Improving scholarship accessibility with reinforcement learning
Introduction. A vast amount of scholarly work is published daily, yet much of it remains inaccessible to the general public due to dense jargon and complex language. We introduce a reinforcement learning approach that fine-tunes a language model to rewrite scholarly abstracts into more comprehensib...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
University of Borås
2025-03-01
|
| Series: | Information Research: An International Electronic Journal |
| Subjects: | |
| Online Access: | https://publicera.kb.se/ir/article/view/47530 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850067818123362304 |
|---|---|
| author | Haining Wang Jason Clark Hannah McKelvey Leila Sterman Gao Zheng Zuoyu Tian Xiaozhong Liu |
| author_facet | Haining Wang Jason Clark Hannah McKelvey Leila Sterman Gao Zheng Zuoyu Tian Xiaozhong Liu |
| author_sort | Haining Wang |
| collection | DOAJ |
| description |
Introduction. A vast amount of scholarly work is published daily, yet much of it remains inaccessible to the general public due to dense jargon and complex language. We introduce a reinforcement learning approach that fine-tunes a language model to rewrite scholarly abstracts into more comprehensible versions.
Method. Our approach utilises a carefully balanced combination of word- and sentence-level accessibility rewards to guide the language model in substituting technical terms with more accessible alternatives, a task which models supervised fine-tuned or guided by conventional readability measures struggle to accomplish.
Analysis. We evaluate our model’s performance through readability metrics, factual accuracy assessments and language quality measurements, comparing results against supervised fine-tuning baselines.
Results. Our best model adjusts the readability level of scholarly abstracts by approximately six US grade levels—in other words, from a postgraduate to a high school level. This translates to roughly a 90% relative improvement over the supervised fine-tuning baseline, while maintaining factual accuracy and high-quality language.
Conclusions. We envision our work as a step toward bridging the gap between scholarly research and the general public, particularly younger readers, and those without a college degree.
|
| format | Article |
| id | doaj-art-e6736e8a10ce41979446b2a94b60ed95 |
| institution | DOAJ |
| issn | 1368-1613 |
| language | English |
| publishDate | 2025-03-01 |
| publisher | University of Borås |
| record_format | Article |
| series | Information Research: An International Electronic Journal |
| spelling | doaj-art-e6736e8a10ce41979446b2a94b60ed952025-08-20T02:48:12ZengUniversity of BoråsInformation Research: An International Electronic Journal1368-16132025-03-0130iConf10.47989/ir30iConf47530Improving scholarship accessibility with reinforcement learningHaining Wang0Jason Clark1Hannah McKelvey2Leila Sterman3Gao Zheng4Zuoyu Tian5Xiaozhong Liu6Indiana University BloomingtonMontana State UniversityMontana State UniversityMontana State UniversityAnt GroupMacalester CollegeWorcester Polytechnic Institute Introduction. A vast amount of scholarly work is published daily, yet much of it remains inaccessible to the general public due to dense jargon and complex language. We introduce a reinforcement learning approach that fine-tunes a language model to rewrite scholarly abstracts into more comprehensible versions. Method. Our approach utilises a carefully balanced combination of word- and sentence-level accessibility rewards to guide the language model in substituting technical terms with more accessible alternatives, a task which models supervised fine-tuned or guided by conventional readability measures struggle to accomplish. Analysis. We evaluate our model’s performance through readability metrics, factual accuracy assessments and language quality measurements, comparing results against supervised fine-tuning baselines. Results. Our best model adjusts the readability level of scholarly abstracts by approximately six US grade levels—in other words, from a postgraduate to a high school level. This translates to roughly a 90% relative improvement over the supervised fine-tuning baseline, while maintaining factual accuracy and high-quality language. Conclusions. We envision our work as a step toward bridging the gap between scholarly research and the general public, particularly younger readers, and those without a college degree. https://publicera.kb.se/ir/article/view/47530Accessible languageLanguage modelText simplificationReinforcement learningProximal Policy OptimizationOpen science |
| spellingShingle | Haining Wang Jason Clark Hannah McKelvey Leila Sterman Gao Zheng Zuoyu Tian Xiaozhong Liu Improving scholarship accessibility with reinforcement learning Information Research: An International Electronic Journal Accessible language Language model Text simplification Reinforcement learning Proximal Policy Optimization Open science |
| title | Improving scholarship accessibility with reinforcement learning |
| title_full | Improving scholarship accessibility with reinforcement learning |
| title_fullStr | Improving scholarship accessibility with reinforcement learning |
| title_full_unstemmed | Improving scholarship accessibility with reinforcement learning |
| title_short | Improving scholarship accessibility with reinforcement learning |
| title_sort | improving scholarship accessibility with reinforcement learning |
| topic | Accessible language Language model Text simplification Reinforcement learning Proximal Policy Optimization Open science |
| url | https://publicera.kb.se/ir/article/view/47530 |
| work_keys_str_mv | AT hainingwang improvingscholarshipaccessibilitywithreinforcementlearning AT jasonclark improvingscholarshipaccessibilitywithreinforcementlearning AT hannahmckelvey improvingscholarshipaccessibilitywithreinforcementlearning AT leilasterman improvingscholarshipaccessibilitywithreinforcementlearning AT gaozheng improvingscholarshipaccessibilitywithreinforcementlearning AT zuoyutian improvingscholarshipaccessibilitywithreinforcementlearning AT xiaozhongliu improvingscholarshipaccessibilitywithreinforcementlearning |