Low-Resource Language Translation Application (LauReL)

Research shows that children whose parents are involved in their education are more successful not only academically, but also socially and emotionally. Unfortunately, it is almost impossible for refugees resettled in the U.S. who speak low-resourced languages to engage in their children’s education because they cannot read the information sent to them by the schools. They are unlikely to understand materials brought home by the student including documents for signatures, registration information, vaccine requirements, report cards, enrollment deadlines, parent-teacher conference schedules, information regarding extra-curricular activities, fundraisers, and changes to the academic calendar. The student, who perhaps understands English better than the parent, is forced to interpret for the parent, which puts everyone at a disadvantage.  

In the last 43 years, the U.S. has resettled approximately 31.5 million refugees. In 2024, an estimated 125,000 more will arrive. These families account for thousands of school-aged children whose parents speak a language that is not yet written or for which translation resources do not exist. Dallas-Fort Worth is among the top 3 resettlement destinations in the country, and most refugee families that arrive in the area settle in the Vickery Meadow neighborhood. Vickery Meadow is home to 20,000 refugees (about the seating capacity of Madison Square Garden). These families come from diverse cultures and languages, but are all served by the same elementary, middle, and high school in that district. There are 76 distinct languages represented in these schools. Many of these languages are low-resourced (having little to no online presence) and are spoken by people who are illiterate in all languages and do not speak English. This puts enormous pressure on the schools serving these children when trying to communicate with the parents. The school cannot provide translation/interpretation for each of these families. The result is that parents have no way to engage with the school and cannot play an active role in their children's education, leaving these students without crucial support. 

For this challenge we are focusing the initial solution on the Rohingya speaking community in Vickery Meadow. There are approximately 800 Rohingya refugees housed in this area. Most parents and children have never attended school and do not possess the ability to navigate the educational system in English. There is one refugee liaison for the district, and she does not have the tools to aid communication with linguistically isolated families. Students from these families are already at a disadvantage having suffered trauma, never attending formal education, and being unfamiliar with the culture and customs of the United States. Factor in the lack of parental guidance and support in schooling, and these students are not set up to succeed.  

This problem exists for thousands of refugee students and their parents throughout the United States. Refugee parents who were denied education know what a gift it is and want to be involved in their children's success, they just need to be equipped.  

A user can upload a school's communications or take a picture of it (a task Rohingya users are already comfortable with). The app then extracts the text from the document/image and sends that to a Large Language Model (LLM). The LLM then processes that information and makes a summary of it in English, in a specified format. This summary is then translated into the target language. That translation then goes through a text to speech process so that the user can listen to the summary. The user can then ask clarifying questions about the specific document/image, and the app will speak those answers (unimplemented in the prototype). LauReL Prototype Demo

The initial proposed solution will target Rohingya speaking parents of students living in the Vickery Meadow neighborhood of Dallas, TX. This refugee population is originally from Myanmar (Burma) and are sometimes referred to as the world’s most persecuted ethnicity. They are primarily pre-literate, meaning they do not yet read or write in any language. The Rohingya language only recently became formally written, so most Rohingya have never even seen their language written. Rohingya people are also largely unfamiliar with formal education because they were denied the opportunity to attend school in their homeland as well as in the Asian countries where they lived prior to resettlement in the United States.  

International Literacy and Development has been partnering with this community for 5 years to provide tools and resources to equip them to successfully navigate their new lives in the U.S. Though our programs primarily consist of in-home, trauma-informed ESL and citizenship education, inevitably we have been involved with school registration, social services, health care, and other essential needs of this community. Through our work, it has become apparent that communication with the schools is inaccessible to Rohingya parents.  

The Dallas Independent School District cannot find adequate interpreters for these parents and there are not digital translation tools to use. Also, Rohingya are unfamiliar with commonly used digital communication tools, like email, compounding the challenges that families and schools face in relaying critical information.  

The Rohingya parents want to be involved in their children's education, but cannot engage because all forms of written communication (email, flyers, and text messages) sent from the school are in English. This method poses two significant problems. The first is that the communication is in English. The second is that it is written. Literacy constraints in any language will be a barrier to successful understanding. Rohingya people are oral communicators and must receive information in spoken form for it to be understood. 

The translation application, LauReL, would greatly impact their lives as it would give parents access to communications from the school district, the school their student attends, and their student’s teacher. The app will take the written English information, summarize it, and speak the summary in Rohingya to the user. This will be a game changer for parents and students. The goal is that by providing a way for parents to be informed and involved in their children’s education, the children would be more successful academically, socially, and emotionally.  

ILAD has partnered with the Rohingya community in Dallas, TX. for 5 years to fill gaps in services provided by other agencies and resettlement organizations. The services we provide were developed from the beginning at the invitation and with the input of Rohingya stakeholders. We continue to rely on and value their input as we expand our initiatives in the community. The most critical thing we have learned through this input gathering processes has been the paramount nature of relationship, as it is the currency for everything that takes place in the Rohingya community. ILAD works within the framework of established social networks to provide services, working through Rohingya people themselves to that end wherever possible. 

ILAD has hired interpreters from within the community to help families navigate the schools, medical care, social services, COVID information and more. Our staff is consistently in a feedback loop with 30 Rohingya families concerning their needs and suggestions on different ILAD initiatives.  During the 5 years ILAD has worked with this population, we have developed a relationship with the Refugee Liaison for the Dallas ISD who has confirmed that the school district is not equipped to provide adequate interpretation/translation services for Rohingya speaking parents. Because ILAD staff has worked directly with the Rohingya refugees and partnered with the school district and other services organizations, we have developed relationships and trust with the insular and outsider-wary refugee population. We have tailored our programs to meet their expressed felt needs and we know that there is no “one size fits all” silver bullet. ILAD is respected and trusted, and we have deep personal relationships with the people who need access to the information sent from the schools their children attend.  

Because of our standing in the community, our steady presence, our on-going services, and our relationship with the schools and the school district, we are strategically positioned to deliver solutions that will be used to solve the problem of lack of access to information. 

Our team also consists of linguists, educators, developers, and experts in Human Language Technology. International Literacy and Development is focused on promoting human flourishing for under-resourced language communities by developing solutions in partnership with the people we serve.  

ILAD's team has developed and tested an app prototype that uses high resourced languages to confirm the pipeline and demonstrate the feasibility of the concept. This prototype allows users to capture and submit an image containing English text and receive back an audio summary in Russian.. All relevant stages of the data processing pipeline are in place including:

• image capture on client device
• submission of image to server
• server-side image English text extraction
• using generative AI to produce a simplified, translation-ready summarization of the extracted English text
• server-side translation of simplified, summarized text to target language (for the prototype: Russian)
• server-side text-to-speech of translated text
• playback of TTS audio on client device

Not included in this prototype:

• translation of summary to a Low Resource Language (planned: Rohingya)
• a chat interface to allow the user to ask follow-up questions about the document (planned)

International Literacy and Development is excited to learn from industry experts and collaborate with peers committed to equitable education for all students.  

The phase, beneficiary demographic, technology component, and legal considerations of our solution make ILAD a perfect candidate for the resources associated with this award.   

The prototype phase is the ideal time to adjust the technology, delivery, and functionality of the solution and to develop a robust plan to scale to new languages and sectors. Because technology and refugee communities are constantly changing, ILAD's LauRaL application must stay current and agile to be effective and beneficial for all stakeholders. The MIT peer network would enable ILAD to identify solution weaknesses and provide guidance in refinement.   

The Team Lead and other team participants would benefit from the Leadership Training offered through the Award. Building a strong team that considers the languages, cultures, and needs of the beneficiary community is of utmost importance. Consulting with other leaders and experts will provide Ms. Holland with tools, strategies, and ideas for ensuring that the needs of low-resource language communities remain a top priority.  

With technology evolving at a rapid pace, it will be imperative for ILAD to be connected to a network of experts in the field ranging from data collection tools, use of large language models, all the way to the UX/UJ front end developers. Working collaboratively to provide the best tools to our beneficiary communities will equip more parents and students for success. 

Additionally, the legal expertise provided through the MIT Solve Challenge will benefit ILAD if there are questions regarding privacy, informed consent, or other regulations that might impact our ability to serve the students in public education institutions. 

The resources, funding, and support provided by this partnership will empower us to overcome these barriers, expand our reach, and create a more inclusive and empowering educational environment for refugee parents and their children. 

The problem of access to information is felt most acutely by those who speak low-resource languages but must participate in society where they cannot read or speak the language of power. The innovation of this solution is its focus on low resourced languages. By providing the school and the parents with tools to communicate, refugee students will have increased opportunities to succeed. The use of translation apps, specifically text to speech, will reduce inequality by making information accessible to all regardless of language, and improving the opportunity for quality education while simultaneously taking the burden off of the overextended school teachers and staff. We believe this solution can be a catalyst for translation apps in all school districts that serve diverse language populations. We also believe that apps focusing on other domains, such as social services, emergency information, and healthcare can be created using the same technology to serve refugee communities and low resourced language speakers around the globe.  

To further clarify how this solution will improve the specific problem in focus, it is important to understand the current situation. The Dallas school system sends messages to all parents via written communication. These are in the form of e-mails, texts, flyers, documents, etc. Rohingya parents cannot understand them, and typically ignore them regardless of their importance.  Paperwork in general is a challenge for Rohingya families as it is an unfamiliar concept. Low-literacy levels compound the problem. Because ILAD staff is often in the homes of Rohingya speakers, we can sometimes help with the school-parent communication barrier. When we are visiting a Rohingya home, our team members are sometimes shown a message, flyer or email from the school and asked for an explanation. We provide the family with an explanation and send the information to our interpreter to record an oral message that we disseminate to our program participants through WhatsApp. They, in turn, share the information with the larger Rohingya population in Dallas. This method is hit and miss and is not consistently helpful in getting all the information needed into the hands of parents. This problem could be significantly improved by giving the school and the parents access to the LauReL translation app. The school could upload the communication and send the spoken Rohingya summary directly to the Rohingya parents. Rohingya speaking parents would also be able to upload documents and have a spoken summary of the information it contained. This solution could be scaled to other languages to meet the needs of more parents and students. The result would be that parents would consistently receive all school communication in a format that they can understand, thereby equipping them to be involved in their children’s education. 

Increased parental involvement is correlated with better academic, socioemotional, and behavioral outcomes for students—especially English language learners (He & Thompson, 2022; Niehaus & Adelson, 2014; LaRocque et al., 2011). Home-school communication is a crucial bridge that facilitates parental involvement (Kelty & Wakabayashi, 2020; LaRocque et al., 2011). For many migrant and refugee families, however, there are significant linguistic barriers impeding effective home-school communication (Kelty & Wakabayashi, 2020; Piller et. al., 2009; Uy, 2015). While most schools communicate almost exclusively in English and through written means (e.g. school websites, emails, backpack letters) (Kelty & Wakabayashi, 2020; Kraft & Bolves, 2022; Piller et al., 2023), many refugee parents do not speak English and may be illiterate—relying only on orally transmitted information in their native language (Rah et al., 2009). 

To improve home-school communication, mobile apps have been shown to be in high demand among parents, especially those of English language learners and those from low-income backgrounds (Kraft & Bolves, 2022). Machine translation tools (Google Translate) have also been demonstrated to be an effective means of conveying information to parents who speak high-resource languages like Spanish (Terantino, 2024). However, such tools are not yet available for speakers of low-resource languages like Rohingya. By empowering parents who speak low-resource languages to actively improve their home-school communication, the LauReL translation app will effectively address this need and improve refugee student success. 

 References 

He, B., & Thompson, C. (2022). Family Involvement and English Learners’ Outcomes: A Synthetic Analysis. International Review of Education, 68(3), 409–440. https://doi.org/10.1007/s11159... 

Kelty, N. E., & Wakabayashi, T. (2020). Family Engagement in Schools: Parent, Educator, and Community Perspectives. SAGE Open, 10(4), 2158244020973024. https://doi.org/10.1177/215824... 

Kraft, M. A., & Bolves, A. J. (2022). Can Technology Transform Communication Between Schools, Teachers, and Parents? Evidence from a Randomized Field Trial. Education Finance and Policy, 17(3), 479–510. https://doi.org/10.1162/edfp_a... 

LaRocque, M., Kleiman, I., & Darling, S. M. (2011). Parental Involvement: The Missing Link in School Achievement. Preventing School Failure: Alternative Education for Children and Youth, 55(3), 115–122. https://doi.org/10.1080/104598... 

Niehaus, K., & Adelson, J. L. (2014). School Support, Parental Involvement, and Academic and Social-Emotional Outcomes for English Language Learners. American Educational Research Journal, 51(4), 810–844. https://doi.org/10.3102/000283... 

Piller, I., Bruzon, A. S., & Torsh, H. (2023). Monolingual school websites as barriers to parent engagement. Language and Education, 37(3), 328–345. https://doi.org/10.1080/095007... 

Rah, Y., Choi, S., & Nguyễn, T. S. T. (2009). Building bridges between refugee parents and schools. International Journal of Leadership in Education, 12(4), 347–365. https://doi.org/10.1080/136031... 

Terantino, J. (2024). Communicating with Parents of Latino English Language Learners: Is Google Translate an Accurate Option? Journal of Latinos and Education, 23(1), 46–58. https://doi.org/10.1080/153484... 

Uy, P. (2015). Supporting Southeast Asian American Family and Community Engagement for Educational Success. Journal of Southeast Asian American Education and Advancement, 10(2). https://doi.org/10.7771/2153-8... 

Outputs: Activities (progress toward ILAD's key deliverables) 

Component of Logic Model: Test LLM/app with end-users 

Indicator: Number and demographics of end-users interviewed during app development 

Description: End-users interviewed during app development will be representative of the target population of Rohingya end-users in DFW. Major findings about translation quality and app usability from interviews will inform changes to the LLM and LauRal app.  

Component of Logic Model: Integrate app with Dallas school district’s communication system   

Indicator: Number of individual communications sent to app by the school system

Description: The school district's third party vendor for home-school communication will send all automated communications for parents (e.g.  emails and text messages) to the LauRel app to be translated and interpreted for Rohingya parents. The number of communications sent to the app will be tracked.  

Component of Logic Model: Inform and train Rohingya community about the app and its use  

Indicator: Number of Rohingya parents trained to use LauReL 

Description: We expect that ILAD and other refugee NGO's will train Rohingya parents to use the app. The total number of parents trained will be recorded.  

Mid-term Outcomes (progress on mid-term community impacts as a result of ILAD’s activities) 

Component of Logic Model: Rohingya parents begin to improve their access to school communications as they use the app 

Indicator: Number active users of app 

Description: The app automatically tracks number of active users. We assume that if Rohingya parents are continuing to actively use the app, it is improving their access to school communications.  

Component of Logic Model: Rohingya parents increase their engagement in their children’s education 

Indicator: Average number of education-related activities engaged in by Rohingya parents in last academic year  

Description: We will survey Rohingya parents before they start using the app about the number of ways they engaged in their children’s education last academic year. Parents will check all boxes that apply from a list of common ways to engage in education (e.g. attend parent-teacher conferences, attend an awards ceremony, help with homework). After using LauReL for at least one academic year, we will survey them again about the number of ways that they engaged in their children’s education in the past academic year.  

Component of Logic Model: Rohingya parents are telling others in the community about the app 

Indicator: Number of new downloads per month 

Description: The app tracks the number of new downloads per month. We assume that if the app continues to receive new downloads each month that it is spreading by word of mouth through the Rohingya community.  

Long-term Outcomes (progress on long-term community impact as a result of ILAD’s activities) 

Component of Logic Model: Rohingya parents are more adequately able to help their children succeed 

Indicator: Pre-/post- average student GPA 

Description: ILAD will rely on the school district to provide the average GPA for students of app users at the beginning of their app usage and after one year of app usage.    

The core technology enabling our solution is generative AI and machine learning, particularly the ability of current Large Language Models (LLM) to summarize English text in simple, easy-to-understand language. This simple summarization allows for less sophisticated machine translation engines to create usable translations that still maintain understandability. The complexity of the translation text (and thus the engine needed to translate it) is an important consideration when extending machine translation to Low Resource Languages (LRLs, like Rohingya) where there may not be a large amount of data readily available for training a highly robust translation engine. The output of the LLM can be tailored to conform to the limited training data set used in training the LRL translation engine.

Other technologies include:

• image text extraction (currently implemented using Azure Vision OCR, experimenting with multi-modal LLMs)
• Text to speech (implemented for Russian in prototype, (planned for target LRL Rohingya)
• LRL ARS for interpreting follow-up questions about text (not implemented in prototype - planned for Rohingya)
• RAG for providing relevant responses to follow-up questions about text (not implemented in prototype)

7 Full Time

1 Part Time

1 Contracted company for data collection

1 Contracted employee UX/UJ Development 

International Literacy and Development has been working in the Human Language Technology sector for 5 years focusing on low-resource languages. We spent these years researching and developing concepts and tools utilizing AI (Machine Learning, Neural Networks). We have developed systems for analyzing low-resource languages and created tools to expedite access to information and opportunities for the speakers of these languages.  

We have been collecting and analyzing Rohingya language data for 4 years and developed the concept of LauReL during that time. We began developing the prototype 6 months ago.

ILAD’s mission is to promote human flourishing through educational, economic and social development in communities that are under-resourced. This can only be achieved in true partnership with these communities to facilitate local solutions to local problems.  

ILAD operates projects around the world grounded in the belief that people have the ability and knowledge to create solutions to the barriers to human flourishing faced by their communities. ILAD’s work serves as only the catalyst for projects operated and managed by individuals from local communities.  Our organizational logic model ends with these communities contributing to society in new ways and teaching others to do the same.  

From the outset, our projects include input from the community and our Monitoring and Evaluation system provides the avenue for relevant improvements.  

ILAD’s workforce is intentionally selected for their capacity to contextualize their way of living and working to the location and culture where they choose to live long-term. ILAD’s workforce learns the majority or minority languages of the people where they have chosen to work, respecting local customs, traditions, and spiritual beliefs. 

Specifically, the Rohingya program employs 2 Rohingya refugee women and has two others in volunteer roles to serve their community. They are equal partners in the program, and their input is highly valued. ILAD relies on the broader community for accurate linguistic data and as experts in the need and relevancy of initiatives.  

ILAD’s business model is based on the need for parental involvement in their children’s education. When parents are not able to engage in their children’s education, there is less likelihood for academic, social, and emotional success.  

The key beneficiaries for this solution are the Rohingya refugee parents whose children are enrolled in the Dallas Independent School District, the DISD itself, and ultimately the students. We know that communication barriers are a pain point for all involved, and we know that consistent and reliable parent-school communication is critical for student success. 

ILAD is offering a translation/interpretation app that can be used by the school and by the parents to solve the communication barrier. Initial funding for the development and deployment of this app will come from individual donations and grants, but as this technology is scaled and adopted for use, we believe school districts will be willing to pay for a license to use the app to support their linguistically diverse student body. 

ILAD will implement a monitoring and evaluation system with indicators for success.  

Indicators will include: 

• Number of Rohingya refugee parents using the app. 

• Parents increase their engagement in their children’s education. 

• Students' academic success begins to show signs of improvement as measured by GPA 

As success of the Laurel app increases, ILAD will scale into other languages and sectors to provide translation and interpretation services through technology to low-resourced language speakers in the United States and globally. 

International Literacy and Development is a 501c3 in the business of promoting human flourishing in under resourced language communities. Human Language Technology is one of the areas of expertise we use to benefit these communities. Members of our staff have advanced degrees in computational and applied linguistics. Individual donors and grants provide the financial resources needed for these staff members and support staff. 

The LauReL app prototype solution was created by ILAD for the purpose of serving the client.  

Collecting and processing low resource language data for the LauReL app will be done through contractual agreements with tech companies specializing in this sector.  

The cost of those contracts, hosting the app, and ancillary expenses will be offset by the licenses sold to the public school system for the use of LauReL. We believe the benefit to the schools, parents, and students will be evident after the pilot phase with the Rohingya language. ILAD will then work with the district to determine which languages are priorities for this app and enter into an agreement with the school district.   

We envision scaling into other sectors including social services, legal, health, and employment. The companies, government entities, and/or organizations within those sectors will license the LauReL app for their use thereby allowing for its development and expansion.