Lourdes Moreno, Rodrigo Alarcon, Isabel Segura-Bedmar, and Paloma Martínez. 2019. Lexical simplification approach to support the accessibility guidelines. In Proceedings of the XX International Conference on Human Computer Interaction (Interacción '19). ACM, New York, NY, USA, Article
https://dl.acm.org/citation.cfm?doid=3335595.3335651
3. 10% of the world’s population,
or 650 million people
lived with some type of disability
3
by 2050 there will be 2 billion
people over 60, or in Europe,
33% of society will be older
than 65
5. The readability and understandability of
texts that contain:
◎ long sentences
◎ unusual words
◎ complex linguistic structures
People with intellectual and learning
disabilities
A lack of approaches that improve
cognitive accessibility
=> Offer simplified texts
Introduction
5
6. Websites which include simplified text
content, the simplification has been
created manually
◎ Manual simplification is expensive
◎ Information is constantly being
produced
Few tools that provide systematic
support to simplification processes.
Introduction
6
8. Web Content Accessibility Guidelines
(WCAG) 2.1
◎ Standard to be followed in the majority
of the regulatory frameworks
Easy to Read guidelines
◎ People with intellectual disabilities
◎ International Federation of Library
Associations and Institutions (IFLA),
(ILSMH)
Plain Language guidelines
◎ All citizens, eAdministration documents
◎ Plain Language Action(PLAIN).
plainlanguage.gov ,
www.plainenglish.co.uk
8
9. The Natural Language Processing (NLP)
provides methods and systematic
support to comply with cognitive
accessibility guidelines
This work:
Guideline: “Use simple lexicon”
Develop lexical simplification system
using NLP methods
Spanish and not dependent-domain
9
11. Step 1) CWI
Machine Learning approach (Support-
Vector Machine (SVM) classification
algorithm)
Data set from Complex Word
Identification (CWI) Shared Task,
Workshop BEA 2018
◎ Information of Wikipedia pages in
Spanish (collection of 14,280 entries)
Features:
◎ Length and Frequency
◎ Binary
◎ Word Embedding
◎ Easy to Read (E2R)
Lexical
Simplification
System
11
12. Step 1) CWI
E2R dictionary which indicates if a
word is automatically qualified as
simple or not.
◎ Sources: web documents and
books in E2R elaborated by experts
(13400 simple words)
Process:
■ Tokenized
■ Lematized
■ Tagged
Lexical
Simplification
System
12
13. Step 1) CWI
Evaluation
Workshop BEA 2018
c
The E2R dictionary provides a
significant improvement of
almost 2 points,
Lexical
Simplification
System
13
14. Step 2) Substitution Generation
Select candidate substitutions for
each complex word
Linguistic database querying
strategy
◎ Dictionaries: Babelnet, Thesaurus
◎ For each complex word, a set of
synonyms are obtained
E2R dictionary and CWI process to
discard those complex synonyms
Lexical
Simplification
System
14
15. Step 3) Substitution Selection
The most appropriate candidate is chosen
based on it being the simplest option within
the context of the sentence
Semantic similarity filtering strategy:
detect the synonyms that had the most
similar meaning to the original word
◎ Word embedding models (Word2Vec,
FastTest)
Lexical
Simplification
System
15
16. User Interface for the system. It
generates accessible interfaces with
the generated simplified text
”Accessible technologies 2019" grant
awarded by:
EASIER tool
(ongoing)
16
17. Conclusions
Complexity of the textual content
◎ Cognitive barriers experienced by
people with intellectual and learning
disabilities
Proposal:
◎ Cognitive accessibility guidelines:
■ Easy to Read
■ Plain Language
■ WCAG
◎ A lexical simplification system
Ongoing and future research:
◎ System: better performance, metrics
to measure lexical difficulty
◎ Tool Easier
◎ Test with users
17
18. Lexical simplification approach to
support the accessibility guidelines
Lourdes Moreno, Rodrigo Alarcón,
Isabel Segura-Bedmar, Paloma Martínez