Hallo Language Assessment

Hallo offers fast, affordable, accurate language assessments for 50+ languages for individuals, enterprise, and education all through AI.

Hallo’s language assessment is scored based on fluency, vocabulary, grammar, pronunciation, and coherence; furthermore, it is mapped to common global standards such as CEFR and ACTFL.

- Assess your speaking skills in under 10 minutes with immediate results on both web and mobile.

- Rely on unbiased, valid scores from assessments all powered by AI and floor/ceiling rating.

- Significantly reduce your costs and time by 95% with our AI-powered tests and dashboard.

- Customize the assessment based on your particular industry such K-8 and K-12. 

Hallo can help schools and teachers ensure accurate language class placements with our streamlined assessments. Say goodbye to misplacement and hello to tailored learning journeys for your students.

Typically, a comprehensive language proficiency interview lasts for about 30 minutes, with an additional 30 minutes dedicated to reporting and communication, making it a one-hour commitment for the teacher and 30 minutes for the student.

Our automated assessment system can eliminate the need for the language assessment interview entirely, saving a full hour per teacher and potentially reducing expenses by $10-20 per student.

The founders of HalloAI (Joon Beh and Andrew Hale) speak 12 languages combined fluently. We've lived in many different countries and understand the importance of learning a new language early on. 

With our expertise in language learning, technology, and product, we look forward to impacting the lives of students and teachers through our solution.

Our goal is to make language assessment fast, affordable, and fun by making it more customized for K-8 learners.

We have built both speaking and writing language assessments for 56+ languages on mobile and web as well as a dashboard and API for B2B organizations.

We have served more than 100,000 users, and we are currently working with global organizations like BYU, Princeton, University of Texas, Majorel, and Berlitz. 

Utah

In schools, teachers spend way too much time assessing their students' proficiency levels each year. It's expensive and time-consuming.

Typically, a comprehensive language proficiency interview lasts for about 30 minutes, with an additional 30 minutes dedicated to reporting and communication, making it a one-hour commitment for the teacher and 30 minutes for the student.

Our automated assessment system can eliminate the need for the assessment interview entirely, saving a full hour per teacher and potentially reducing expenses by $10-20 per student, depending on their hourly rate.

We are automating all of these through AI for more than 50 languages so teachers can focus on the most important thing, teaching and meeting their students' needs. 

We can save 90% of the time and cost by automating language assessment through AI.

Website

All the front-end components have been developed by us, and we work with Google and Amazon for our databases.

Our assessment questions and formats are proprietary and developed by us.

For evaluating proficiency levels and pronunciation, we work with third party models such as ChatGPT and Azure.

Our adaptive assessment engine(Machine Learning) makes our assessment questions dynamic, which means that the difficulty level of the next question is tailored uniquely to each test taker based on their previous responses. We pull different questions from specific pools corresponding to the determined level, ensuring that every assessment experience is distinct.

Unlike human-rated proficiency interviews, Hallo's AI-driven assessment process eliminates human error and bias, achieving a remarkable accuracy rate of 95%. Our AI has undergone extensive training with thousands of samples to align with global standards such as CEFR and ACTFL, ensuring the delivery of accurate and valid scores.

Our assessments use criterion-referenced (CR) assessment methods, where the test taker's performance is evaluated against predetermined criteria rather than compared to the performance of others. This approach ensures that proficiency ratings are based on the individual's ability to meet specific criteria consistently. 

Floor and ceiling rating procedures are also applied to determine the highest level of sustained ability demonstrated by the test taker. The floor level represents the highest proficiency level where all criteria are consistently met, while the ceiling level indicates the next higher level where the test taker does not meet all criteria. 

Our adaptive assessment engine makes our assessment questions dynamic, which means that the difficulty level of the next question is tailored uniquely to each test taker based on their previous responses. We pull different questions from specific pools corresponding to the determined level, ensuring that every assessment experience is distinct.

In order to assess our test takers’ proficiency levels accurately, our scoring algorithms use 5 key criteria:

1. Fluency: Measures how smoothly and naturally a person can speak in the language without frequent pauses or hesitations. It reflects the ability to maintain a conversation or convey ideas without difficulty.

2. Grammar: Assesses the correctness and accuracy of the language structure, including sentence formation, verb conjugation, tense usage, and agreement. Good grammar is essential for clear and effective communication.

3. Vocabulary: Encompasses the range and richness of words and phrases a person can use accurately and appropriately. A broad vocabulary allows for more precise expression and understanding of ideas.

4. Pronunciation: Evaluates the clarity and accuracy of speech sounds, intonation, stress patterns, and rhythm. It can automatically detect miscues and provide accuracy score simultaneously, which provides more detailed information on omission, repetition, insertions, and mispronunciation in the given speech. Our pronunciation score represents an aggregated assessment of the pronunciation quality and includes four subaspects: 

• Accuracy: Evaluates the correctness of pronunciation.

• Fluency: Measures the level of smoothness and naturalness in speech.

• Completeness: Reflects the number of words pronounced correctly.

• Prosody: Assesses the use of appropriate intonation, rhythm, and stress including unexpected break, missing break, and monotone.

5. Coherence: Measures how well ideas are organized and connected within a conversation. It includes logical sequencing of thoughts, cohesion between sentences and paragraphs, and overall clarity of expression.

• Ray Clifford (2016). A rationale for criterion-referenced proficiency testing. Utah, USA: Brigham Young University  

• Brown, J. D., & Hudson, T. (2002). Criterion- referenced language testing. Cambridge, UK: Cambridge University Press. 

• ACTFL. (2020). ACTFL oral proficiency interview tester training manual. White Plains, NY: ACTFL. from http:// www.actfl.org 

• ACTFL. (2012). ACTFL proficiency guidelines— speaking.from http:// www.actfl.org 

To ensure a fair and personalized assessment experience, we have curated hundreds of different questions for each level pool in our assessment platform. This means that each user will encounter a unique set of questions tailored specifically to their proficiency level. By drawing from a diverse pool of questions, we can accurately gauge the test taker's language skills while also providing a fresh and engaging experience for every individual. This approach not only minimizes the risk of bias but also ensures that each user's strengths and areas for improvement are accurately assessed, contributing to a more comprehensive evaluation of language proficiency.

We have 3 full-time team members and 1 part-time team member.

So far, we've been working mainly working with learners and educators in higher ed, but we are starting to work with school districts in Utah and in Canada soon. For them, we've been developing questions that are more appropriate and geared towards K-8 and K-12 students, and they will be ready by the end of May, 2024. 

We are excited to work on more questions that are customized for public Transitional Kindergarten and Pre-K children as well as K-8 students, focused in particular on Black and Latino learners, and all learners experiencing poverty.

This is one of the main reasons why we built our product: to make language assessment faster, cheaper, and better.

Our pricing is designed to be approximately 5% of the standard exam fees. School districts in the United States are using a huge budget for language assessment each year, and we'd like to help save both time and budget significantly throughout the nation. 

We are mainly a product team, and we are good at building products. However, we don't have a marketing/sales team yet, so it would be super helpful if we could get some public exposure and the opportunity to work with more schools through the program.