Everything you need to know about PubTrawlr!
There are a few steps in the process. Broadly, we use Natural Language Processing algorithms. After running some descriptive statistics, we then determine what the optimal clustering of the topics is. We then look at which articles are most likely to occur within a given topic. For the summarization, we then determine what sentences are most representative of the topics, then use this to guide our text generation. Read more about our methods in this open-source article.
Right now, we query the PubMed database, which indexes the biomedical literature. This is a vast area that includes social services, public health, sociology, criminology, and some educational literature. We are looking to expand in the future by more explicitly pulling in the educational and business literature.
Interpreting my results
We give you a lot of information back. While we want it to be as straightforward as possible, we recognize that some things may be unfamiliar to you.
"Best" means different things to different people. Because of that, we flag different types of articles. Review articles are those that try to summarize trends and findings for a specific question, such as “what’s the best treatment for immigrant children with autism?” We flag articles that come from prominent journals. These are journals that are frequently read and cited by fellow researchers. We also look for journals that are frequently publishing on a specific topic, and show a selection of articles from these. Of course, your full results are available for download in a variety of formats.
Additionally, our Premium Search selects the article most representative of each topic using a specific algorithm.
Patience, you must have my young Padawan! Running analyses on datasets in the hundreds (or thousands) takes time. However, we are continually upgrading our distributed computing to bring you results faster and faster.
That's not really a question 😉. Our clustering and summarization methods are entirely computer-driven. The computer doesn't care whether a human can interpret the results; it just looks at the words as data and goes from there. That said, we are always at work, improving the quality of our underlying methods to make sure that you get results that are meaningful to you.
When you don't get any results, usually the problem is in spelling. First, check to make sure that all your search terms are spelled correctly. Our team is working to implement some predictive text mechanisms to help you out. Next, try to broaden your search. Additionally, your topic area may not be covered in the PubMed database!
Everything you need to know about your account and data
Yes, we offer steep discounts for institutional or business accounts. Reach out to us for pricing at email@example.com
We have optimized our pricing for students, so for the time being, we do not offer discounts for those with .edu addresses.
We want to make things as simple as possible. You can cancel your subscription at any time and drop down to the free tier, while still retaining access for the time you have purchased.
A few words about us.
Yep, we are specifically looking for people to lead our organizational/institutional sales team right now. We also are always improving our underlying algorithms, so we are also looking for methodologists with expertise in LSTM architectures.