Literature Review

Introducing our new Literature Review feature, designed to streamline your research process by summarising key publications within your current subgraph. This feature leverages advanced language models (with GPT-4o) to combine a list of genes with your focus topics, finding relevant information within the publications in your subgraph. You can choose between short, medium, or long reviews, and depending on the Resource you’re accessing (Free or Premium), you have a monthly limit on the number of summaries you can generate. For those needing more summaries, you can also use your own GPT API Key from OpenAI.

Back to nova
new
From Search to Summary: Literature Review
Our new Literature Review feature revolutionises how you interact with the publications found within your current subgraph. By leveraging the power of GPT-4o, it can summarise the content of these publications in relation to a list of genes and focus topics you provide.

You have the flexibility to select the length of the summary – short, medium, or long – based on your needs.
How To Get Started
To get started, generate a graph and click AI Summary at the top. Now, input a list of up to three genes (or more, but three is ideal) and add your focus topics. These also auto-populate based on your initial search in KnetMiner. You can choose whether to review all publications within the current subgraph or only those visible on screen. Once you hit generate, in under a minute, the knowledge within your graph is summarised and presented back to you in a large popup where you can read your review. Evidence is backed by citations, which, when clicked, will open the publication up in the graph, allowing you to read further details in the Info Box. You're also able to generate multiple reviews for each network, and they'll all save when you save the network to your account. You can also mark specific ones as favourite, rename, and delete ones you're not so fond of.

Fine Tuned & Enhanced

We’ve put significant effort into reducing the chances of model hallucinations by fine-tuning our prompt settings and utilising a Retrieval-Augmented Generation (RAG) based approach, with low temperature, ensuring the work remains as scientific as possible. The RAG approach is used to find the closest matching publications from the subgraph, based on your parameters (genes & focus topics). Our prompt then instructs GPT 4o to focus solely on the information retrieved from these publications when generating the review. This helps us ensure that the generated literature review is highly reliable and does not deviate from the source publications found in the subgraph. We’re continuously exploring more in this space, working towards adding more ML/AI enhancements into KnetMiner for better and more intuitive search capabilities and text mining. Watch this space for future developments, and please share any feedback or suggestions you may have.

We’ve worked on these updates alongside the experts who use them most and believe they make KnetMiner a more powerful, user-friendly tool, tailored for those who need results fast.
‍
‍We welcome any and all positive feedback as we begin rolling out the all New KnetMiner.

Explore NovaContact Us