Annotations panel in the legacy Topic editor

The Annotations panel displays the text that is matched to a topic. Use the annotations panel to assess topic accuracy, read relevant phrases, discover new keywords, and when necessary, remove phrases associated with certain themes.

In the annotations panel, text from records is divided into smaller segments (that is, phrases or comments). However, the totals at the top show the number of records associated with the text (not the number of segments). Multiple phrases or comments can originate from the same record, so there are often more segments than records.

The annotations panel shows the number of records matched to a topic, the number of records matched correctly, the number of records matched incorrectly, the number of unmatched records that should match the topic, and the total number of records with comments (both matched and unmatched).

Tip: When multiple segments of text from a record match the same topic, those segments can only count once towards the number totals at the top of the annotations panel. For example, suppose all these conditions are true:

Medallia Experience Cloud is analyzing text at the phrase level.
The total record count in the annotations panel is 28.
The number of correctly matched records is 6.
An incoming record has three phrases that match to the Service - Office/Environment topic.

When this happens, all three phrases will appear as separate segments in the annotations panel, but the total record count will only increase by one (to a total of 29). Additionally, if you clicked the thumbs up button ( Thumbs up icon

) next to each of the three phrases, the total number of correctly matched records would also only increase by one (to a total of 7), since all three phrases originate from the same record.

Topic accuracy for phrases captured by themes

To help you understand why certain phrases are included in themes, Experience Cloud highlights the words that cause a phrase to be included. As you hover over each highlighted word, tool tips reveal the underlying concept that the Experience Cloud is applying to that word. For example, in the screenshot below, Quality of service; Amenities and programs is captured by the Club Facility theme because the word Amenities is included in the FACILITY concept.

In this example, the word "FACILITY" appears when you hover over the word "Amenities."

When you find an inaccurate phrase that is captured by a theme that you've linked to or added to your topic list, you can remove it by clicking the minus icon in the phrase:

In this example, clicking the minus icon will remove the phrase, I wish I could send money straight into a bank account in the recipient's country.

Note: You can only remove phrases from themes after you add them to your topic list or link to them from an existing topic. Also, in the case where one theme is linked to multiple topics, removing a phrase from one of those topics will not remove it from the other topics.

Topic accuracy for phrases captured by rules

For phrases that are captured by rules, it can be helpful to mark which phrases are accurate and inaccurate. Based on the number of inaccurate phrases, you can then decide whether it is necessary to modify or create new rules. Accurately tagged phrases belong in the topic and should be marked with the thumbs up button ( Thumbs up icon ). Inaccurately tagged phrases do not belong in the topic and should be marked with the thumbs down button ( Thumbs down icon ). After you mark a phrase, the phrase is highlighted in green or red.

For example, suppose there is a topic called “shoe department” and you decided to create a rule to gather customer phrases about finding shoes:

In the rule, the word find is connected to the word shoe.

When assessing topic accuracy, you would mark the following phrase as accurate because the customer is describing their experience finding shoes within the store, which is relevant to the shoe department topic.

This example comment says: "The only issue I had was that I had bad luck in finding a shoe in my size that I really liked. The two pairs that I wanted, did not have them."

However, the following phrase would be considered inaccurate because it describes an experience buying shoes online instead of in a store, so although it is captured by one of the topic’s rules, it does not fit within the scope of the shoe department topic.

This example comment says: "I liked finding shoes that were not available in the store. It is always hard to chooses, however, when you can't try them on."

Some inaccurate phrases are normal and are not an indication that something is wrong with your topic. However, it is important to monitor the percentage of records with inaccurate matches to ensure that the majority of matches are accurate. See Assessing captured comments for more information.

Filtering options

You can filter phrases into three categories: Captured, Marked But Uncaptured, and All Comments.

Captured phrase category

Phrases in the Captured category have been tagged by one or more rules in your topic. The words highlighted in orange are associated with the rule that matches the phrase.

As you hover over words from phrases in the Captured category, Experience Cloud provides hints about how they were processed:


Behavior	Description
Highlighted	The word is processed as-is. Example:
Not highlighted	The word is not in our dictionary and is considered a stop word, meaning that it would be an ineffective keyword to include in a rule. Example:
Captioned	The word was processed differently than the form of the word used in the comment. One of these circumstances applies: The Experience Cloud's spellchecker determined the word was misspelled The Experience Cloud's processed the root version of a conjugated or pluralized word The word is a user feature The word was defined as part of a pre-defined collocation The word is part of a word group Examples: Misspelled word: Plural word: User feature: Pre-defined collocation: Word group (“terrible” is part of the “#smell” word group):

If a caption appears above a word, the word was processed differently than the form of the word used in the phrase. In the following example, although the plural form of issue is used in the phrase (“issues”), Experience Cloud processed the root version of the word (“issue”).

In this example, although the plural form of issue is used in the phrase (“issues”), Experience cloud processed the root version of the word (“issue”).

All Comment category

Phrases in the All Comment category include both tagged and untagged comments. One strategy when brainstorming ideas for rules is to search these phrases for keywords related to your topic.

In this example, the user is searching for the keyword "shoes."

Marked but Uncaptured category

Phrases in the Marked but Uncaptured category are phrases that you marked as accurate or inaccurate in the All Comments category, but that aren’t captured by your rules. Phrases that have been marked accurate always filter to the top of the list.

Phrases in the "Marked but Uncaptured" category are highlighted in either green (accurate comments) or red (inaccurate comments).

Update options for the annotations panel

To refresh the rules and phrases of a particular topic, open the topic and click Update.

When you click Update, Experience Cloud refreshes the phrases assigned to your topic set. Consider refreshing the phrases in the following situations:

You haven’t logged into this topic for an extended period of time.
New phrases or dictionary terms (user features) were added to your Experience Cloud.
You changed a rule or added a new rule that is not reflected in the rule record counts.
When the Experience Cloud does not automatically update the record counts.

There are three options for updating the phrases assigned to your topic set: fast, balanced, or precise.

The Experience Cloud updates the phrases according to these options:


Option	Description
Precise	Loads all the phrases. Precise is the default option.
Balanced	Loads a moderate amount of phrases.
Fast	Loads a small subset of phrases. Choose Fast if you have a very large data set, and are experiencing system slowness.