Orthogonality based Semantic Encoding

2021-2022 ősz

Nincs megadva

Téma leírása

The topic is available also in Hungarian.

Semantic encoding gains incredible attention nowadays. This domain is extremely competitive as tech giants conduct research on this topic. The goal in this case is to assign a semantic vector to a text, so that the vector should hold the meaning of the text. Some examples on algorithms are: https://github.com/huggingface/transformers. The student will work on developing a novel encoder, which involves the orthogonality (or independence) in the encoding process.

Maximális létszám: 2 fő