Technology Opportunities

Available Knowhow

Appropriate for any type of text, including reviews and tweets

Professor Ari Rappoport, Oren Tsur, Department of Computer Science and Engineering, Dmitry Davidov, Interdisciplinary Center for Neural Computation (ICNC), The Rachel and Selim Benin School of Computer Science and Engineering

Highlights

• An algorithm that learns the type of words and patterns that distinguish sarcastic remarks – those that mean the opposite of what they literally convey, or that convey a sentiment inconsistent with the literal reading
• The algorithm identifies sarcastic tweets on Twitter and sarcastic sentences in product reviews.
• Sarcasm recognition contributes to the performance of review summarization and ranking systems.

Our Innovation

A novel semi-supervised algorithm for sarcasm identification (SASI).The algorithm identifies sarcastic sentences using a machine learning technique in which a small number of sarcastic sentences act as the basis for the software to learn and generalize upon. The algorithm is then able to identify sarcastic sentences even if they differ from the examples.

Key Features

• Algorithm recognizes sarcasm in any type of text
• Can recognize sarcastic sentences in online product reviews with 77% precision
• Improves review summarization and opinion mining systems

• Recognition of sarcastic statements generates improved personalized content and recommendations to human users
• Benefits opinion-mining systems that troll the Web to measure public sentiment about a product or idea
• Improves the personalization of content ranking and recommendation systems

Researcher Information: cs.huji.ac.il/~arir/, staff.science.uva.nl/~otsur/, alice.nc.huji.ac.il/~dmitry/