Menu

An intelligent system that reads, learns, and reasons about science.

Live Demo

Aristo is an exciting, multidisciplinary project that aims to develop systems that have a deeper understanding of the world and are capable of demonstrating that understanding through question answering and explanation. A key metric we use is Aristo's performance on science exams, as these questions test many of the key skills required for machine intelligence. Read more about Aristo and our recent success with 8th grade science exams in the article From ‘F’ to ‘A’ on the N.Y. Regents Science Exams: An Overview of the Aristo Project.

Our research integrates multiple AI technologies, including:
  • Natural language processing
  • Deep learning
  • Information extraction
  • Knowledge representation and reasoning
  • Commonsense knowledge
  • Diagram understanding
Explore Aristo questions

The Aristo System

The most recent Aristo system is comprised of eight solvers that work together to answer non-diagram, multiple-choice (NDMC) questions. Several older solvers, not part of the current system, address other types of question and other aspects of question-answering (see Aristo Research Areas). Aristo's most recent solvers include:
  • The Information Retrieval, PMI, and ACME solvers that look for answers in a large corpus using statistical word correlations. These solvers are effective for "lookup" questions where an answer is explicit in text.
  • The Tuple Inference, Multee, and Qualitative Reasoning solvers that attempt to answer questions by reasoning, where two or more pieces of evidence need to be combined to derive an answer.
  • The AristoBERT and AristoRoBERTa solvers that apply the recent BERT-based language-models to science questions. These systems are trained to apply relevant background knowledge to the question, and use a small training curriculum to improve their performance. Their high performance reflects the rapid progress made by the NLP field as a whole.


Want to Try?

As part of Aristo, we have created and released several datasets and accompanying leaderboards. Can your system do better?
Our public datasets include:

•  The AI2 Reasoning Challenge (ARC) - a large collection of real science questions
•  OpenBookQA - reasoning over knowledge in a science book
•  SciTail - textual entailment
•  ProPara - scientific processes

For fun, you can also try competing against a 2016 version of Aristo in our interactive quiz.

Aristo Research Areas


“Knowing is not enough, we must apply. Willing is not enough, we must do.”
—  Johann Wolfgang von Goethe