Nested Knowledge

Bringing Systematic Review to Life

User Tools

Site Tools


wiki:autolit:utilities:artificialintelligence

Artificial Intelligence in Nested Knowledge

Nested Knowledge offers a variety of AI-enhanced tools that make the systematic review process smoother and easier for users. Listed in order of when you would use each tool in the review process, here are our four key AI features:

RoboPICO

Search Exploration

  • RoboPICO works to provide the most commonly reported Population, Interventions, Comparators and Outcomes (PICOs; in red) in Search Exploration to help you build an effective search query. When you use the Search Exploration tool, RoboPICO automatically runs when you hit “Refresh Exploration.” You can select these terms, view their definitions, and build them into your search query (in orange).

Screening

  • RoboPICO also auto-highlights PICOs found in Abstracts during Screening to aid you to more efficiently make a decision on inclusion or exclusion of the record but may be toggled off (in red).

Bibliomine

The Nested Knowledge Bibliomine feature auto-extracts citations when you upload a pdf of any previous systematic review or landmark study and imports all cited references as records directly into your nest. This allows for fast updating of existing reviews and turns them into living reviews through quick implementation in the software. Alternatively, citation mining from an existing project pdf builds a solid foundation for a new project in your field of interest.

Robot Screener

Robot Screener replaces one human reviewer with an AI reviewer in nests with a Dual Screening mode. It does require training (50 adjudicated screening decisions and 10 advancements or inclusions) prior to being switched on, but continually trains itself thereafter. Then, a human adjudicator reviews the preliminary screenings and makes the final decision.

Smart Tag Recommendations (Enterprise users only)

Smart Tag Recommendations uses OpenAI's GPT-4 to search each full text for the most relevant evidence to extract alongside a tag. This is unlike Standard Tag Recommendations (available to all) which performs a simple keyword search (Standard Tag Recommendations). This feature helps to better answer questions (in Form-based modes) and saves time reading through pdfs to retrieve the data of interest.

Core Smart Tags

Search Exploration

After inputting concepts of interest and refreshing exploration, Core Smart Tags identifies and displays Study Type via Sunburst diagram, Location via Choropleth map (below), Size via Histogram, and frequently reported Acronyms. Interact with each page and view associated records from a subset of 250 records from PubMed.

Tagging/Data Extraction

After adding a search to your nest, you will want to configure a data extraction/tag hierarchy. Use Core Smart Tags to input your research question, generate a hierarchy of PICOs, Study Type, Location and Size and either recommend evidence or immediately extract this data from all studies in your nest.

This not only produces a hierarchy for you, but you can then go on to immediately filter to subsets of studies to view and screen using Study Inspector:

Dashboard

Core Smart Tags also allows you to display additional cards in Dashboard. This includes tags as Sunburst Cards, Study Location as a Map/Choropleth, and Study Types as a Histogram:

Ask AI Support Chat Bot

While not directly involved in the review workflow, our Ask AI Chat Bot is incredibly useful for any questions you might have on how to use the software and is accessible on any page in your nest in the top right.

wiki/autolit/utilities/artificialintelligence.txt · Last modified: 2024/09/20 12:04 by jthurnham