In Zurvey.io, you can create custom labels in three ways.
1. Manual Creation on the Portal
2. Uploading from an Excel file
3. Creating & Managing During Analysis
4. Analysing Data with Custom Labels
1. Manual Creation on the Portal
To create custom labels one by one, go to the custom label section on the portal.
Click on the "New Custom Label" button.
After selecting the New Custom Label button, you will be faced with the following panel:
This panel consists of the following:
- Custom label name: this will be the name of the custom label, and it is determined by the user. This parameter is mandatory.
- Category: you can create categories to group the custom labels, thus creating a hierarchy. This parameter is optional.
- Code: while the name of the custom label is created by the user, the code will be generated by the system, however it can be modified by the user (this is especially useful in SPSS outputs).
- Tags: you can attach tags to the custom labels. This is not a category-like grouping, instead it helps find custom label sets via search. For example, project names can be used as tags, this way you can see which custom labels are used for each project. This parameter is optional.
- Language: you can choose the language from a predetermined list. This parameter is mandatory.
- Synonym set (include at least one of): This parameter is mandatory, as synonym sets primarily determine if a custom label is a hit in a text or not. You can add synonyms to the custom label one by one or in bulk. Adding them one by one requires pressing ENTER after each synonym, while bulk adding means pressing ENTER only when all synonyms have been typed in, separated by commas. The custom label will generate a hit only if any of its synonyms match the beginning of a phrase in a text. For example, the word import will be a hit not only for import, but also for imported, importer, important or importance. Of course, it may not be the desirable result, but to avoid this, you can opt for an exact match, using the special character ]. By adding the synonyms import], imports, imported and importer, the custom label will be a hit in case of the word import and the suffixated forms determined above. We can be sure that imports and imported are strings that are not followed by any other suffixes, while by not adding ] to the end of importer, we allow the word importers to be a match. It is also advisable to be prepared for misspellings and different forms, such as login, log-in and log in. Synonyms are not case sensitive, so only including lowercase forms is alright. Accented forms are equal to their non-accented counterparts, but not vice versa. This means that the synonym café will be a hit if cafe is in the text, but only adding the synonym cafe will not be a match for café. Despite the above examples, a synonym does not have to be only one word, a synonym can also be a phrase. If so, space-separated synonyms will behave like the above mentioned one-word synonyms, while using the AND operator will have a more complex result. If we use the AND operator (the word AND in capital or the characters && or the character ;) between two or more parts of a synonym, it means that every part of the synonym must match the beginning of a phrase in the same part of the text (subsentence as default - more on that in Text splitting) to be a hit. For example, the synonym answer AND complaint will be a hit for answered my complaints and my complaints have never been answered. This can also be combined with exact match, like answer] AND complaint.
- Exclude set (do not include): exclude phrases work exactly like synonyms (meaning matching the beginning of a phrase, the use of exact match, not being case sensitive, accented forms being equal to their non-accented counterparts and the use of the AND operator). However, determining exclude phrases is optional. If an exclude phrase is present in any part of a text, it means it can never be a hit for the custom label regardless of what the synonyms are.
- Add new language phrase set: with this option, you can create multilingual custom labels. It is useful for multilingual datasets (more on that in Language recognition and mixed language analysis), where you apply the language recognition function. The verbatims will undergo language recognition and if the recognized language has a corresponding phrase set within the custom label, that language’s synonym and exclude set will be applied to the verbatim itself.
Read more about this feature on our blog:
Receiving feedback both in Latvian and Estonian? No problem!
2. Uploading from an Excel File
If you have a list of custom labels, it's more efficient to upload them as an Excel file. The Excel file should include columns for custom label name and synonym sets, and optionally, categories, tags, and exclude sets.
Example: Uploading a list of product-related custom labels from an Excel file
When you have a whole custom label set ready, uploading the whole set as an Excel file is more desirable than adding them one by one on the portal. The Excel file must contain the following columns: custom label name and synonym set in at least one language. Category, tags and exclude set are optional columns. If the custom label is multilingual, each language’s synonym and exclude set has to go in separate additional columns. When the Excel file is ready, it can be uploaded by clicking the Import Custom Labels button. Doing so will lead to the upload screen:
Each column’s type can be selected from a predetermined list, thus the order of the columns is not important in the Excel file itself.
3. Creating & Managing During Analysis
While analyzing text data, you can create custom labels on the fly. One way is to select parts of the text in the Feed view and select Add as a synonym or Add as an exclude phrase from the popup menu. In the following example, we want to add the misspelled MiFi word as a synonym to the custom label WiFi.
After selecting the Add as a synonym… option, you are faced with the following menu where you can choose from the existing custom labels already applied to the analyzed data, or you can also create a New Custom Label.
However, this is not finalized until you select the Reanalyse button that appears at the bottom of the page.
Another way to create Custom Labels is by using automatically recognized labels and their synonyms. You can select one or multiple labels to work with. When all the desired labels are selected, you once again need to click on the three dots next to any of the selected labels and choose ‘Custom label actions based on selected label(s)’.
After selecting, a popup appears where you can decide whether a new CL should be created based on selected label(s), or if you want to utilize the synonyms of the selected label(s) as additions to an existing Custom Label. If the synonyms are already added to the Custom Label, it won't be possible to select from the list.
Once you choose the Custom Label to assign the synonyms to, the changes will be collected in a footer element and reanalysis can be triggered similarly to Quick Custom Label Reanalysis.
Read more about this feature on our blog:
Get your perfect labeling model faster: Quick Custom Categorization has arrived4. Analysing Data with Custom Labels
If the Custom Labels are ready, they can be applied to new or already existing analyses by selecting them when uploading the data or reanalyzing the texts. Already applied Custom Labels can also be removed from the analysis during the reanalysis process.
