This feature filters the steps where questions appear based on keyword cues that the worker spoke during filming. When you specify a target keyword in "Step conversion settings," questions appear only in the steps where that keyword was spoken, and questions are removed from other steps.
Questions are available with Core, Pro, and Enterprise plans. Filtering is OFF by default, so unless you change the settings, all questions are applied as before.
When to use
AI processing creates questions aggressively whenever a confirmation scene is detected. In longer videos covering a single task, the number of steps increases, and so do the questions. If you ask for answers to all of them, it creates a large burden on staff viewing them at the worksite.
With this feature, you can place questions only at the points where the worker intentionally signaled during filming. Use it when you want questions placed only at key points, or when you have many steps and want to filter down the questions.
Filtering uses captions created from the video audio. If your video has no audio, words cannot be found, so the checkbox cannot be selected.
How to set it up
- When you create an SOP from source material, "Step conversion settings" appears
- Open "Step-level questions" and verify that "Apply questions to steps" is checked
- Check "Keep questions only for steps where a keyword was spoken"
- Click "Check" to open "Target keyword settings"
- Specify the keyword to target and click "Use these settings"
- Return to "Step conversion settings" and click "Confirm with this content"
Your specified settings are saved and become the default values the next time you open "Step conversion settings."
Specify target keywords
In the top section of "Target keyword settings," you decide which words to target. The list contains three types:
- Templates … Words registered by your team
- Add … Words you type on this screen
- Default … Words pre-built into Dive (Caution, Point, Important, etc.)
Switch whether to use a word with the checkbox on each row. Words with unchecked boxes are excluded from matching but do not disappear from the list. If you operate the header checkbox in the table, you can switch all rows currently showing at once.
To add a new word, type it in the input field at the top and press Enter.
The "Matching steps" column on each row shows how many steps in this video the word appears. Words with 0 instances are not used in this video. Default words with no matches in this video are initially collapsed and can be displayed using the link at the bottom of the list.
Matching is flexible with writing variations. Katakana and hiragana, full-width and half-width, and the presence or absence of long vowel marks are treated as the same. Differences between kanji and kana (such as "確認" and "かくにん") are distinguished, so specify the wording used in the video.
Choose from words that appear often in this video
Even if your team has decided on specific signal words, actual filming may use different phrasing. For example, you might have decided on "Caution," but the worker might say "Be careful."
Below the input field, words that are used across multiple steps in the video appear as candidates. Click to add them to targets. Since you can choose from words actually spoken, you can avoid a situation where no matches are found for the specified word.
Candidates appear only in videos with a reasonable number of steps. In videos with few steps, there are no differences in word usage, so candidates are not displayed.
Check results
The bottom section of the screen shows what happens with that setting. You can see the total number of questions and how many remain.
Switch between "Keep" and "Remove" to see the list of steps that apply to each. The list shows the step number, step title, and question text, plus the spoken phrase that was the basis. Spots matching the keyword are highlighted in color.
On the "Remove" side, you can see what was actually said in that step. If a step you wanted to keep is being removed, add the words found in that utterance to your targets.
Register team-common keywords
Signal words decided by your team can be registered on the dictionary page, so they appear as candidates each time you create an SOP.
- Open "Dictionary" from the left menu
- Open "Question keywords" in the "Templates" tab
- Type the word and click "Add"
Registration is done by users with Team Administrator or higher permissions. Templates are available with Pro and Enterprise plans.
Registered words appear as "Templates" in "Target keyword settings." Whether to use them can be switched by each user, and that choice does not affect the registration.
Operating during filming
This feature is most effective when you decide on signal words at the filming stage. When you want confirmation at a certain scene, have the worker clearly speak the decided word, and a question will appear in that step.
If you choose a word that is not often used in regular task instructions, you can pinpoint exactly where you want. If you choose a common word like "Confirm" that appears frequently during work, many steps will apply and filtering won't be effective. Adjust based on the count shown in "Matching steps."
Changing later
Questions removed by filtering have their question type set to "None." Since the question content itself remains, if the SOP is still in Draft status, you can reset the questions for each step. If you want to create it again with different keywords, create the SOP again from the same source material.
Related information
- Step conversion settings (Narration, Captions, Questions, Related files)
- Step-level question settings
- How to create and use question templates (Apply to multiple steps at once)
- Remove questions in bulk (Take questions only out of steps)
- Pass/fail judgment questions and work records (Checklist operation)