Grafos, a digital platform to master Spanish spelling

Spanish spelling is one of the easiest in the world, yet many people still struggle to write properly in their mother tongue. Bad spelling is particularly high among students from the public education system, and is often seen as a symptom of poor education and communication skills. All these factors add to a person's unemployability.

Grafos is an adaptive online learning system for young people and adults who want to improve their writing. Based on a diagnostic test, the platform creates a customized learning path of micro-lessons with simple but cognitively progressive exercises and personalized vocabulary. Lessons are created on the fly and can be repeated as many times as needed, with different input. 

Writing is a basic cross-disciplinary skill required for the vast majority of professional activities. Grafos can help anyone with access to an electronic device to write without errors! 

Our project seeks to broaden pathways for employment to young adults thanks to a digital learning platform that improves writing skills through automation, AI and data.

Bad spelling is a widespread and widely recognized problem in Latin America and the Caribbean. A myriad of elements lie at the root of the problem, amongst which are low educational levels, poor teacher training and inadequate teaching strategies, input riddled with errors in social and other media, the loss of a culture of reading, and a general lack of interest in correct writing. These factors are unlikely to disappear with time. 

Moreover, it is hard to find quality tools to improve spelling as an adult. Face-to-face spelling courses abound, yet they often share the methodology and syllabus of the school training plans that have proven to be ineffective for so many years. Also, due to their cost and on-site modality, they tend to be inaccessible for those who need them most. Online resources on spelling and grammar are plentiful too, yet in many cases they are just stand-alone lessons with very few examples and exercises, or they lack the interactive and pedagogical strategies needed to achieve in-depth learning.

It is extremely common to read texts, written by students, professionals or teachers, that evidence poor spelling. Google’s search engine confirms the dimension of our potential public: tens of thousands of searches are performed every month with keywords like “curso de ortografía” o “cómo escribir bien”.

The problem is particularly pressing among people with lower educational levels (35.6% of Latin Americans do not complete secondary education) and students of the public education system (making up about 80% of the total), but is by no means limited to this population. Many universities, including private ones, are forced to offer special preparation courses to push students’ spelling and writing to acceptable levels.

A lot of adults enroll on continued learning courses, either through personal initiative or sponsored by their employers. However, these options are not accessible to everyone, and they often lack a modern, holistic, interactive and personalized approach to learning.

With 4G mobile wireless penetration in Latin America set to reach 90% by 2022, an online adaptive learning system like Grafos, which allows people to study for free, any time and anywhere, through short two-to-ten-minute sessions, can go a long way in helping to address the problem.

A different path for every user

Flexible settings. Grafos is composed by a myriad of 2-minute lessons that focus on very specific language topics and, put together, cover the whole of the Spanish spelling landscape. Users choose the time, place and density of their practice.

 Multi-level. Lessons are divided into four levels of mastery according to their impact on improving spelling. The structure progresses from the most frequent errors and vocabulary to spelling exceptions and professional lexicon.

 Personalized content. The learning path will adapt continually in view of the students' work and results. Lessons will include users' frequent mistakes and vocabulary related to their fields of interest.

Effective pedagogy

Choosing the most effective learning tools to maximize results is at the core of Grafos’ design.

 Multiple skills. Grafos uses different learning methods based on scientific research on language acquisition and memorization. It has four main lesson types:

●     “Palabras difíciles”: lessons concentrate on memorization to learn the spelling of difficult stand-alone words;

●     “Reglas y patrones”: encourage pattern identification of highly productive spelling rules;

●     “Parejas en conflicto”: focus on the context in which words appear, to distinguish homophones through grammar functions;

●     “Derivados”: motivate morphological analysis to infer the spelling of new words.

 Ample, varied and interweaved practice. Regardless of the level and topic, practice will be ample, with abundant and varied input (both visual and oral), presented within multiple cognitive challenges. Everyday examples take precedent over literary or technical texts since familiar texts allow users to concentrate on form over meaning.

 Progress analytics and gamification. In order to keep learners’ motivation, Grafos' user interface will include progress graphics and gamification features like badges, scoreboards and personal challenges. 

From spelling to writing

Through its exercises focused on spelling and grammar, Grafos contributes to develop multiple other skills such as visual memory, deductive and inductive thinking, attention to detail, and the analysis of language and ideas. These skills are essential to compose any kind of text and are transferable to other domains.

 Even so, proper spelling and specific vocabulary are not sufficient to write well, which is the ultimate goal of Grafos. With this consideration in mind, in a second phase of the project (4-5 years), we envisage an extension to Grafos’ original clusters of automatic lessons: a peer-review editing platform, where users can submit their texts and receive feedback.

A different course for every user

Spelling courses available today are not fit to provide quality training to young students and adults on a large scale. They either are expensive, face-to-face, or limited to a few random lessons that offer the same content and exercises to everyone. Our target has already received language training to different degrees, so their knowledge and needs will vary significantly. Grafos offers a holistic training program that will adapt to each individual learner.

Design that maximizes results

Grafos prioritizes the topics most needed to improve spelling. Its learning sequence takes into account the frequency of personal mistakes, the productivity of spelling rules, and the lexical field in which the words are used. Existing spelling courses does not follow these considerations, as they tend to organize their content only by linguistic criteria (problematic letters, rules and exceptions).

Effective pedagogy to develop multiple skills

Unlike traditional writing courses, Grafos’ exercises promote more than visual memory. We designed different learning paths to encourage pattern identification, attention to form and context, morphology and syntax analysis, among other essential skills to address the trickiest of spelling challenges.

AI, language corpuses and databases for educational purposes

Both the data sets and AI technologies required to create Grafos have been available for quite some years now (ie. in spell checkers, automatic translators, text recognition software, etc.). We propose to leverage the potential of these resources to address a broad educational issue in Latin America: poor spelling and writing.

Attainable goal

Concrete verifiable results are key to engage adults to study outside of formal settings. Improving someone’s spelling is a clear, short-term goal. Studies show that, in Spanish, around 5 words are responsible for 10% of the mistakes and 67 of 60%. The basic lexicon of Spanish speakers varies from 2000 to 5000. The immediate impact that Grafos is bound to have on users’ spelling is likely to motivate them to continue their training and develop more complex skills.

A bridge towards higher-order skills

Writing words without mistakes requires visual memory, a fair understanding of the syntax and phonetics of the language as well as a constant analysis of its internal structure. It is also closely related to lexical acquisition. Grafos lessons are designed to promote all these linguistics skills, which will have an impact on the user’s overall communication skills.

A renewed attitude towards written language

Bad spelling reveals more than an incomplete language training; it can be a sign of indifference, even disdain, towards the written language. As experts point out, behind the ability to write correctly necessarily lies the desire to do so. However, students often have a troubled relationship with writing due to bad experiences in school, and many think that writing proficiency is relevant only for certain professions. We believe that, provided with the right strategies, these students can change their attitude towards writing and embrace it as an essential tool in their personal toolkit.

Sources: Paper, Virtual Educa 2015.

• In one year: deploy a pilot project in Mexico in collaboration with some high schools, universities and research centres (approx. 1000 users).

• In three years: launch the system in Mexico and some other countries, primarily through educational institutions, in web mobile-first format.

 
If we reach 0.1% of Mexicans aged 15+ (approx. 94 million), then we would be serving 94,000 people.

If some of the major public high schools like UNAM’s CCH (with over 58,000 students) adopted the system, an equally large number of users would be very easily reached.

• In five years: offer the system to the general public across the continent, in web and App format, with the necessary adjustments for certain country-specific linguistic usages. 

If we reach 0.1% of Latin America’s approximately 300 million Spanish-speaking citizens aged 15+, then we would be serving roughly 300,000 people.

Over the next five years, we plan to unfold the different levels and functional layers of Grafos.

Year 1: fully functional web prototype for the basic level (first level out of four).

This includes:

• Definition of the complete learning roadmap for level 1 

• Creation of the user interface

• Manual and semi-automated creation of micro-lessons for level 1 + diagnostic test

• Definition of the system’s underlying database structure

• First manual and semi-automated DB population through sourcing and tagging of item input.

Year 2-3: full-blown system for levels 1-3, with automatic lesson creation, in web mobile-first format.

This includes:

• Definition of the learning roadmap for levels 2 and 3, and corresponding diagnostic test

• Algorithms to create items and lessons automatically

• Data mining for fully automated DB population with real-world text examples for all levels

• Addition of individual user progress and monitoring tools

• Services to enterprises and institutions, connecting the platform with their user DBs and providing group-specific content, testing and monitoring.

Year 4-5: all four levels, in web and App format, with tropicalization and community-based services.

• Development of level 4: Master

• Adjustments to DB material for certain country-specific linguistic usages

• Inclusion of community-led services and gamification: online editing and proofreading, achievement (con)tests, ...

• Development of iOs and Android App with updating mechanism for offline study

Building a complex adaptive learning system like Grafos requires a multidisciplinary team of at least nine people: a project manager, a linguistic expert, two language teaching experts, an IT expert, a database/data mining specialist, an HTML/mobile App programmer, a web designer, and two more junior language experts (e.g. students) in the supportive role of reviewing and tagging the input that will be stored in the databases, and testing. We already have candidates for these posts, including the five members of the original team; however, we need financing to get them on board on a part-time or full-time basis.

Another key factor in Grafos’ success is the involvement of external experts and institutions like schools and universities: sharing ideas and receiving input from teachers and linguistic experts who are not actively participating in the construction of Grafos, and getting constant user feedback are all essential in order to adequately define, then tweak and improve the overall functionality and user experience of the tool. In this respect, gradually growing our customer base will also be a challenge.

Finally, we will need access to corpuses like CREA (Corpus de Referencia del Español Actual) hosted by the Royal Spanish Academy; the Spanish Corpus, managed by Mark Davis of the Brigham Young University; and even Google Book’s databases. These resources offer texts and word lists classified by country, field, and frequency of use; the Spanish Corpus also includes part-of-speech tags. We also need to improve our knowledge of data mining/management.

The main barriers can be easily overcome with the right resources: at present, we have (most of) the local know-how to build Grafos, but need time and money. 

Our first step is to apply for financing through initiatives like TPrize and governmental funds, in order to build a fully functional prototype with a nuclear team of 6-7 people. 

The next step will consist in approaching universities, schools and research centres and invite experts to participate in the more long-term design, development and deployment of Grafos. At the same time, research centres would need to be contacted in order to gain access to corpuses and other online lexicographical resources.

In terms of customers, in the pilot stage, our end users will be students of the institutions we’ll be partnering with. However, during the second phase of development (2-3 years), the team will need to be extended with a PR/Communications Coordinator, in order to grow our user base through third parties and other direct promotion channels.

The two people who are applying to this TPrize Challenge are currently working as independent contractors:

• Adriana Xhrouet, an e-learning and language teaching expert
• Tine Stalmans, an e-learning specialist and online project manager

Three more people have participated in the development of the Grafos demo at UNAM:

• An IT expert
• A programmer
• A graphic designer

In different development stages, we will also need:

• One or two linguistic experts
• A second language teaching expert
• A database specialist
• A data (mining) expert
• Two junior language experts (e.g. students) for the reviewing and tagging of input, and testing.
• A PR/Communications Coordinator

As ‘spiritual parents’ of Grafos, apart from being extremely motivated and enthusiastic about this project, we also have an intrinsic understanding of its nature, objectives, complexities, and developing needs. Moreover, our own cross-disciplinary profile ―with learning and professional experience spanning the instructional, technological and linguistic realms― allows us to have an overall understanding of the project and coordinate the definition and development of its multiple aspects. 

Adriana Xhrouet Aguilera, BA in History and a Specialization Degree in Spanish Language Teaching. She has worked for seven years as an e-learning specialist, focusing primarily on the  development of online resources for young Spanish speakers to help them improve their writing skills.

Tine Stalmans, BA in Languages and Chinese, MA in Publishing Studies, has more than 15 years of experience in designing and developing online learning systems and courses, especially for teaching Spanish spelling and writing to native speakers. She also has solid programming knowledge and experience in online project management.

We are not actively partnering with any organization, but have ties with the following institutions, which all have expressed interest in the project:

• DGTIC, Universidad Nacional Autónoma de México (UNAM): provided technical counselling and support for the development of the Grafos demo; several of its employees are considered as the IT specialists of the development team.

• Academia Mexicana de la Lengua: can provide linguistic feedback and expertise.

• Universidad del Claustro de Sor Juana: interested in participating with teachers from a pedagogical perspective, and for user feedback and testing purposes. 

• Universidad Autónoma de México (UAM): possible collaboration as academic consultants and participation in pilot project.

Our main target are adults with an incomplete language training, which often is the cause of an unsatisfactory performance both at work and in the academic environment. As language skills are essential to develop other high-order skills, this population struggles to fulfill basic job requirements and provide quality services. Grafos is an adaptive learning platform that will help them improve their writing skills, raise the quality of their work and gain access to better-paid employment possibilities. 

Universities, governmental departments, and private enterprises are concerned with the poor writing skills of their students and employees, since it affects their image, productivity, and results. They are our principal stakeholders and, as such, they also are key candidates for partnerships and financing in the first development stages. Funding through private trustees and foundations will also be sought.

Once the platform is up and running, we plan to offer new and more specialized services to improve Spanish writing, like a peer-review editing platform, online workshops, paid-for language-level tests, and customized implementations for enterprises.

Our services will be available on the web and as an app. We have two main strategies to reach our public: through partnerships with universities and schools, and through online campaigns (mainly on social media). 

We will evaluate our results regularly measuring the number of users versus the number of potential users; their rate of improvement and retention, and our coverage in Latin America.

We are targeting the middle to low income population. This implies keeping the main services free for those who most need them, while charging certain specific services to those who have the necessary resources. We will also look for partnerships and community work to develop the platform.

We plan to fund our work with different revenues in the different stages of the project:

• Donations, grants and fundraising campaigns. Universities, governments, private trustees and international organizations offer funding to projects like Grafos. This funding will be essential to complete the first (prototype) and second stage (levels 1-3) of development.

• Partnerships with universities, internships, co-op. Graduate and undergraduate students contribute their expertise and acquire work experience; universities assign academics and teachers to collaborate with the project.

• Community volunteers. Once the platform is launched, we will encourage a community-based approach to offer new content and reduce the maintenance costs to a minimum. The peer-review editing extension planned for phase 3 will also provide us with categorized, tagged and corrected input for the lesson clusters.

• Paid services. Towards the end of the second development phase,  we plan to offer paid services to enterprises, i.e. company-specific implementations of the platform with customized vocabulary, programmed and supervised tests, collective progress monitoring and reporting, etc. 

• Publicity. Even if we are reticent to use it, adding publicity is a possible strategy to generate revenue.

We believe TPrize can help on several fronts:

1. Partial financing of the Grafos development team

2. Mentorship and training on data mining and AI

3. Connection with experts on the above-mentioned technological issues

4. Support with servers and hosting of Grafos

5. Expertise on social business development
   

As mentioned in previous sections, partnerships with academic and other institutions will be crucial for different purposes:

• We want to collaborate with academic institutions and research centres like UNAM, UAM, Academia Mexicana de la Lengua y Universidad del Claustro to receive input and feedback on linguistic and pedagogical issues.

• We will need the involvement of schools and universities to deploy and test the platform in its different stages of development.

• Partnership with tech companies or universities will be optimal for training and support with data mining and AI.

• Access to corpuses created by the Royal Spanish Academy, Google and other organizations will also be essential to the project’s success.