Frequently Asked Questions
- I forgot my password, how to recover it?
Password can be recovered using the following button:
- How to change my current password?
Password can be changed within the console under “Account | Change Password” or using the following button:
- I do not want to receive daily digest emails any more but still like to receive conferences digests and keep access to the other services/resources, what can I do?
You can disable/re-enable the daily digest service using the button shown below. If daily digest service is disabled, you will not receive daily digest emails any more. You can still read daily digests through our console. The service can be easily re-enabled using the same link. The service can also be updated within the console (under “Account | Enable/Disable Daily Email”).
- How to cancel the service?
You can cancel the service any time using the following button:
- Why does my account become inactive?
When a recipient of our daily digest emails clicks the “spam” button (or its equivalent, such as “junk”, “block”, “reject” buttons) in their email application, their email service provider (such as Yahoo, etc) typically sends a message to us indicating this recipient has marked our emails as spam. When we receive such an email, we will deactivate the complaining recipient to make sure the user will not be in your future email sends. To continue using the service, affected users need to create new accounts, or contact us to restore the previous accounts.
- How to check service status?
You can check the service status using the following button:
- Do you send out digests every day?
Digests are sent out Monday-Friday.
- I can see daily paper digests using the console, but never received any daily digest email, what happened?
Probably when you registered the service, you chose to disable the daily digest service. You can re-enable the service using the following button:
It is also possible that emails sent to you are put in spam folder, or blocked by your email service providers. You can mark our digests as “not spam”, and whitelist us.
- I do not receive daily digest emails on some weekdays, what happened?
There are three possible reasons:
(1) if the areas you signed up are not very active, then it is normal that you do not receive updates everyday.
(2) The paper websites that we are tracking (like arxiv.org) may not release new papers in time due to holidays or outages. Once their services are back to normal, our users will be able to access the digests through paper digest console. We will not send out digest emails if the delay is longer than half a day, but will notify users about the issue in the next daily digest emails. Such scenarios do not happen very often (1-2 times per year).
(3) Emails sent to you have been blocked by your email service provider. If emails sent to a user are bounced back multiple times, we will have to disable the associated account. If a large number of emails sent to a domain are bounced back, we will have to limit #emails we can send to that domain. We notice that users using emails from qq.com and 163.com experience this problem more frequently than the others.
(4) Universal Solution: Users can login to the console and read digests under “Digest | My Daily Digest”, which provides the same contents as email sent outs.
- How do you decide what papers will be included in my daily digest?
(1) When a user creates an account, we asked the user to select at least 1 out of >200 research areas. Our default setting is to push all new papers under the selected area(s) to the user. An average user receives dozens of papers on a daily basis.
(2) We allow users to select keywords and authors to track: users can key in the keywords or authors’ names using Account|Update Interests. Users can also click on follow this author on an author’s profile page or i-Search page. Papers that can pass those filters are called “tracking results” and will be put on top of each daily digest email (marked with *).
(3) Users can decide the best way to get notified. The default setting is to receive a daily digest email if there are new papers under the area(s) that the user signed up with. Users can also choose to receive the daily digest email only when tracking results are available. We also allow users to completely disable the daily digest email and read the digests online.
- How to update my interests & author/keyword tracking list?
Such information can be updated at “Account | Update Interests” (login required). Note: “email” is not editable. Here is an example: John Smith’s update interests page. Users can also click on “follow this author” link on an expert’s profile or i-Search page (for example, Andrew Ng’s profile page or Andrew Ng’s i-search page).
- I get digests for more than 200 papers everyday, and I cannot read all of them. How can I receive fewer?
This is the #1 complaint we receive from our daily digest subscribers, and we have been working hard to address it.
Root Causes: Some areas are active like ‘cs.AI’, so it is expected to see a lot of new papers everyday when multiple such active areas are selected. The problem gets worse in January-March, since deadlines of many conferences fall in this time range.
Solutions: We recommend five solutions:
(1) We always encourage users to narrow down the selected interests (see How to update my interests & author/keyword tracking list?). With fewer interests selected, users get fewer updates.
(2) Papers in daily emails are sorted by primary category, readers can decide what papers to read based on categories if he/she does not have time to read all.
(3) We have a console for users to filter out papers that they are not interested in. Users can login the console, and then click on “Digest | My Daily Digest”. This returns a list of links to all daily digests in the last two weeks. The contents here are the same as what we sent out in daily digest emails. For each day, one can choose to read new papers only, updates to previous papers, or both. Users can also sort papers based on columns and type in a keyword in “filter” field to only show papers with the desired keyword.
(4) We encourage users to specify authors & keywords to track (see “Account | Update Interests”). Users can login the console, and then click on “Digest | My Daily Digest | Tracking Results”. This returns a list of papers that are either published by the authors in the tracking list or are associated with keywords in the tracking list. This will significantly decrease the number of daily papers to browse. For most users, setting author/keyword filters can decrease the number of daily papers to a single digit. Users can also choose to only receive papers by the selected authors or with the selected keywords.
(5) Users can suspend daily digest service. When daily digest service is suspended, users will not receive any daily digest emails, but still have access to the daily digests through console and be able to receive conference digest notifications. Users can always re-enable this service using the same link.
- Precisely, what is a “highlight”?
By our definition, a highlight is a sentence that can immediately tell readers what the paper is about. With such highlights, readers should be able to quickly browse a large number of papers, keep up with the most recent work and find the papers that they like to focus on.
- What does the number in your daily digest email title represent, for example, 0151832 in “Paper Digest-2020.09.08 0151832”?
It is a random number we purposely append to email title. As many of you know, some email service providers may block an email sender if that sender sends out a large number of emails with the same contents (bulk emails). Paper Digest has a large number of daily digest subscribers, and needs to send out a lot of emails on a daily basis. Even though our contents are customized for each individual and thus different, the email titles used to be the same (e.g. “Paper Digest- 2020.09.08”) for all subscribers. To lower the chance that our emails will be marked as spam or rejected, we append a random number to each email title to make titles look different.
- What is the difference between Daily Digest and Topic Tracking?
Daily digest emails are generated based on user selected categories/keywords/authors. There are more than 200 such categories in total, like artificial intelligence, computer vision. Topics in topic tracking typically have a much narrower coverage compared to user selected categories. We currently track ~60 topics, covering trending topics in Biology/Health, Computer Science, Finance, Math, and Physics. Depending on the demand, we may frequently add new topics and remove old topics. The categories in daily digests do not change very often.
Best Paper Digest:
- In “Best Paper” digests, how are the most influential papers selected?
“Best Paper” digests are created for top conferences/journals. All papers published on a given year are ranked based upon #paper citations + #patent citations. The ranking lists are also updated regularly to reflect the most recent changes.
- I have a paper with a lot of citations, but it does not appear in your “Best Paper” digests, may I know why?
There are several possible reasons:
(1) the papers in best paper digests have more citations.
(2) our system does not have all citations that you are aware of or our system does not associate citations to the right paper. For example, our system may fail to match a paper and its citations, when the paper title or author names contain foreign characters.
(3) feel free to let us know if you find any problem, we will look into it and get the problem fixed. All changes will go into the next updates.
- Do I need an account to use the literature review service?
For the time being, this service does not need users to sign in. However, given the high demand of this service, in the near future, it is possible that the service will only be available to the signin users.
- When I clicked on submit button, I received a message “Our literature review service is busy, please try again in a minute.”. What should I do?
Literature review service requires a lot of resources: CPU, Memory, Bandwidth, etc. To avoid overloading the service, we set a limit on #users that we can simultaneously serve. Once the limit is reached, the system will reject further requests and pop up the service busy message. Users can wait for a minute and click the submit button again. Based on our statistics, the chance to see this message is low: under 0.1%.
- How much time should I wait to receive literature review results?
It depends. 99% results are dynamically generated. Typically, it takes ~30 seconds. Sometime, it may take up to 1 minute. We also cache some review results for a couple of hours. If the review result is already in cache, users should get it immediately. Cache is refreshed frequently to guarantee the results are up to date.
- How can I share the literature review results with my colleagues?
Once you are happy with the review results, click on “copy link” under the submit button. This will create a direct link that you can share with your colleagues.
- Will review results remain the same for a while?
The review results could be different from hours ago. More specifically, 99% reviews are dynamically generated to include the most recent research work. We do have a cache to store some review results, and the cache will be refreshed every few hours to include the up to date information.
- What is “IF” and how is it calculated?
“IF” stands for impact factor. It is a score in 1.0-10.0, calculated based on paper citations, patent citations, etc. A higher value indicates a broader impact. This is not the score used to rank “most influential papers”, which is ranked based on a much simpler metrics: #paper_citations + #patent_citations.
- What is the advantage of using your search console compared to using other academic search sites?
There are a couple of advantages:
(1) Most of our search results come with highlight sentences. This helps users quickly decide if a paper is worth reading or not. We are the first to provide this feature in industry, and able to offer high quality highlights for all subjects.
(2) Every paper in our result list is associated with related papers, patents, grants, experts and organizations. This feature is unique in industry.
(3) Author profiles of each paper are also provided. Users can click on author names and see their recent papers, grants and patents. Users can also follow some authors and get alerts on their new publications in daily digest emails.
Technologies behind Paper Digest:
- What technologies are you using to generate paper digest and search results?
We have a fairly complicated text analysis platform (comparable to the platforms developed & used by those largest tech corporations) to support scientific literature tracking, search and summarization services.
The platform contains
(1) a central knowledegebase storing hundreds of millions of documents providing literature search results;
(2) a number of site crawlers tracking new research work in real-time;
(3) a pipeline with dozens of built-in natural language processing and machine learning components to analyze documents. For example, we have several different algorithms (from rule-based to transformer-based) for paper summary generation. We have been improving them over time with the most recent technologies.
- Can you explain how paper highlights are generated in detail?
Given a paper, our system first processes the paper to get the parse results. Then a set of candidate generation components start working on the parse results to generate highlight candidates. Different generation components are based on different strategies. For example, any single sentence in the paper can be generated as a candidate; adjacent sentences can be merged to a candidate; sentences connected by co-references can also be formed as candidates, etc. We usually generate a lot of candidates for each paper. After candidates are created, a set of scoring components will be applied to score each candidate from different perspectives. For example, some scorers are rule-based to check if the candidate has the desired patterns; some are deep learning based scoring candidates based on pre-trained neural networks; some model the location of candidates (from abstract, from conclusion, etc.); some check if the candidate aligns with the title well or not. We have dozens of such candidate generation and scoring components. Techniques used in some of these components are exclusive. Together these components assign each candidate with a set of features. In the last step, a final ranking model ranks all candidates using the features created in the previous steps and produces a confidence score for each candidate. We return the top candidate as the paper highlight if the confidence score is above a threshold. Depending on applications and data availability, we can choose what components to integrate in the pipeline.
- Are you using deep learning in your system?
Some of our components are based on neural networks. They are integrated in our pipeline system.
- Where do your highlight sentences come from?
This depends on what the input is. The input can be as long as a full paper or as short as title plus abstract. Highlight sentences can be created from anywhere in the input. Most of the time, we use title + abstracts as our input.
- What services do you provide?
We currently offer three paper digest services: daily paper digest (introduced in early 2018 to serve subscribed users), conference paper digest (introduced in mid 2019), and “best paper” digest (introduced in early 2021). “Paper Digest” is our trademark granted by the United States Patent and Trademark Office. We also offer search services (paper, patent, grant, person) and real-time topic tracking services.
- Who is running paperdigest.org?
We are a group of researchers and engineers working on machine learning and natural language processing.
- How to contact paperdigest.org?
Questions, requests and comments can be sent to firstname.lastname@example.org.