Importance of Accurately Labeled Data for Topical Machine Learning
Accurate data plays a critical role producing useful machine learning models. Most available training sets create models outputting either generic topic classifications or unstructured flat entities. Attempts at granular, hierarchical outputs are sub-optimal, even when trained on corpuses in a single vertical, because of the significant ambiguity of natural language. Automatically labeling training data with hundreds of thousands of hierarchical topics produces flexible, structured classifiers in any vertical. In this session, we’ll address: • The role taxonomy plays in machine learning • Benchmarking opportunities for better machine learning results • Improving the machine learning model with labeled topic data
As CEO of Info.com and eContext, Stephen is responsible for all aspects of development at both companies and has more than 20 years of experience in managing businesses. Stephen has a strong marketing background and a passion for big data and analytics.
Info.com is an independent search platform with 8 million unique users. From a single search query, Info.com provides results from the leading search engines. Info.com is also partnered with eight vertical search providers. Info.com owns Info.co.uk and Info.com.au and has Chicago and London offices.