The body of scientific literature is growing at an ever-increasing rate. For example, in the field of artificial intelligence (AI), the number of publications is growing every month with more than 500 new papers every day. The situation is similar in other fields of science such as astronomy with more than 40k new articles every year. As a result, it is increasingly difficult for researchers to keep an overview over the current state-of-the-art with the consequence that nearly identical research is advanced in parallel, relevant references are overlooked, or important connections between methods are ignored. This hinders scientific progress as a whole and leads to a suboptimal usage of resources including research funds, compute, energy and intellectual capacity.
While a number of tools (eg, Google Scholar, Semantic Scholar, Connected Papers, Research Rabbit, Arxiv Sanity) simplify literature retrieval today, none of them is able to deliver accurate personalized recommendations on a daily basis.
This motivated us to develop Scholar Inbox, a personal paper recommender which enables researchers to stay up-to-date with the most relevant progress by delivering personal suggestions directly to your inbox - free of charge. Scholar Inbox learns which papers you like and dislike and recommends papers to you that are similar to papers you like based on neural language processing techniques. In particular, Scholar Inbox embeds scientific literature into a vector space that allows for computing paper similarity and training a classifier specific to the users research taste. Scholar Inbox focuses entirely on the written contents and hence also suggests relevant papers that don't receive social media hype.
Scholar Inbox collects papers into daily or weekly digests, based on the user's frequency preference. Importantly, papers are suggested to users only once, no matter if they first appear on arXiv or on a conference website. Using just a few paper votes, users can improve their personal AI-based recommendations. We also offer advanced and very fast search functionality and several paper skimming aids. You can use Scholar Inbox from PC or mobile and you have the option to star papers for later reading (they will appear on your personal "Starred" page). Registration is easy and requires only a few steps.
Scholar Inbox has been created by scientists for scientists. As we are using this tool ourselves, we understand the community's needs. This allows us to better implement new features or improve existing ones. But of course we are also excited to learn about the user's feedback and improve our platform over time. Together, let's give scientists the best tools so they can focus on science!
We love open access and believe that all research should be freely available. Scholar Inbox hence daily indexes all of arXiv, bioRxiv, medRxiv and ChemRxiv including physics, math, computer science, biology, chemistry, health sciences, finance, statistics, electrical engineering and economic. Additionally, we index many open access proceedings in computer science.
The main features of Scholar Inbox include daily or weekly email alerts, training of your personal recommender system by voting for papers, seeing which papers other users of the platform like, enhanced paper search functionalities, paper skimming aids, easy access to the full paper pdf, responsiveness to various screen sizes and fast registration.
Registration is fast and easy. After entering your name and affiliation (your information will be kept confidential and is stored securely) and confirming your email, you will be forwarded to the 'Bootstrapping' page and receive a personal login link that you can use for fast and easy access to our website. We recommend bookmarking this link on your PC and mobile device. Besides, you can of course also login using your email and password. On the Bootstrapping page, you can select 'base authors' (feature powered by Semantic Scholar) that align with your research interests and add papers from them which you find relevant. If you have published papers, you probably want to add yourself. If you are a PhD student, you can add your advisor(s). We recommend to not add more than 3 base authors and to not add more than 100 papers initially. Typically, 20-50 papers is already a good starting point. It is more important to choose the right set of papers than to choose plenty of papers. Note that this is only bootstrapping your recommender system. To improve it further, continue to the iterative training page and vote for suggested papers. Also, continue voting for papers while reading digests later on to further improve your recommendations. Note that after iterative training, the recommender is only retrained once a day, hence the effect of voting on papers in a digest will not become immediately visible.
After selecting 20-50 relevant papers during bootstrapping, you move forward to the 'Iterative Training' stage which proposes papers for you that you can vote upon (or skip if undecided) by pressing the thumb up bottom (if the paper is relevant to your research) or thumb down (if it is not). You can decide how long you want to repeat this process. As a rule of thumb: after voting for another 30-50 papers, your recommender system is trained well enough to get started. After iterative training, you get forwarded to the 'Finish' page that lets you choose a threshold which determines how many papers on average will be recommended to you based on your trained recommender model. We recommend to start with the default choice, and continue to the scholar inbox page. Note that base authors can also be added later on. Moreover, while using our system, we recommend voting for papers to further improve your personal recommender. From now on, you will receive paper recommendations directly into your inbox. While the recommendation emails provide a preview of the most relevant papers, we highly recommend to visit the scholar inbox page to benefit from the full feature set. You can also change the frequency at which you receive emails.
You will now see your 'research cockpit' which displays currently trending papers based on reads and likes at the top as well as you personal digest underneath. When visiting this page for the first time, you will see your most recent digest. Colors indicate relevance to your research interests. You can configure Scholar Inbox to show you relevant papers for a single day only, or for a longer period such as an entire week or month. You can also determine on which day you like to receive email alerts. This can be useful if you want to read papers only on specific days or when returning from vacation. By clicking on thumb-up or thumb-down you train your paper recommender based on your research interests (hence you should thumb up papers that are relevant to you, not only the ones you particulary like). Note that after iterative training, the recommender is only retrained once a day, hence the effect of voting on papers in a digest will not become immediately visible. You can also star papers for later reading, visit the arxiv or github page, share papers on social media or obtain a link to share with your colleagues. Key sentences in paper abstracts are highlighted in yellow.
By clicking on the icon at the bottom right (red box below), you can view the figures and tables in the paper without opening the paper. This might help you to determine if the paper is of interest to you or not. Scholar Inbox also has a responsive mobile web interface that you can access via your phone or tablet. This is useful if you want to quickly check your current digest on the go. You can star papers that you find interesting but want to read later on. We recommend to use the web interface on mobile devices to access and enjoy the full feature set of Scholar Inbox.
When clicking on the user icon, you get to the settings page, where you can change your password, the frequency of email digests (by choosing certain days of the week only), change your relevance threshold or delete your account. You can also choose to hide the trending bar in the settings and the location from which paper PDF files are served from. The voting weight parameter determines the importance of your votes vs. the base papers you have added during onboarding via author search. It is automatically increased based on the number of papers you have voted, but you can also change it manually. Our server hosts PDF files for several months and can be used as fast cache for users in Europe. Besides, it also renames papers with the author name and title. For users outside Europe, we recommend to switch the PDF server to the Arxiv server.
The 'My Papers' page allows you to show positively and negatively rated papers, along with the predicted relevances. This page allows you to quickly find papers that you liked or correct mistakes (by re-rating wrongly rated papers). The 'Starred' pages shows all papers that you have assigned a star to. You can use this feature to remember papers for later reading, eg, when browsing Scholar Inbox on a mobile device.
The conference explorer shows conferences (currently we only support computer science) that are relevant to you based on the papers that appeared in these conferences. You can also use this page to find papers from a certain conference and year, and sort them based on your personal relevance. This allows you to quickly parse conference proceedings of conferences that you did not have a chance to attend. Conferences are colored based on relevance to you.
By clicking on the search icon, you can quickly find papers based on keywords, searching for title, authors or abstract. You can sort the results by personal relevance, query match or publication date. You can also filter for read, starred or arxiv papers.
When writing a paper it is important to not miss any relevant prior work. Semantic search can help you with this! Entering a text (eg, your abstract or introduction) into the text field will retrieve semantically similar papers based on the paper embeddings that we have computed for all papers. You can also limit the publication range for your search to find only new or old papers related to your query.
Modern web browsers (Chrome, Firefox, Safari) support shortcuts for searchable websites. For Scholar Inbox you can setup a shortcut such that typing "s mysearch" into your browser's address bar will directly open the Scholar Inbox search page with the search text "mysearch". This is super useful if you want to quickly find a paper. Here is how to set this up for your browser:
Chromium browsers (Chrome, Brave etc):
Settings -> Search Engine -> Manage search engines and site search -> Site Search -> Add
Enter "Scholar Inbox" for the name, your preferred shortcut (eg, "s") and "https://www.scholar-inbox.com/search?searchPrompt=%s" for the url
Firefox:
If shortcuts are already enabled:
Settings -> Search -> Add
Type in Scholar Inbox for the name, your preferred alias (eg, "s") and "https://www.scholar-inbox.com/search?searchPrompt=%s" for the url
If shortcuts are not yet enabled:
Enter about:config in the address bar
Enter browser.urlbar.update2.engineAliasRefresh
If it already exists, switch the value to true from false. If it doesn't exist, create it as a boolean and set it to true after which the website shortcut can be added as described above.
Safari:
Settings -> Search -> make sure 'Enable Quick Website Search' is enabled
Make sure you've visited scholar-inbox.com so that the browser picks our search engine up
Verify that scholar-inbox is available for search by clicking 'Manage Websites' (Safari knows that our site has search and creates an option to search by pressing the down arrow key after typing scholar-inbox.com)
Remark: Safari does not have an option to add a custom shortcut key out of the box, but this functionality is available via extensions such as http://safarikeywordsearch.aurlien.net/ or https://github.com/EddieCameron/keyword-searches-in-safari)
We hope this blog article could provide you with a good overview of Scholar Inbox. We'd love if you to try it out! We welcome your feedback and suggestions for improvements at scholar-inbox@googlegroups.com!