Our mission is to structure the world’s information.
The electronic world needs structured data. From search engines to your favorite humanoid electronic assistant, structured data elevates functionality to a new level, previously limited to the realm of science fiction.
Locu was founded by a team of MIT researchers who are passionate about data and ambitious enough to aspire to bring structure to complex, messy data sets that pose a challenge for human intelligence, never mind artificial intelligence. Locu's first stop: local data.
Innovation
We are building technologies to efficiently digitize real-world content through a novel combination of document analysis, machine learning and online human computation workflows. The resulting structured and semantically annotated data sets will be made available through a set of APIs.
Document Analysis
Machine Learning
Human Computation
Our patent-pending approach improves upon prior machine learning techniques by introducing additional classification layers that are inspired by human pattern-recognition. Building upon classifiers that extract contextual information, our approach looks for hints in visual artifacts throughout the documents we crawl that indicate high-level semantic categorization.
Another key component of our classification process revolves around preparing the document such that it can be augmented and touched up by 'human computation' provided by Amazon Mechanical Turk or oDesk. The burgeoning field of human computation allows a shift in the traditional computing processes by providing a framework for outsourcing certain computational steps to humans. This enables computer systems to solve problems that are otherwise difficult to solve purely algorithmically. In our model, misclassifications that occur during a machine learning phase are easily spotted by human workers. By recording corrections performed by the human workers, we not only improve data quality, but also build an active feedback loop that retrains our machine learning components.
