You may want to add additional information to your dataset. With the help of AI technology, Keatext’s Metadata Enrichment feature allows you to add additional columns of metadata that can then be used for filtering or graphing purposes.

Examples of this would be adding a field describing the customer journey steps, the departments within a company that deal with particular issues, cancellation reasons, quality issues, etc.

Our AI technology will then analyze the text in your dataset and apply a value according to your description.

Start by enabling the feature on the dataset you want to add metadata to on the Sources page. Select the three dots at the top right of the source you want to apply metadata to.

Move to the Enriched Metadata tab and slide the purple toggle to on. At this point don't worry about the dropdowns because you have not yet created your classification schema (field) yet.

Now let’s do exactly that, let’s come up with a classification schema. We do this in the management section of the tool.

Before we get started let’s look at a few definitions:

Field: The name of your classification schema - for example “Hotel Industry Customer Journey Steps”. Think of this as the label of the new metadata column in your dataset.

Value: These are the different categories in the classification schema - “Booking”, “Payment”, “Check-in” etc.

Description: Describes either the field or the value. This field is very important because our AI technology uses it to determine how to classify the text. You want to be as precise as possible in when describing you Fields and Values

There are two scenarios that are possible. Either you already have a list of categories (values) you want to use or you have an idea of how you want to classify the data but you do not have the categories.

Let's go with the first scenario. You already know the categories you want to use.

Perhaps you have a preexisting code book or classification schema that you have been using to manually categorize your data. Well, we can automate that process to save you loads of time and effort.

The enriched metadata feature has an AI-powered chatbot to help you design your classification schema. But as we already know our classification schema, we can work in two different ways, with or without the aid of the chatbot.

The most basic thing we can do is manually enter the categories (values) and the descriptions.

To manually add your categories (values) and descriptions, select the “Add metadata field” at the top left of the page.

You will first be asked to add a field and describe it. Remember the field is the name of your classification schema.Think of it as the name of a new column in a dataset. Then enter an overall description.

For example we could title it “customer feedback categories” and then in the description say what this categorization will do. “This field will classify customer feedback using the following categories: Authentication, Service, Billing, Technical Issues.”

From here, enter each category (value) and description.

The description of the field and of each value are very important and should be as clear as possible as they are used by the AI to decide which tag, if any, to apply to each verbatim.

You can use complex logic in your value description. For example you could say for the value of Service:

“Apply the tag ‘Service’ if the two following conditions are met. The text includes feedback pertaining to overall service experience. The text does not mention automated chatbot.”

And now queue the chatbot…

We can also do basically the same thing but with the help of the chatbot.

Why not make your life easy and let the chatbot write the descriptions for you! In our example, the same as above, we want to classify customer feedback using the categories: Authentication, Service, Billing, and Technical Issues.

Just type that into the field - and voilà, your descriptions will be automatically generated.

The chatbot will generate a description of the field and of the values. You can edit or delete as desired. You can also manually add more values if you wish.

If you are happy with all of your values you can select “save this field” on the panel on the right.

In the case that you don't have a preexisting code book or classification schema, but you have a few ideas of what you want to do, the chatbot can guide you.

You have a few possibilities on this page. You can choose from three commonly requested metadata enrichment options by selecting one of the purple cards.

However, no dataset is the same and the metadata enrichment options are endless. For this reason, there is an AI powered chatbot here to help you!

Simply enter a basic description of what you would like to do - how you want your categorization ideas.

For example:

You are a food and beverage producer and you would like to generate a classification of food quality issues.

You are a real estate company and you would like to generate residential property categories.

You have a hotel and resort chain and you would like to classify your data by customer journey step.

Simply type your request into the search bar and hit enter.

Your categories (values) will automatically be generated. From here you can edit, delete, or add categories as you wish.

Remember, the description of the field and of each value are very important and should be as clear as possible as they are used by the AI to decide which tag, if any, to apply to each verbatim. You can use complex logic in your value description.

If you are happy you can save the field.

If you are not happy you can further describe how you would like to categorize your data in the chatbot, now located on the right, and new results will appear.

You can have a number of different classification schemes (fields) that can be applied to one or multiple datasets.

Now we have to apply the classification schema to your dataset. We move back to the Sources page to do this.

Select the source you wish to add the classification to. Then choose the classification schema you wish to use from the dropdown on the left and the long form textual data field you wish the categorization to be applied to on the right and save.

If you want to add another, simply select “Add a field” and repeat.

Now let's move to the records page to see if your categories have been applied to the dataset.

It will take a few minutes for the values to start populating. The larger the dataset the longer it will take.

Now go back to the sources page and and choose the field you want to work with from the dropdown. This is most likely the one you just created. Choose the associated field, which will be the long form text field you want the values to be applied to.

It will take a few minutes for the values to be applied.

If you move to the records page, you will now see the new field added to the dataset. This field can now be used for filtering and graphing.

If for whatever reason you are unhappy with the fields and values you have chosen and you have already added the new metadata field to the dataset, you can still modify it.

Simply go into your field and make desired changes, whether it be adding, deleting, or editing a value. Then simply run the enriched metadata feature again.

There are two places, on the records page and on the sources page. Simply select the rerun button.

You will be given two choices when you rerun your metadata enrichment.

Apply new values: This will apply values to a record that are not already associated with them. The result could be biased towards these new associations.

Delete and replace existing associations: All associations will be deleted and replaced. The result may not be the same as previous executions.

Choose and select run. An hourglass will appear in many different locations in the app.

Simply scroll over the hourglass to see what progress has been made.

Your new metadata fields will appear like any other metadata column in your dataset and can be used for filtering and graphing.

Note that the representative comment may not show text related to the filtered value. This is because it is applying the value to the whole entire verbatim. If you drill down to the verbatim you will find text matching the value.

Below is a video that explains howe to enrich your metadata in detail.