Subscribe
Sign in
SciSpace — Your AI assistant to discover and understand research papers
SciSpace is the easiest way to find, understand, and learn any research paper. For every paper you read, get simple explanations and instant answers from AI and discover a network of connected and relevant papers — all in one place.
211
Replies
Awesome product! The results on the first papers I tried it on was amazing, although now with the last few ones I tried the results were completely wrong. Apparently the AI thinks I’m talking about a completely different paper. Maybe it’s overused at the moment, I’ll try again in a couple of days.
I've been waiting for a tool like this! I have 100's of papers on ML that I haven't made the time to read. I'm looking forward to trying Typeset to get through my backlog.
@derekneuland We're glad you're excited about Typeset! We hope it helps you get through your backlog of papers quickly and easily.
Hi Product Hunt! 👋
I'm Saikiran. I'm super excited to introduce the PH community to what we've been working on over the past year. 🤩
SciSpace is your AI copilot that can break down any research paper on the go in plain simple language.🤖
⚙️Here’s how it works:
🔍 Search for a paper on our repository or upload the PDF directly
🖍️ Highlight the confusing text, mathematical formula, or tables
📋 Get a concise and straightforward explanation on the same screen
🗨️ Ask questions about the highlighted section or the paper in general
💬 Receive answers in the same chat box
☝️ Follow up with further questions to get more context
💾 Save your progress and come back to it later
We launched SciSpace Beta earlier this year. We began as a repository of research papers across domains — with metadata of 200 million+ papers and 50 million+ Open Access full-text PDFs. Progressively, we designed new features to simplify research discovery and learning.
The reaction has been overwhelming. Over the past three months, we attracted more than 1 million readers worldwide. And with the addition of this AI-based reading assistant, we felt the time was right for us to release SciSpace publicly to everyone. 🎆
🧐 Why did we build SciSpace?
My co-founder, Shanu, and I worked closely with the research community in our previous venture. It helped us get a firsthand experience of the different stumbling blocks that slow researchers down. What struck us the most were the frustrations many experienced when trying to find papers or absorb critical insights from these complex texts. Sometimes, it felt like looking for a needle in a haystack.
More than 1.8 million research papers get published every year. So it is fair to say research papers are still very much the way to go for information sharing. But anyone who has stepped into the research world knows that it comes with baggage.
All the complicated words, abbreviations, technical clutter, the clunky PDF reading experience, and, let’s be honest, we had better attention spans like twenty years ago! Now, where were we? Oh yeah, diminishing attention span and also information overload–you don’t even know where to begin!
To make things worse, it is not like you can access all these papers in one place. It is scattered across different journals, archives, and digital libraries.
This is what prompted us to build SciSpace. We want researchers to get more leeway to focus more on finding answers to critical global problems. We hope it will deliver value to the research community and anyone interested in reading papers.
🚀 What can you do with SciSpace?
Most researchers spend hours reading papers every day. We want to help them make the most of it.
And here’s how we do that:
➡️ Highlight confusing text, math, and tables to get a simple explanation
➡️ Ask follow-up questions and get instant answers
➡️ Explore and find highly relevant papers on a topic without specifying keywords
You can perform the three actions across all the 200 million+ papers on our platform. 💡
Pretty cool, right? We’re super excited about this feature. We hope it will help researchers worldwide wade through the clutter, find what they need quickly, understand key concepts and apply them to their projects or share them with their students.
The discovery and reading parts of SciSpace are completely free to use. No trial period. No sign-up is needed. 🫶
🔜 What’s next?
➡️ Improved search and recommendations
➡️ Directly extract insightful answers from papers
We're still very early in our journey, but we're thrilled to share SciSpace with the PH community. Please share any feedback or questions you may have. We'd really appreciate it. 🙏
PS: By the way, we're still using our old domain, typeset.io, but we'll be migrating to scispace.com very soon. ⌛
A bit more about SciSpace Copilot,
It's an easy-to-use tool that works this way - Highlight any text, table, or maths equation. Ask a follow-up question. Get instant answers.
Here's a paper where you can test it out: https://typeset.io/papers/attent...
Or if you have a paper of your own, feel free to upload the PDF. It works the same.
Do share your thoughts on it. It's really going to help us improve.
@mailharishv Glad you liked it. If you're looking to explore how Copilot works, here's a paper to get you started: https://typeset.io/papers/attent...
Congrats on the launch, guys! The project looks very interesting. Just out of my curiosity, what technology is behind your AI? Dialogflow, OpenAI, or something proprietary? Another question is why do your stick with the research papers only? I'm pretty sure this technology can be used for a broader range of documents, brining value to more people.
@pavel_pavlenko Glad you liked it. We are hybrid - proprietary tech and GPT-3 combined. Yes, we did think about other domains - what should be the next domain you'd like us to solve for?
@pavel_pavlenko @saikiranchandha Sorry to jump in on this, but I was just thinking the exact same thing! I think there is definitely a business application for this for understanding large volumes of information and making a decision about it, but it could also be used to help create technical blogs by utilising all of the research papers you have at your disposal! Congrats on the launch!
@pavel_pavlenko @glimpseapp Yes, I agree. We are seeing legal documents uploaded on the site. We are open to trying it out in other domains. Do let us know.
@glimpseapp @saikiranchandha Legal is the first thing that comes up actually because of the complex language. It can really help non-legal people do reviews on their own and propose changes. Lawyers cost a lot, so this can drastically decrease legal expenses for companies. Very clear value proposition. Save up to 50% costs on legal work - smth like that.
This is super cool.
Do you use an open source tool to read/recognise/extract data from research papers? Or is it an in-house tool?
@prithiv We do rely on open source parsers but to really get to good quality, we had to fine-tune them.
@phanis Thanks a lot! We are hybrid - for few questions it triggers proprietary tech and for the rest it goes to GPT3. Going forward we plan to charge users directly or labs/universities. We are currently speaking to various stakeholders to figure out a price point that works. But, we intend to keep the core co-pilot conversations free and charge on features around workflow. Pls do share if you have any ideas around this.
This is indeed a radical innovation!!
Data and AI backed power tool would stand in good stead to researchers, philosophers and change drivers in their field.
An excellent platform to excel and accelerate out of your enormous and tedious research process!!
The highlighting feature to me is that it demystifies complex aspects of research papers and yields a simplified output.
I would also like to underscore the other relevant research paper suggestions rendered to elevate your study.
This is best dated platform which supports your research, enhances your study and prompts invention!!!
Hope it will be free for students and independent researches. Thanks for your work! It can be very useful.
@cryptoved That's the plan. Consider joining our Discord to get more updated: https://discord.gg/cR3r8wP92C
Awesome - there's so much prior work you need to reference and understand to get the gist of a paper that if you're not a researcher in the field, you end up needing to read multiple papers to understand one. This looks like a great way to manage that process and shortcut some of the understanding.
Love it! Something that I would pay for: emailing me a newsletter with your TLDR and links to newly published articles on my topics or *questions* of interest.
Conoto