000008748 001__ 8748
000008748 005__ 20240124114308.0
000008748 0247_ $$2DOI$$a10.6083/f1881m64w
000008748 037__ $$aETD
000008748 245__ $$aOvercoming limitations of categorical language modeling
000008748 260__ $$bOregon Health and Science University
000008748 269__ $$a2020
000008748 336__ $$aThesis
000008748 502__ $$bPh.D.
000008748 520__ $$aNeural language models typically employ a categorical approach to prediction and training, leading to several well-known computational and numerical limitations. These limitations are particularly evident in applied settings where language models are employed as means for communication. From speller systems employed as assistive technology to texting applications on smartphones, all language models revolve around category-based prediction. Research shows that neural-category approaches to language modeling are questionable for predicting low-frequency words that are essential for user personalization. It is also challenging to adapt these architectures to a changing vocabulary due to the initially learned vocabulary constraints, which limit predictions of relevant categories (i.e., words) a user can type. Recently, such categorical models were shown to be relatively complex with long inference times, which may be detrimental for user engagement. In this thesis, I reevaluate neural-category approaches and propose an alternative: continuous output prediction.
000008748 6531_ $$aaugmentative and alternative communication
000008748 6531_ $$alanguage models
000008748 691__ $$aSchool of Medicine$$041369
000008748 7001_ $$aDudy, Shiran
000008748 8564_ $$9c2b5a7aa-3c7d-468d-8445-3a570cf0c035$$s9167791$$uhttps://digitalcollections.ohsu.edu/record/8748/files/Dudy.Shiran.2020.pdf
000008748 905__ $$a/rest/prod/f1/88/1m/64/f1881m64w
000008748 909CO $$ooai:digitalcollections.ohsu.edu:8748$$pstudent-work
000008748 980__ $$aTheses and Dissertations