User features
User features enable Text Analytics parsing for terms that would otherwise be ignored
User features are terms that are added to the Medallia Experience Cloud dictionary because, otherwise, they would be ignored by the parser. Often, they are specific to your company and your company’s industry, such as product names, brand names, or frequently used combinations of words. Once a user feature for a word is created and the feedback is parsed, Experience Cloud can search for the exact word or phrase defined within the user feature.
Properties
- Name
- Descriptive name. A best practice is to use the same name in the Canonical form, and the Name properties. The name should be unique (not the same as a normal English word).
- Language
-
Option
Description
Global
The user feature is accessible to every language you build topics in.
• If you plan to build topics in one language only, there is no need to select Global
• If you are building topics in multiple languages, only enter user features as Global that are shared across languages. User features such as brand or product names can usually be entered as Global.
Note: Language-specific user features are parsed before global user features. In the case that both a language-specific and global user feature are a match to the same text, only the language-specific user feature is applied. One way to get around this is to build every user feature in each language and abandon global user features entirely. However, that may require you to rebuild the same 10 user features across 10 languages.A specific language
The user feature is only accessible when you build topics in that specific language.
- Definition
-
Define the user feature by using a regular expression. The regular expression tells the system what to look for when identifying a match for your user feature.
Keep these guidelines in mind when writing regular expressions:-
Use the /b word boundary to retrieve whole words only
- When possible, avoid using wildcards to ensure that the system picks up only the exact words you intend
- Always test your regular expressions before adding them to the system Note: Use an online tool (such as http://regexr.com/) to rest your regular expressions. As you test, keep in mind that some online test tools have different rules for regular expressions. In Medallia Experience Cloud, these rules apply:
-
OR expressions must be enclosed in parentheses. For example, hotel | inn will not work unless there is parentheses around the clause: (hotel | inn).
-
Negative look-ahead expressions (?!word) are permitted.
-
Negative look-behind expressions (?<!word) are permitted.
-
Here are some examples:User feature name Regular expression Part of speech DUE_TO \bdue\sto\bNoun follow_up \bfollow(s|ing|ed)?[-\s]?up\bVerb ON_TIME \b([Oo]n[\s-]?time)\bNoun -
- Canonical form
-
The user feature identifier.
In order to reference a user feature from a topic rule, use the text from the Canonical form field. A best practice is to use the same text in the Canonical form and Name properties. The following images show the Canonical form field for a user feature in Setup and a Topic Builder rule with the canonical form of a user feature as a component in Admin Suite.
Important: The Canonical form for a user feature should:- Be unique (not the same as a normal English word).
- Be all lowercase.
- Only include alphanumeric characters (a-z,1-9), underscores(_), and hyphens (-).
- Have 3 or more characters.
- Not begin or end with an underscore.
- WordType
- The part of speech that applies to the user feature.
When to add user features
When a keyword does not capture any feedback containing a certain word, Medallia Text Analytics might be identifying the keyword as a stop word (see Stop words section in Standard topics for more information about stop words). Sometimes, Text Analytics might not have the word in the dictionary at all. In some cases, it might be necessary to create a user feature for that keyword so the system recognizes it.
A list of available user features can be viewed from the Edit Topics screen, as shown in the following image.
Typically, user features are industry or company-specific words and phrases (“Google,” “wifi,” “GPS”). However, user features can also be necessary for commonly used phrases, such as “deal with,” that contain stop words. Stop words are words (often conjunctions and prepositions) that the parser ignores during the parsing process. When you need to capture an exact phrase with a stop word (such as “deal with”) you must create a user feature for that phrase.
To determine whether a user feature is needed, search for the term or phrase in the a report with searchable comments to ensure that the term or phrase is actually used. If the term or phrase is found in your search results, double-check the syntax you are using. If the syntax is correct, a user feature might be needed to pick up the term or phrase. However, if the term or phrase is not found in your search results, you probably do not need it in your topic.
If you identify a term or phrase that needs to be added to the system dictionary, add it as a user feature. Note that new user features are not available until the feedback in the system is reprocessed.
-
Nothing prevents user features from overlapping with each other and other normal word taggings, so you must verify that there is not an existing user feature that matches the same string.
For example, suppose you have two user features, one for Orion and another for Orion Rewards. In the regular expression for the Orion user feature, you need to include a negative look-ahead for the word rewards (?!\srewards). Otherwise, all instances of Orion Rewards will match the Orion user feature.
User feature Regular expression Orion \borion(?!\srewards)\b Orion Rewards \borion\srewards\b - Consider how the new user feature will impact topic building. For example, if you want a topic to capture instances of both Orion and Orion Rewards (as defined in the previous example), you would include both user features in the rules for that topic. If you included the Orion user feature but not the Orion Rewards user feature, you would not capture any instances of Orion Rewards.
-
Consider which normal dictionary words will lose matches if you create the user feature. For example, suppose you have a user feature for the phrase over the phone. Because that user feature exists, a keyword rule for phone will overlook any instances of the over the phone phrase, resulting in fewer matches.
Add user features in Setup
To add a user feature in Setup:
-
Navigate to the Reporting > Text Processing > User Features screen.
-
Under Filter by language, select the language (or languages) to which the user feature applies. The options are outlined below:
Option
Description
Global
The user feature is accessible to every language you build topics in.
• If you plan to build topics in one language only, there is no need to select Global
• If you are building topics in multiple languages, only enter user features as Global that are shared across languages. User features such as brand or product names can usually be entered as Global.
Note: Language-specific user features are parsed before global user features. In the case that both a language-specific and global user feature are a match to the same text, only the language-specific user feature is applied. One way to get around this is to build every user feature in each language and abandon global user features entirely. However, that may require you to rebuild the same 10 user features across 10 languages.A specific language
The user feature is only accessible when you build topics in that specific language.
-
Click New.
-
Enter information for the following properties. See Properties, above, for more information.
- Name
- Definition
- Canonical form
- WordType
-
Click Save.
Note: New and modified user features are not available until you re-process your comments. For more information, see Text Analytics processing and initialization. - After you have added user features and re-processed your comments in a sandbox, test them before adding them to a production environment.
-
From the Reporting > Text Processing > Topic Sets screen, create a fake topic called Test User Features.
-
Add each user feature as a single keyword.
-
See which comments are captured by the user feature and confirm it returns the results you expect. If you do not see the results you expect, update the regular expression in the Definition property.
-
Add user features in Admin Suite
To add a user feature in Admin Suite:
-
Navigate to the Text Analytics > User Features screen.
-
Select Create. The Create UserFeatureDb screen displays.
-
Enter information for the following properties. See Properties, above, for more information.
- Name
- Language
- Definition
- Canonical form
- WordType
-
Select Save.
Note: New and modified user features are not available until you re-process your comments. For more information, see Text Analytics processing and initialization.
The following image shows the Create UserFeatureDb screen in Admin Suite.
