IIT Bombay’s new web application, IMPART, allows researchers to track changing water surface temperatures and can help to track climate change

IIT Bombay’s TARA mobile app to help achieve nationwide oral reading fluency

Read time: 1 min
Mumbai
30 Nov 2024
Teacher Facilitating the test of a student (Credits: Researchers)

Over the past decade, the focus of educational reform has been gradually shifting from increasing school attendance to improving the quality of education. Despite this, more than half of Class 5 students could not read Class 2 level text as per the Annual Status of Education Report (ASER) of 2022. While the NEP 2020 report indicated that at least five crore students in India were yet to attain foundational literacy and numeracy (FLN) skills, the Covid-19 pandemic worsened the situation as 90% of students lost at least one specific language ability, such as describing a picture or reading with understanding. 

With reading considered a gateway skill to all other learning, governments across the globe are focused on action to address this aspect of the learning crisis by developing processes for structured teaching methods and training teachers to deliver content. In such a scenario, it is easy to appreciate the potentially crucial role of monitoring learning outcomes and using the data to guide instructional strategies. 

Regular assessments play a key role in ensuring that FLN outcomes are measured and improved over time. They can inform course corrections for better learning trajectories of a child. Literacy and language skills are traditionally assessed with a battery of tests on a one-on-one basis employing large numbers of well-trained evaluators. As an example, an essential component of the test and a critical indicator of reading proficiency is oral reading fluency (ORF), which is implemented by listening to a child read aloud from printed text and manually scoring attributes such as accuracy, speed and smoothness. 

In a unique effort to transform literacy assessment to be more scalable, objective, and reliable, researchers from Indian Institute of Technology (IIT) Bombay, led by Prof. Preeti Rao, Department of Electrical Engineering, have teamed up with language experts and teachers to create a mobile app to measure oral reading fluency automatically using speech processing and machine learning technology. From an audio recording of a child reading a level-appropriate passage aloud, the app, called TARA (Teacher’s Assistant for Reading Assessment), extracts rubrics for ORF including the widely employed WCPM (words correct per minute). Expression is another important dimension of fluent reading that is strongly linked to the reader’s understanding of the text. With TARA, phrasing (grouping of words), intonation and stress in speech are also measured to obtain a holistic score that is indicative of the precise stage of reading development.

“The system is trained on expert-annotated recordings of children’s reading and currently works for English and Hindi, with its reliability verified to match that of human experts,” shared Prof. Rao.

 

Sample report card generated by TARA-based evaluation (Credits: Researchers)
Sample report card generated by TARA-based evaluation (Credits: Researchers)

Dr. Shailaja Menon, reading pedagogy expert and Lead, Center of Excellence in Early Language and Literacy at Tata Trusts says, “Organisations have long felt the need for a digital tool offering real-time data on learning levels.”

TARA addresses this gap with an end-to-end system that facilitates audio recording and provides performance data for each child, as well as for cohorts such as class, school and region, on a dashboard. 

The project received funding from the Tata Centre of Technology & Design and Abdul Kalam Technology Innovation Fellowship as well as some traction in the school education community.

TARA has been recently adopted by the Kendriya Vidyalaya Sanghatan (KVS) for English and Hindi ORF assessment for Grades 3-8 involving over 7 lakh students in 1200 schools across India, making it by far the largest such exercise undertaken in the country. With the baseline test already completed in October this year, valuable data on ORF benchmarks has been generated for six school grades for both the languages. 

The association with KVS is especially significant in view of the NIPUN Bharat National Mission where KVS schools are expected to serve as model schools for the attainment of FLN by all students by the end of Grade 3, thus acting as pioneers for competency-based education at primary level and adoption of learning outcome metrics. 

The TARA team continues to collaborate with KVS to develop effective remedial instruction for students to help improve their reading capabilities. The effectiveness of the remediation will be apparent in the next assessment phase, resulting in overall benefits to learning from the regular test & practice cycles distributed across the calendar year. IIT Bombay welcomes new partnerships and collaborations with the larger goal of facilitating evidence-based solutions for school education.