作者: Stephen Taylor , Tomas Brychcin
DOI: 10.1109/ICNLSP.2018.8374386
关键词:
摘要: Vector semantic spaces, in which a multi-dimenstional numeric vector is used to represent the meaning of word, are making new natural language applications possible. Word analogies have become standard tool evaluate spaces. They also teach us something about what kinds information vectors space can embody. Arabic orthography has morphological constructs realized syntax some other languages: presence or absence article ??; bi-??, ka-?? prepositional prefixes; verbs with object suffixes constitute an entire sentence. The structured word-forms offer opportunity study how representations interact form verb phrases, noun and phrases. We provide corpus focused on participate these conducted examination ten different spaces see them most appropriate for this set analogies, we illustrate use examine phrase-building.