While summarizing research articles with AI is not new, contrasting information in whole libraries into succinct critical lit-reviews is new. This is exactly what SciSummary does; it is a brilliant tool optimized for working with large libraries of papers rather than a single paper. It searches for and extracts information from them, understands figures, and is one of the best tools to finish your lit review in record time. With the right workflow, you can “read” up to 25 papers in one hour.
Learn which tools can speed up your research: This free 21-day email series introduces you to a range of academic tools and how they work together to speed up the time needed for a literature review. Most of these tools have been around for less than a few years, some use AI. Knowing them gives you an almost unfair productivity advantage for your research.
SciSummary stands for Science Summary and requires you to upload several papers (ideally as PDFs or connect to Zotero) and then generates three types of summaries:
- An individual summary for each paper
- A combined summary of all papers at once
- A contrasting summary (with respect to a key phrase or question)
The resulting summaries are “living documents” that evolve as you interact with the AI behind them. Not only can you ask questions about figures and uncommon expressions, but SciSummary will also recommend new papers that match your questions. This combination makes it an incredibly powerful tool. Let’s dive into the details and see just how well SciSummary performs and what it costs.
Uploading Research Papers to SciSummary
Uploading research papers to SciSummary is the first step you have to do before you can use the tool. It is straight-forward. The recommended option is to upload the files as a PDF or to connect to Zotero if you have the PDFs uploaded there. Go to Library and click the “Add Articles” button at the top right.
After uploading, your papers will be shown in the library, and you can add folders and tags to organize them better.
Creating Combined Multi-Article Research Summaries
SciSummary allows you to combine many research articles and analyze them at once, creating one summary that gives you an overview of an entire field. This is useful when, for example, you start researching a new topic and end up with a large number of interesting papers. Instead of skimming them individually, you can upload them to SciSummary and use the AI to build understanding fast.
After you hit the summarize button, you will receive a comprehensive summary (you may choose between a short, medium or long version). This is how it looks:
The real magic begins when you hit the “Chat” button. You know have the ability to chat with an AI that has the knowledge of all 25 papers. Asking it the right questions is just as important as picking the right papers to read (Learn more about how AI works in my AI for Academics course). The prompts are key here. Here are my three favourite examples to ask an AI when I try to make sense of a large number of papers or try to get into a new field.
Research Prompt 1: Ask the AI about key theories in your field
This is a simple question to get a few keywords. In my case, for example, I learn about “niche theory”. This gives me the ability to dig deeper into this topic and understand the new field I am getting in better.
Moreover, SciSummary points out the papers where these aspects are mentioned. The numbers behind the titles describe which page the AI found the relevant information. A “2.1” means “first half of page 2”.
Research Prompt 2: Ask the AI about Key Review Papers
If you want to reduce the number of papers you read in-depth, find the most influential review papers and read them first. This prompt is about this.
Here, the AI correctly identified a very central review paper. However, the key problem with AI is that it might be incomplete rather than wrong. Since I have read all the papers I uploaded for this demo, I know there are two more essential reviews in my collection that the AI did not mention.
Research Prompt 3: Ask the AI about Questions
Asking the right questions about a topic is central to understanding. This is what the AI chat can help you with. Figure out what people are thinking about in the field.
When you start reading any of the papers, try to answer these questions in your own words.
Research Prompt 4: Asking for Areas of Focus
This prompt gives you areas to focus on when studying this collection of papers. It is similar to the prompt for questions to build understanding. However, the intention here is different. We are going to use these areas of focus to have SciSummary create specific summaries of specific papers to build an understanding of each area individually.
Pick which areas seem relevant to you, and we will use them in the subsequent steps. Using AI to get reliable insight is as much a skill as it is an art. If you want to learn more check out the effortless AI online course.
Creating Contrasting Summaries
The other type of summary in SciSummary is a contrasting summary or compare summary. A key advantage here is that you can specify an Area of Focus, which will narrow your conversation to a specific niche within your article collection. Alternatively, the area of focus can only be used to compare sections of the articles with one another (e.g. materials and methods). Here is how to set it up:
A caveat here is that you can only add ten papers. For best results, use papers that have something to do with your focus area. For example, review papers will rarely have a methods section and would lead to poor results if you asked the AI to compare the methods sections. Identify review (and/or other types) papers with the AI chat described in the previous section. Here is the result of the contrasting summary:
Notice that in the example above, the combined summary now starts with a materials and methods section since I specified this as my area of focus.
Semantic Search with SciSummary and Carbon AI
Combining summaries has given us an overview of 25 papers and allowed us to identify the key areas of focus needed to understand the entire collection. Naturally, we want to know which papers deal with each area of focus. Then, we can use a contrasting summary on this subset and an area of focus to quickly build knowledge on this topic. This is precisely what semantic search does.
Semantic search means “searching by meaning”. Google, for example, does a “keyword search”, which means looking for the occurrence of certain words. A semantic search is much more potent because it can pick up synonyms and jargon, allowing you to use long descriptive queries.
Go to your library and click the search bar to start searching. Option 2 in the screenshot above is the semantic search option. When doing it for the first time, your papers will be uploaded to carbon.ai, an external service that runs the semantic search. According to the founder of SciSummary, your data remains private and is not shared with any other parties.
In semantic search, you can search with broad, simple terms, e.g. “review papers” or “papers dealing with X”. We will use this method to find papers on each focus area we identified earlier using the initial comparing summary. Here is the resulting search:
Important: Run this step multiple times using slightly different queries. Results will differ and might give you a few more papers than a single search. Aim to get a decent amount of papers before proceeding (~ 3-5). If you do not find enough papers, you can use the discovery search (step 1 in the infographic above) to find more articles on a specific topic.
Summarizing Research Papers in Bulk With SciSummary
SciScpace can summarize all your articles individually as well. This is useful when you want to dif a little deeper into each paper and need a clearer understanding what the paper is about and whether it is worth reading, skimming or skipping for you.
Every summary also allows you to define specific words while reading. Select a word in the summary that you do not understand, and a small popup will allow you to get an AI-generated definition:
A history of all previously defined words helps you learn the terms you are not familiar with.
Interacting With Figures in SciSummary
AI is capable of analyzing images as well as text, combined with having the text of the entire paper loaded into the AI you can use this method to get explanations on papers.
For best results, you should make sure that the automatically discovered figures have captions and titles. The caption helps the AI to understand the content better, and the title is used to find mentions in the paper’s text. Captions and titles are normally added automatically but sometimes this fails in which case you should update them manually:
All your changes are stored permanently in the library so you do not have to do it all over again when you navigate away from the site.
Generating an Abstract for Your Paper
A neat additional feature that SciSummary has is the abstract generator. Click on “AI Abstract Generator” and copy and paste your manuscript to get a first draft of an abstract.
In this example, I copy and paste the text directly from Google Docs (arguably the best choice of academic writing software), including all the formatting code. The abstract is a very nice first draft that is quite similar to what the paper’s abstract actually ended up being!
Accessing References of a Paper in Seconds
Accessing a paper’s references takes time; you need to scroll all the way to the bottom, copy and paste the name or DOI, search for it and add it to your reference manager. It is distracting to the reading experience. Instead, SciSummary can add the references of the paper you are reading with one click to the library and auto-generate a summary. This saves you time during your literature review and helps to keep your mind focused on the subject matter.
It is useful to directly generate a summary and add the reference to your paper.
How to read 25 papers in one hour with SciSummary
The different features of SciSummary can be combined into a unique workflow that will allow you to read/skim up to 25 papers in under one hour. Here are the steps:
- Upload the papers you have discovered on a topic (see my lit review workflows for tools)
- Run a Combine summary and get a first understanding of the subject matter
- Ask the AI about key questions, key reviews and areas of focus.
- For each area of focus, use the semantic search to identify which papers speak about this.
- Run a Compare summary on each area of focus (from 3.) on the identified papers (from 4.)
- Gather your summaries and start taking notes, trying to answer the questions (from 3.) to maximize understanding.
Here is an overview:
The disclaimer here is that relying entirely on an AI analysis will lead to missing a few key points. To build a complete understanding, you should start by slowly reading the critical review papers instead. However, if you aim to get an overview of a new field or topic, you can speed up your journey with this workflow.
Remember that note-taking is your one super skill if you also want to remember what you read. This free note-taking tutorial will give you everything you need to get started.
Difference between SciSpace and SciSummary?
While both tools use AI to summarize papers, SciSpace main strength is the discovery of new papers, while SciSummary performs better at extracting information from a large amount of papers and allows you to contrast or connect these findings. In this tutorial on how to use SciScpace AI for your literature review, you can learn more about the tool.
Pricing and SciSummary Discount Code
SciSummary costs 3 – 7 USD monthly making it quite cheap compared to other tools. Additionally you can get a 25% discount with the Code: “ILYA”.
SciSummary vs ChatGPT for Summarizing Research Articles
Primarily, SciSummary is built for working with research articles in bulk, something that would take much longer with chatGPT and require custom prompts or a custom GPT. But as Max Heckel, the founder of SciSummary, points out, the tool exploits the structure of research articles for better performance:
The biggest difference here is that we are able to exploit research papers having a semi-standardized format (aka IMRAD) which lets us better prompt LLMs for summarization and chat. We’re basically able to better load in context than other tools that are for just generic documents.
Think of it this way ChatGPT is a multi purpose tool like a knife it might not be optimal for the different use-cases e.g. a kitchen knife is very different from a whittling knife.
SciSummary vs Scholarcy
Scholarcy and SciSummary both focus on summarizing papers. Still, while Scholarcy focuses on giving very detailed summaries of a single paper, SciSummary condenses and compares information from many papers at once into brief and quick-to-digest chunks. Scholarcy is aimed at early career researchers and students who use papers to learn about a subject and thus need more time reading a single paper. SciSummary, on the other hand, attracts more advanced academics (grad students and above) who have to go through many papers in their field of research, e.g. as part of a literature review.
Weaknesses of SciSummary
My impression was that the SciSummary website is still in development in some parts, which is true for most AI tools that have big money behind them. It only gets better from here, and during the article’s writing, the developers fixed a few issues I reported in a matter of hours.
The summaries were good for the most part but sometimes incomplete, as mentioned in the “finding review example” above. This, too, is not uncommon for AI to be incomplete and can often be fixed by altering the prompt minimally and rerunning it a few times to be more precise.
In conclusion, I find SciSummary worth the money; 7x cheaper than ChatGPT might be a good alternative for those wanting to summarize massive amounts of research articles.
Summary
SciSummary is a new tool that was recently launched, and it is incredibly promising because of its bulk analysis capabilities and narrow focus on scientific literature. To use it well, combine the three main functions it has:
- Comparative summary of up to 25 articles.
- AI chat to drill deeper into aspects of the summary
- Contrasting summary of a subset of articles with a focus question/area in mind
Use (1) to get a broad understanding and then break away key questions and areas of interest using (2) and drill deeper into each of them using (3).
