Register for InstructureCon25 • Passes include access to all sessions, the expo hall, entertainment and networking events, meals, and extraterrestrial encounters.
Smart Search is a feature of Canvas that is currently in development. Leveraging semantic algorithms and AI, the new Smart Search feature understands the context of queries, providing more accurate and relevant results without the need for traditional boolean operators or other search tools. This approach is an improvement over traditional keyword-based searches, and will enable users to swiftly and efficiently navigate the vast array of educational resources available on Canvas.
Our Smart Search feature relies on an AI-adjacent technology called “embeddings.” This technology “reads” the course content and creates a complex mathematical representation of each piece of content. When you perform a search, the technology “reads” your query and converts it into a complex mathematical representation. The tool then compares the mathematical representations of both your search and course content to return relevant results; this result is returned with a “Confidence” rating.
Because the search operates on “understanding” both the content and the queries, you can type in keyword(s), content, or just give a general gist of what you are looking for (“guitar”, “what are the steps of photosynthesis?”, “math to turn a function into frequencies“). We think you will be surprised at how well the system understands the intent of your queries and returns results accordingly.
Our Smart Search feature relies on an AI-adjacent technology called “embeddings.” It currently uses the Cohere Embed Multilingual model hosted on AWS. The “nutrition label” below answers some key questions at-a-glance related to the AI model, data, privacy and security.
The Cohere Multilingual Model is supported in all Instructure regions except the Canadian Region (YUL).
As of April 10, 2024, the Search feature is querying the following items within a course: content pages, announcements, discussion prompts and assignment descriptions. The intention is to expand this scope as the Smart Search tool continues development.
In the latest release, we do on-demand indexing within the course. That means when the search is initiated within a course the very first time, users will see a status indicating that results are being prepared. See screenshot below. This indexing will only be seen on the initial launch of Search and may take a few minutes to load; subsequent searches will return results instantly.
When it reaches General Availability, Smart Search will be a feature option, defaulted to disabled, that Admin can choose to make available in their instance of Canvas at no additional cost.
Smart Search supports a multilingual model. This allows for a non-native speaking user of the course language to search for course content using their native language. The supported languages can be found here. Please note that performance may vary across languages.
The UI is a work in progress. In its current iteration, it has passed basic accessibility checks.
If a course is concluded, a user with view access to that concluded course could use the Smart Search feature.
The Smart Search feature is currently classified as a feature preview (available July 17, 2024). We want to provide opportunities for users to provide feedback as often as possible. Clicking thumbs up and thumbs down on your search results and telling us why the results are good or bad will help us determine when results should or shouldn’t be shown to you. Additionally, posting to the Smart Search group in the community is another way to share your feedback.
The Smart Search feature is currently classified as a feature preview. We want to provide opportunities for users to provide feedback as often as possible. Clicking thumbs up and thumbs down on your search results and telling us why the results are good or bad will help us determine when results should or shouldn’t be shown to you. Additionally, posting to the Smart Search group in the community is another way to share your feedback.
As of July 17, 2024, Smart Search is a feature preview option that your institution can unlock or enable. To enable at a course level, go to Course Settings > Feature Options > enable the Smart Search feature option. You may want to manage course navigation to ensure Search is visible to users.
If Smart Search is available in one of your courses, select the Search area of Course Navigation.
In the search box, type in keyword(s), content, or just give a general gist of what you are looking for (“guitar”, “what are the steps of photosynthesis?”, “math to turn a function into frequencies“). We think you will be surprised at how well the system understands the intent of your queries and returns results accordingly.
Smart Search is intended to be available free of charge within core Canvas. The current functionality searches within a course, scoped to specific item types. While we intend to expand this over time, if other search tools better meet your institution's needs, you are welcome to use those alongside smart search.
At this time, Smart Search is not visible in the Canvas for Elementary/Homeroom View. We look forward to adding support for this view in the future.
At this time, Smart Search is not visible on Canvas mobile apps. We look forward to adding mobile support for this feature in the future.
For an initial search within a course, it may take a little bit of time for your results to load. Subsequent searches should return results quicker, as the content has already been indexed.
Updates to Smart Search will be shared in the change log of the community group. Please subscribe to be notified when updates are posted.
Not at this time. We are aware that account-level/cross course searching is desired functionality, but we are prioritizing search within the context of a course and exploring expansion in the future.
There is a lot of information here about how it works and how wonderful it is but no explanation of what its purpose is or what a user - whether teacher or student - would actually use it for. I can see it's a search tool but what does it search and what results does it produce that I can't already find? Why would I bother investigating this tool and learning to use it if i can't understand what it is intended to be useful for?
Does a Smart Search query include both published and unpublished items in a single course? I see a variety of comments in the community related to this question, and users have opinions, but I don't see an official answer. It would also be good to know if the query has changed over time to include (or not include) published content so those of us who are testing understand the development timeline. Any info you could provide would be helpful.
@osnow My testing indicates that it is role-based, searching only the content your role has access to. As a teacher, unpublished pages are included in the results that a student's results would not include.
My lingering question is: After the initial indexing, if the instructor adds/modifies content, does that trigger a new index to be generated for the next search?
Note: You can only embed guides in Canvas courses. Embedding on other sites is not supported.
To interact with Panda Bot in the Instructure Community, you need to sign up or log in:
Sign In