Custom topics allow you to organize comments into groupings that make sense for your unique situation.
The Bellomy Text Analytics Toolkit provides three complementary approaches for creating custom topics. They are:
- Create custom topics from the term discovery tool using search terms.
- Create custom topics and apply them to comments one-by-one.
- Create custom topics and apply them using the bulk assignment tools.
These approaches are complementary and not exclusionary. You may mix and match the approach that gets you to the desired outcome.
Create Topics From the Terms Discovery Tool
From the Terms Discovery tool, click the + sign to the right of the Term picker.
This will present the “Create a Custom Topic” dialog as shown below.
Every custom topic should have a unique name (within the dataset – you can use the same topic name in differing datasets).
The key to building custom topics using this approach is the use of search terms. You will need to provide one or more search terms for the tool to use. Once provided, the tool will locate all comments that meet one or more of those search terms and apply the topic to those comments. You may also exclude comments based on specific keywords.
Let’s consider an example. Let’s create a custom topic that organizes comments about damage to a consumers yard. Let’s start by giving the custom topic a title of “Yard Damage”. Then, let’s add a few search terms. Let’s use the following search terms:
- yard + debris
- yard + damage
- yard + rut
- backyard + debris
- backyard + damage
- backyard + rut
- limbs + yard
- branches + yard
- yard + torn
Add each of those terms to the “Terms Included” list by entering the phrase in the “Enter a new term or search” box and clicking the “Add Term” button for each. You should end up with a dialog populated as shown here:
(Note: Each of the search terms shown such as “yard + debris” instructs the system to find comments that contain that specific pair of words. So, only comments that contain the word ‘yard’ AND ‘debris’ would be located for this search statement.) The terms included are OR’d, not AND’d, so any comment that meets ANY of the terms will be included. Read more about search terms below.)
Click ‘Save’. It may take a second or two for the system to locate the target comments. Once it does, you will see a dialog that looks like this:
Click GO.
A progress bar will appear that keeps you informed on the progress of the operation. Depending on how many records are involved, it may take several seconds or even a minute or two to complete. When the operation is complete, a message will appear that indicates it has completed. The terms tool will refresh and your new topic should appear.
In some cases it may take a minute or two for the system to ‘catch up’ with the newly assigned topic. You can use the refresh icon (located in the upper-right of the term tool) to refresh the term list to see updated results as they come in.
Once the topic has been processed, it will appear in the ‘Topics’ term alongside any other topics you have for the dataset. You can interact with the new topic as you would for any other topic or term. See example below.
Once created a custom topic can be edited. When editing a custom topic, you can add or remove search terms, change the name, etc. The process for finding the relevant comments is repeated each time you hit ‘Save’.
Formulating Search Terms
Search terms used for custom topics are always case insensitive (e.g. uppercase, lowercase is not considered when matching). Search terms can be any of the following:
- Simple single word searches (e.g. “yard”)
- Phrase searches (e.g. “my yard”)
- AND searches where all stated words/phrases MUST be included in the comment (e.g. “yard + debris”).
- OR searches where any stated word/phrase MUST be included in the comment (e.g. “yard OR backyard”)
What are Fuzzy Matches?
By default, search terms used by this tool require an exact match to the stated search. You can enable fuzzy matching for any search term when you add it to the terms list. When fuzzy matching is enabled, the system will locate exact matches as well as any comment that closely matches. A comment is considered a close match if it could match with only one character change to the comment. Example: The search term is “power”, fuzzy matching would also match on “powr” because one character change – in this case adding an ‘e’ would match. “Powers” would also match as would “Lower”. Fuzzy matches can be quite useful but may also introduce unwanted comments. By default fuzzy matches are not included but you may encounter situations where you want them included. If so, just check the box when adding the search term to the custom topic.
Frequently Asked Questions
Q: I made a mistake and included a search term I did not want. How can I fix it?
A: No problem. There are several approaches to rectify that concern. TODO: Click here to read about bulk updates.
Q: I have manually assigned some comments to an existing custom topic. Will changing the search terms (adding, editing or removing them) remove those manually assigned comments?
A: No. Making changes to custom topics, by default, only adds more comments to the topic. Removing comments already assigned to a topic are handled separately to prevent this situation.
Q: There are some comments included in a custom topic, that match a search criteria, but that I do not want included. How can I fix that?
A: No problem. You can remove individual comments, or bulk comments, from a custom topic. TODO: Click here to read …
Q: Is there a maximum number of comments that can be applied to a custom topic?
A: Nope. A custom topic can have as many comments assigned to it as there are comments in your dataset.
Q: Is there a limit to the number of custom topics I can create?
A: Nope. We do not enforce a limit to the number of custom comments you can create. There may be practical limits you should consider but the software will not stop you from adding more.
Q: Is there a limit to how many topics an individual comment can be assigned?
A: Nope. Topics are not mutually exclusive. An individual comment can have as many topics assigned as there are topics in the dataset.