{"id":{"repo_id":"washington","oai_identifier":"oai:digital.lib.washington.edu:1773/48282"},"canonical_url":"https://search.dev.ndltd.org/etd/washington/oai:digital.lib.washington.edu:1773/48282","repository":{"repo_id":"washington","name":"University of Washington","base_url":"https://digital.lib.washington.edu/server/oai/request"},"display":{"title":"Latent Compositional Representations for English Function Word Comprehension","abstract":"This paper investigates whether biasing natural language models toward tree-compositional structure and systematic token representation can improve performance on tasks that require the use of function words. The method used treats tree-structure as latent and thus requires no gold parse labels. Results show that across four function-word-focused NLI probing tasks, tree-compositional models perform as well as LSTMs, but lag behind BERT to varying degrees between tasks. Context-dependent behavior of tree-compositional models highlights a potential weakness of the architecture in the absence of grounding information.","abstract_html":"This paper investigates whether biasing natural language models toward tree-compositional structure and systematic token representation can improve performance on tasks that require the use of function words. The method used treats tree-structure as latent and thus requires no gold parse labels. Results show that across four function-word-focused NLI probing tasks, tree-compositional models perform as well as LSTMs, but lag behind BERT to varying degrees between tasks. Context-dependent behavior of tree-compositional models highlights a potential weakness of the architecture in the absence of grounding information.","abstract_has_math":false,"creators":["Barnes, Megan"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Steinert-Threlkeld, Shane"],"committee_chairs":[],"committee_members":[],"year":2022,"date_issued":"2022-01-26","date_published":"2022-01-26","updated_at":"2026-07-24T05:58:01Z","subjects":["Linguistics","Computer science"],"languages":["en_US"],"rights":["CC BY"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/1773/48282","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Steinert-Threlkeld, Shane"]},{"key":"dc:creator","label":"Author","values":["Barnes, Megan"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2022-01-26T23:25:26Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2022-01-26T23:25:26Z"]},{"key":"dc:date.issued","label":"Date","values":["2022-01-26"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Linguistics","Computer science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en_US"]},{"key":"dc:rights","label":"Dc Rights","values":["CC BY"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["Barnes_washington_0250O_23738.pdf"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://hdl.handle.net/1773/48282"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (Master's)--University of Washington, 2021"]},{"key":"dc:description.abstract","label":"Abstract","values":["This paper investigates whether biasing natural language models toward tree-compositional structure and systematic token representation can improve performance on tasks that require the use of function words. The method used treats tree-structure as latent and thus requires no gold parse labels. Results show that across four function-word-focused NLI probing tasks, tree-compositional models perform as well as LSTMs, but lag behind BERT to varying degrees between tasks. Context-dependent behavior of tree-compositional models highlights a potential weakness of the architecture in the absence of grounding information."]},{"key":"dc:format.mimetype","label":"Dc Format Mimetype","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Latent Compositional Representations for English Function Word Comprehension"]}]}],"canonical_facts":{"dc:contributor.advisor":["Steinert-Threlkeld, Shane"],"dc:creator":["Barnes, Megan"],"dc:date.accessioned":["2022-01-26T23:25:26Z"],"dc:date.available":["2022-01-26T23:25:26Z"],"dc:date.issued":["2022-01-26"],"dc:description":["Thesis (Master's)--University of Washington, 2021"],"dc:description.abstract":["This paper investigates whether biasing natural language models toward tree-compositional structure and systematic token representation can improve performance on tasks that require the use of function words. The method used treats tree-structure as latent and thus requires no gold parse labels. Results show that across four function-word-focused NLI probing tasks, tree-compositional models perform as well as LSTMs, but lag behind BERT to varying degrees between tasks. Context-dependent behavior of tree-compositional models highlights a potential weakness of the architecture in the absence of grounding information."],"dc:format.mimetype":["application/pdf"],"dc:identifier.other":["Barnes_washington_0250O_23738.pdf"],"dc:identifier.uri":["http://hdl.handle.net/1773/48282"],"dc:language.iso":["en_US"],"dc:rights":["CC BY"],"dc:subject":["Linguistics","Computer science"],"dc:title":["Latent Compositional Representations for English Function Word Comprehension"],"dc:type":["Thesis"]},"updated_at":"2026-07-24T05:58:01Z"}