We've received a lot of requests for guidance on how to apply custom event coding schemas to track media reporting of events in Ukraine in realtime, especially surrounding highly specialized taxonomies relating to specific Russian equipment, divisions and weapons, as well as tactics such as cluster bombs and civilian targeting that are outside of the core CAMEO taxonomy. Both grammar-based and neural automated event coding systems can typically be applied to the Web NGrams 3.0 dataset using the tutorials we published last month.

In particular, augmented coding systems can apply classifiers at either document or snippet levels to identify relevant coverage for human review, including identifying likely mentions, while fully automated coding systems can leverage the tutorials on entity extraction and knowledge graph linking to adapt most coding systems to utilize the ngram dataset.

For those performing narrative assessment, such as of domestic Russian media, we recommend the ngrams dataset and also the Global Similarity Graph coupled with TFProjector for embedding visualization for augmented analysis of specific narrative landscapes or using the ngrams to perform more traditional topical a