Topics

Overview

The Topics page manages the topic dictionary used by open-set topic detection. Each topic has:

• a topic label — the class documents are tagged with, and the value you filter on in Search;
• seed words for each language — the words topic detection uses for that topic in that language’s documents.

The page shows each topic once, with its seed words in every language side by side, so you can see at a glance which languages a topic still misses.

Open the page

You need the Modify Settings permission. Click Administration in the navigation bar, then click the Topics card. If you are already in Administration, click Topics in the section bar.

Click to enlarge

• 1 Dictionary coverage - the totals for topics, seed terms, and topics complete in all 5 languages.
• 2 Language cards - coverage per language. Click a card with missing topics to list them.
• 3 Search - finds topics by label or seed word.
• 4 Has words in and Rows - the language filter and the page size.
• 5 All / Complete / Gaps - shows all topics, complete topics, or topics with missing languages.
• 6 Add topic - opens the Add topic dialog.
• 7 Showing line - which topics are shown, active filters, and page numbers.
• 8 Topic table - one row per topic with its seed words, coverage, and actions.
• 9 Seed words - one slot per language. Click the words to edit them.

Dictionary coverage

The band at the top summarises the whole dictionary: the number of topics, the number of seed terms, and how many topics have seed words in all five languages the AI services run (Romanian, English, French, Russian, Hungarian).

Each language card shows how many topics have words in that language, a coverage bar, and how many are missing. Click a language card that has missing topics to list exactly those topics; click it again (or clear the filter chip) to show every topic.

Toolbar

• Search — finds topics by label or by seed word in any language. Results update as you type; press Enter to search at once, Esc to clear, or / anywhere on the page to jump to the search field. Matches are highlighted.
• Has words in — shows only the topics that have seed words in one language.
• Rows — how many topics each page shows (10, 25, 50 or 100).
• All / Complete / Gaps — all topics, only topics with words in every language, or only topics missing at least one language. The numbers count the whole dictionary.
• Add topic — opens the Add topic dialog.

The line under the toolbar says which topics are shown (for example “Showing 1–25 of 28 topics”), lists the active filters with a button to clear them, and holds the page numbers.

Click to enlarge

• 1 Search - the search text.
• 2 Filter chip - the active search. Click × to clear it.
• 3 Highlighted matches - the matching text in labels and seed words.

Topic table

On tablets and phones each topic becomes a card: the label, then every language’s words, then coverage and actions. Nothing is hidden on smaller screens.

Add a topic

Click Add topic to open the Add topic dialog.

Click to enlarge

• 1 Topic label - the label documents are tagged with.
• 2 Seed words - the words for this topic in the chosen language.
• 3 Language - the language of the seed words.
• 4 Close - closes the dialog without saving.
• 5 Create topic - saves the topic.

Click Create topic to save, or Close to cancel. To add another language to an existing topic, use its dashed Add words slot or the Add words in another language action. The dialog, titled Add followed by the language and words, keeps the label fixed; click Add words to save. If the topic already has words in that language, a note says that saving adds another entry beside them.

Edit words

Click a language’s words in the table to open the Edit … words dialog.

Click to enlarge

• 1 Topic label - changing it moves only these words to another label.
• 2 Seed words - the words for this topic in this language.
• 3 Language - the language of the seed words.
• 4 Delete words - removes only this language’s words after a confirmation.
• 5 Close - closes the dialog without saving.
• 6 Save changes - applies the changes.

You can change the label (this moves only these words to another label), the seed words, and the language. Click Save changes to apply, or Delete words to remove only this language’s words after a confirmation. Seed words must be unique per language: if the words already belong to another topic in that language, the dialog says which one.

Reset the dictionary

The Reset the dictionary panel at the bottom of the page deletes seed terms in bulk. Open it, choose Every language or one language (each option shows how many terms it deletes), type the confirmation phrase shown (delete all topics, or the language name in lowercase), and click the delete button. Resetting one language leaves the other languages untouched.

Documents may lose topic classification until topics are recreated and processing runs again. There is no undo.

Use in training

The topic list is used in training. By default all languages are trained using the Romanian topics, because the Romanian topics are considered to be more up to date and because the Romanian documents are used in training all the languages. Romanian topics should only contain Romanian words. Topics in other languages can mix Romanian and specific terms.

Notes

• Deleting a topic, a language’s words, or resetting the dictionary is permanent.
• Changing a filter or the page size returns the list to the first page.
• Search covers topic labels and seed words in every language.

Related pages

• Administration
• Search