: Vocabulario de PLN: BigARTM

NLP methods and tools > NLP software > BigARTM

BigARTM

BigARTM is a fast and flexible library for topic modeling (Frei and Apishev 2016) based on Additive Regularization of Topic Models (ARTM) formalism (Vorontsov 2014). (Bulatov, Alekseev, Vorontsov, Polyudova, Veselova, Goncharov & Egorov, 2020)
BigARTM is a tool to infer topic models based on a technique called Additive Regularization of Topic Models. (Yusupov & Kuratov, 2018)

Another shortcoming of BigARTM is the difficulty in extending it. (Bulatov, Alekseev, Vorontsov, Polyudova, Veselova, Goncharov & Egorov, 2020)
As an example of dealing with auxiliary information BigARTM makes it very easy to include document metadata (e.g. authors timestamps tags and n-grams) in a single model. (Bulatov, Alekseev, Vorontsov, Polyudova, Veselova, Goncharov & Egorov, 2020)
BigARTM library adds several other metrics such as sparsity purity and contrast (Vorontsov and Potapenko 2015). (Bulatov, Alekseev, Vorontsov, Polyudova, Veselova, Goncharov & Egorov, 2020)
Therefore as applications of BigARTM were becoming more diverse and the algorithms were gradually refined the high-level interface of BigARTM was getting less well-suited for "best practices". (Bulatov, Alekseev, Vorontsov, Polyudova, Veselova, Goncharov & Egorov, 2020)

http://data.loterre.fr/ark:/67375/8LP-PP7K2BLW-S

RDF/XML TURTLE JSON-LD última modificación 3/5/24

Loterre