Dyslexia impacts about 20% of the world’s population. We envision a future in which AI allows people to overcome this learning disability. It can be devastating, 48% of prisoners are dyslexic, according to a study by the University of Texas Medical Branch in conjunction with the Texas Department of Criminal Justice. Dyslexia is a general term for disorders that involve difficulty in learning to read or interpret words, letters, and other symbols, but that do not affect general intelligence. We have developed a strong intuition for a partial solution to their disorder: sentence simplification Natural Nanguage understanding+processing+generation. This task consists of modifying the content and structure of a text in order to make it easier to read and understand, while preserving its main idea and approximating its original meaning. A simplified version of a text could benefit low literacy readers, in every language, children and adults alike, for people with aphasia, autism, not just dyslexia. In addition, simplifying a text automatically could improve the performance of other NLP tasks, such as parsing, summarization, information extraction, semantic role labeling, and machine translation.
- TurkCorpus: MUSS Multilingual Unsupervised Sentence Simplification by Mining Paraphrases https://paperswithcode.com/paper/multilingual-unsupervised-sentence
- Newsela: Neural CRF Model for Sentence Alignment in Text Simplification https://paperswithcode.com/paper/neural-crf-model-for-sentence-alignment-in
- ASSET: Sentence Simplification with Deep Reinforcement Learning https://paperswithcode.com/paper/sentence-simplification-with-deep
- PWKP / WikiSmall: Hybrid Simplification using Deep Semantics and Machine Translation https://paperswithcode.com/paper/hybrid-simplification-using-deep-semantics
- EASSE: Package that facilitates the process of of evaluating and benchmarking sentence simplification models on commonly used datasets. (link)
- MUSS paper (link)
- Perform simple prediction using MUSS model:
- Validate on the test datasets (as per table 2 in the paper):
- Perform simple prediction using MUSS model:
