Categorization workflow: Industry-specific or Custom models
The fastest, easiest way to categorize
Stephane_SimpleDecisions
Last Update 3 years ago
This approach is about categorizing datasets along a stable list of preset categories, that are specific to a given industry, in order to analyze different datasets with the same analytical grid over time.
Use cases
This workflow is suited if your dataset fits into one of the preset models available in SimpleX.
For now, there are 5 industry-specific models available in SimpleX, all pre-trained for optimal performance on proprietary labelled datasets:
- CRM services
- E-commerce reviews
- HR feedback
- Retail reviews
- App reviews
Using these models is a guarantee of high performance and consistent over time categorization of reviews, comments, provided one of the models is relevant top your datasets.
On a broader perspective, whenever you have a recurrent flow of text answers that you want to categorize always with the same coding plan, a pre-trained classifier is the way to go.
Note that the existing models can be used as a starting foundation for more detailed analysis. Indeed, once data is categorized in 4 or 5 broad topics, you can build sub-topics, using one of the other approaches.
If you have such a flow, but would like a custom model with your very specific coding plan, we can build it for you and embed it into SimpleX.
If you have already built the model and would like to leverage it within SimpleX interface, we can also connect with your API to do so.
Key SimpleX functions
SimpleX industry-specific models are only available for Entreprise Plans.
The relevant SimpleX functions are: Build topics/Categorize…
The 5 industry-specific models available in SimpleX have the following settings:
CRM services: a multi-label classification model with 5 categories: Communication, Skills, Efficiency, Pricing & Contract, General
E-commerce reviews: a multi-label classification model with 5 categories: Product Quality, Product Pricing, Delivery, Customer service, General Satisfaction
HR feedback: a mono-label classification model with 5 categories: Corporate Culture, Leadership, Organization & Process, Strategy, Work
Retail reviews: a multi-label classification model with 5 categories: Product quality, Pricing, Location, Service & Staff, Satisfaction
App reviews: a mono-label classification model with 4 categories: Feature request, General feedback, Information seeking, Problem discovery
Custom models in SimpleX can be mono or multi label and present as many categories as required.
To use a categorization model, click on the Build topics button and select Categorize…, then select the relevant model. Proprietary models appear here too, next to the industry-specific models available.
You can chose to have the value delivered by classifier (eg Product quality for a e-commerce specific review) used in SimpleX as a topic (default) or as a custom tag (if for ex. you would like to build topics along another angle and combine both views).
Selection criteria and Pros & Cons
Industry-specific or custom models are recommended when:
- you are dealing with ‘standard’ data such as retail reviews, and looking for a first solid first level of categorization
- it is critical for you to keep exactly the same categorization from one dataset to another
- you regularly deal with datasets > 3-5k
Pros:
The analysis workload is reduced to the minimum
Instant results (10k quotes analyzed in 30 sec in one click)
Cons:
None, if you have the model that suits your peculiar needs!