One language model for many chatbots
Thirteen chatbot tasks need thirteen separate models?
When visitor ask question for a website, a classifier pick one reply from the answers wey the site keep. This study test whether different sites fit share one multilingual BERT classifier while each chatbot still dey pick only from im own answers.
The comparison cover 13 classification tasks and 333 categories for Turkish, English and German. Dem compare how often the shared model pick right answer, the server memory and the storage e need, with separate models for each bot.
Original title
Performance Trade-Off for Bert Based Multi-Domain Multilingual Chatbot Architectures
Original documents
Original source · dergipark.org.tr dergipark.org.trHow to cite am
Davut Emre Taşar, Şükrü Ozan, Seçilay Kutal, Oğuzhan Ölmez, Semih Gülüm, Fatih Akca, Ceren Belhan · Performance Trade-Off for Bert Based Multi-Domain Multilingual Chatbot Architectures · 2021 · IZ:JA77UY82ER · https://dergipark.org.tr/en/pub/jaida/article/1060221