NovaCards is a machine learning web app designed to help medical students reduce the amount of work needed to study with Anki. Currently, NovaCards helps medical students immediately find relevant cards and tags from the popular AnKing deck. This can save medical students an hour or more each day of time spent searching for relevant material, and free up time to actually study.
In order to find flashcards and tags, you must first upload material that covers what you are learning to the NovaCards home page as shown in the image below. This can include textbook chapters, lecture notes, slide decks, or just about any other text file. While NovaCards will return results for almost any relevant medical text, its algorithm works best with text that is in natural language. This means that things such as textbook chapters with complete sentences will often return better results than bulleted lists of notes, for example. These files can be copied and pasted, or uploaded in formats such as a .pdf or .docx file.
Once your text is uploaded, you will have several options available to customize your search using the Advanced Options. You can choose to only search for certain subsets of tags, such as First Aid tags, if you are studying using a specific resource. You can also specify the number of tags or cards you would like to see. Additionally, you can specify the version of AnKing you are using to ensure that you get the most accurate recommendations. These options can be accessed by clicking the hamburger icon to the left of the NovaCards logo at the top of the page.
When you submit your search, you will first see a list of relevant cards. These cards are ranked in order from most relevant to least relevant. It is encouraged that users review each card to see if they think it will be useful. If a card isn’t relevant, users can click the trash can icon to the right of the card to remove it from the list. When you are ready to import the cards into Anki, scroll to the top of the page and click the copy icon to the right of the word “Card.” This can be pasted into Anki in order to pull up the selected cards.
To find relevant tags, click the “Tags” icon to the right of “Cards.” If there are high confidence tags, which there are in the example image below, they will show up first. Clicking the button below for low relevance tags will reveal tags that we have less confidence in. When you find a tag that you would like to use in Anki, simply click the tag to copy, and search for it in the Anki tag search bar to pull it up.
Currently, we are working hard to add more advanced features to NovaCards including the ability to automatically create flashcards from notes, a plugin to Anki that will increase ease of use, and the ability to search through users' own decks instead of just the Anking deck.
NovaCards utilizes state of the art Natural Language Processing to find the most relevant flash cards based on the text you give it. As its name suggests–Natural Language Processing works best with just that: natural language.
So what is natural language? Natural language is more or less just a fancier term for the language we use everyday; like the text of this paragraph, or the conversation you might have with a friend. What natural language is not is text that is structured in a different way, such as bulleted or numbered lists.
Although NovaCards works best with natural language, it is also able to do a pretty good job with other text inputs like lists and bullets, provided there is a little bit of natural language mixed in. In the future, we will be adding a feature that summarizes unnatural language and converts it to natural language, but for now it's best to just submit the highest quality text you can. Below are some examples of really good sources of text–along with some not-so-good ones as well.
Sources of "good text" that would yield the best NovaCards results:
Sources of text that might not lead to great NovaCards results:
Currently, you can upload word documents, PowerPoints, .pdf documents, and raw text files. In addition to this, you can also copy and paste text into the text box, and submit all of these at once.
No, multiple files cannot be uploaded and submitted at the same time. This is only temporary to limit the amount of data being uploaded to the server. Regardless, we recommend submitting multiple searches for different topics you may be interested in.
Each time you submit text to NovaCards, your text has to undergo a lot of computationally intensive math to best find your cards. Even a single paragraph will take a few seconds, with larger chunks of text taking even longer. In addition to the size of your text, the number of users currently using NovaCards will also affect the time it takes for results to load. If you feel like you are waiting abnormally long, or if something is wrong, please feel free to let us know here.
Once you get results from the text you submitted, you can easily get these cards into Anki by copying and pasting all the card ID's into the Anki Search Bar. To do this, click the button in the upper-right hand corner of the card results table, which copies all of the card ID's to your clipboard. Then, open the browse tab in Anki, and paste all of the card ID's you just copied into the search bar (note that this might be a lot of text–but don't worry! Just paste it into the Anki seach bar and let Anki handle the rest). This should make all of the cards from your search populate the browse tab. Just un-suspend them, and you're good to go!
If you want to explore the results from the tags section you can do one of two things: 1) manually go through the tags in the browse tab of the Anki until you get to your desired tag, or 2) in the search bar of the Anki browse tag, type "tag:" followed by your tag of interest, with no space in between.
Yes! For most users the standard settings will do, but for those looking for a more customized experience, we offer advanced options to help you precisely identify the cards and tags you're looking for. Those looking for this experience can take advantage of our hamburger page on the text entry page, which contains the following options:
Seeing as the tags in AnKing can get a bit messy, we chose to restrict the tags we suggest to the five highest yield tags:
The choice to only return certain tags is our attempt to strike a balance between returning a comprehensive list of tags relevant to your text, while also trying to reduce unhelpful tags that might confuse the user or obscure good results. If you feel that some high quality tags are being missed, or maybe some lower yield ones are being included, feel free to give us feedback here.
In developing our tag-finding algorithm, we came up with two approaches: a high confidence approach where we show very few tag's that we are almost positive relate to at least one sentence in your document, and a slightly lower confidence approach where we return more tags that also relate to your document. In order to best help you, we chose to implement both algorithms and let the user decide which tags to use, seeing as the tag list is relatively short, compared to the card list.
To most efficiently interpret the results of the tags found, we suggest looking at both lists, and critically thinking about which tags are most relevant to the text you submitted, using the ones you feel are most appropriate to your content.
Yes, the order of the cards and tags returned does matter. Cards and tags returned first are those that we consider most relevant, while those returned after are decreasingly less relevant.
Yes! As new versions of AnKing are released, we aim to provide support for them within the week that they are released.
Yes! We welcome and thrive off of the feedback of our users. Feel free to submit feedback of any type via the form located here.