* Messages in the Discussions area between 2018-2020
* Topics/subtopics for JMP used in Sirius
* Defect components for JMP
* Pre-defined labels used in JMP Wish List
* Main menu items in JMP

The following are the results of using Text Explorer/Latency Class Analysis on the messages from the Discussions area using 2000+ terms, a minimum term frequency of 100, and 5 clusters:

Cluster1

Cluster2

Cluster3

Cluster4 Cluster5 Term

Score

Term

Score

Term

Score

Term

Score

Term

Score

calculate

0.1226

column

10.193

script

10.232

model

4.8902

graph

5.5348

pro

-0.171

table

7.3356

file

4.9529

analysis

3.7828

plot

5.2238

remove

-0.222

columns

6.1466

table

4.2519

test

3.3579

graph builder

4.5729

p

-0.289

rows

4.1276

open

3.9491

factors

3.2343

change

3.3102

text

-0.394

script

3.9949

window

3.4885

design

2.9353

x

3.186

pass

-0.435

new

3.7383

code

3.3302

results

2.8212

script

3.136

version

-0.455

value

3.7088

run

3.1653

question

2.7171

y

3.0149

legend

-0.467

row

3.2049

save

3.0242

factor

2.6662

show

2.7342

ideas

-0.472

values

3.1432

user

2.4876

variables

2.4231

make

2.5478

numbers

-0.489

formula

2.3068

error

2.472

run

2.3767

values

2.334

I tried various combinations of settings for analysis with the results being similar to those above.

I loosely mapped the clusters as follows: Data Table (Cluster 2), Scripting (Cluster 3), Analytics (Cluster 4), Graphics (Cluster 5), Other (Cluster 1).

If you increase the number of clusters, an argument could be made to include a mapping for DOE. As the number of clusters increased, the terms overlapped so that cluster X was very similar to cluster Y. I found 5 to provide the most distinct separation of terms.

These are essentially the pre-defined labels used for the JMP Wish List. They also align with the main menu in JMP (Tables, DOE, Analyze, Graph). The JMP User Community Newsletter uses Data Visualization, DOE, Statistics/Predictive Modeling/JSL.

So, how many labels should we have? The following articles were used as reference:

* Categories vs. Tags
* The Best Ways to Organize Your Files with Tags and Labels
* The Right Way to Organize Your Blog with Categories and Labels

Most agree that you do not want too many (20 or less). The purpose is to help users find information without having to select several options. Labels should be at a higher level. A message should have at least 1 label but not more than 3-5.