This wizard will guide you in creating your models and will ensure that the data arrives to our servers in a format that we can process successfully.
Enter your name and description model:
Select Data Type
You can upload two types of Text Data:
Data Type: Conversations
Select Model Type
You can select one type of Text Data:
Topic Detection
This model lets you do a deep dive into your conversations to better understand what exactly customers are talking about with your agents.
You will detect topics on three nested levels:
First Level: Captures the main objects of interest of your customers. For a bank, this may be a "bank account"
Second Level: Captures which actions the customers want to take on these main objects. In the example of the bank these may show that customers want to open or close a bank account.
Third Level: A precise grouping of customer issues within the second level. Here we find specific recurring problems that customers may have in opening a bank account.
The model will, in a first step, analyze all conversations and identify which messages of your client contain his/her problem statement.
In a second step, the model then creates the nested hierarchy of topics that will be presented to you on the results page.
... After validating the data:
Data Type: General Text Data
Select Model Type
You can select one type of Text Data:
Text Clustering
Compared to conversations, general text data is more easily represented.
We only require a .csv file with one row per text item and one column with the content.
We encourage you to include further columns in your dataset.
Metadata of any type may be helpful to train better models. Also, we can cross-reference the metadata and the results to provide you with useful insights.
... After validating the data:
Finally your data is ready for analysis:
