Joining data from unstructured data to Structured data

Hello Team,
I have an unstructured text data with only reviews from customers and I have another dataset with structured data of customer details. Now I had thought that Squirro can help me by automatically joining these data based on ID (key) and bring in reviews by one customer id under that particular customer in a single dataset.
Can you please suggest how can this be done or recommend the relevant course to learn this.

Kind regards

1 Like

Hello Sagar, great to hear from you!

Thanks for the question. Please allow me to restate your problem to make sure I am understanding correctly:

  • You have a structured dataset of customer information, which contains a ‘customer_id’.
  • You have an unstructured data set of customer reviews. This data set ALSO has a ‘customer_id’ for each review. (Please see below if this is not the case).
  • You would like to label customer reviews with the relevant data found from the structured dataset by using the customer_id.

NOTE: For all these options, we are assuming you have some knowledge of creating Labels and adding them to your data set in Squirro

If your unstructured data set DOES have a customer_id:
Option #1:
If you have coding experience, you can build a custom pipelet that will run when data is loaded in. This pipelet can then attach the desired metadata to each item as it is indexed. This will lead to the most flexibile solution, as there is much you can do with these tagged items.

Option #2:
You also can use the data as-is and build a dashboard that shows data from both sources. For example, one widget to show customer information and the other to show every review related to that customer. In this case, you would use the label as a query, like ‘customer_id: 1’ (for example)

If your unstructured data set does NOT have an available customer_id
Option #3:
We have a built in process for this called Known Entity Extraction (KEE). With this, you can enrich a set of unstructured data with a relevant set of structured data. We are happy to show you a few examples below.

We hope these answers bring you closer to your desired solution. Please feel free to write again or share more details in case you have more questions!

Have a great day,
Hubert Shon, Solutions Engineer