Understanding at scale


We take fun seriously. Fetching, crunching, analyzing, and visualizing data is a complex endeavor. We create understanding for the ever-evolving language of online communication.

Beyond Mention Counting

Spiketrap combines a structured database of thousands of media and entertainment-related brands (such as companies, franchises, characters, movies, TV shows, games and so on) with a network of NLP classifiers to figure out the entities a piece of content refers to, even when not explicitly mentioned in the text. Our AI is able to deal with ambiguities, abbreviations, and acronyms: think of companies such as “2K” or “Blizzard”; TV shows such as “Friends”; games like “Anthem”, “Control”.

Emotion AI

Our sentiment classifiers are trained against large proprietary datasets, and vastly outperform latest classifiers and off-the-shelf sentiment services when applied to the media and entertainment landscape. Over the years we evolved the family of emotions that our AI can detect, now encompassing excitement, sarcasm, toxicity, and intent of purchase.

Putting Things in Perspective

One of our core R&Ds, besides Emotion AI, revolves around bringing the important information to the surface by identifying the top topics of conversations around your property. Our AI is able to discern trends in real-time, connecting the dots across different data sources and time intervals

World-Class AI Team

Our team is an active part of the AI scientific community. We are continuously contributing, learning, and evaluating our technology against the state of the art.

Research Advisors

Research Publications
A Reduction for Efficient LDA Topic Reconstruction
Almanza, Chierichetti, Panconesi, Vattani – NIPS 2018
The equivalence of Single-Topic and LDA topic reconstruction
Chierichetti, Panconesi, Vattani – Pre-print
Optimal Probabilistic Cache Stampede Prevention
Chierichetti, Lowenstein, Vattani – VLDB 2015
Fast Greedy Algorithms in MapReduce and Streaming
Kumar, Moseley, Vassilvitskii, Vattani – SPAA 2013
Scalable k-means++
Bahmani, Moseley, Vattani, Kumar, Vassilvitskii – VLDB 2012