Have you ever experienced a classroom where the teacher seems to do all the talking, they rarely pose open-ended questions, and do not foster active discussions? Some lessons revolve entirely around students listening, capturing, and echoing the teacher's words.

Teacher development and coaching programs strive to reshape this dynamic by organizing training events that bring teachers together, encourage role-playing, and the adoption of fresh techniques for delivering student-centered lessons. As a teacher, you may have experienced a handful of these sessions, and done your best to implement the advice, but there remains a gap – a lack of insight into what truly happens within your classroom.

Are you interested in understanding the distribution of talk time in your class? Now, you can use Loquat to discover practical insights for your lessons. Loquat is designed to support educators and provide guidance in facilitating balanced student participation in each class. There is no one formula to bring success to your lesson but with Loquat you will be able to measure and ensure sufficient student involvement during each activity.

Why use artificial intelligence and machine learning?

The use of technology can revolutionize a classroom. Currently, there is a lot of talk about Artificial Intelligence (AI)--one type of AI is Machine Learning (ML). It uses pre-trained models to perform various tasks. The most basic and yet powerful of these is the Speech Activity Detection (SAD) model which can detect when someone is talking. This model enables one to precisely measure the duration of speech within the classroom. This metric serves as a valuable indicator of “active time”. Why stop there? By incorporating additional speech characteristics, such as pitch, speed, and energy levels, the analysis can be further enhanced.

Children exhibit distinct speech patterns. They have higher pitched voices, speak faster, and their speech has lower energy levels. By integrating these extra details into the “active time“ analysis, a powerful picture of who is talking in the classroom, and for how long, can be built. Loquat provides a detailed visual breakdown of the duration of speech for both children and teachers. Let's explore how teachers can benefit from this information.

Teacher gains from speech analysis

Formal training and coaching programs are crucial for teacher development, however, not all educators have access to such resources. That's where ”Loquat” comes in. By capturing audio from an activity or entire lesson, teachers can leverage the app to send this audio for processing. The recording is uploaded to the Loquat server and once fully processed, the analysis provides a range of reports that are available to teachers on their phones. These reports are designed to initiate self-reflection, prompt immediate corrective actions, and aid teachers in enhancing talk time balance in the classroom.

One of the notable reports used in Loquat is the “Speaker Wheel”. This visual chart provides teachers with a snapshot of talk time distribution in the processed recorded session. Empowered with this report, teachers can engage in self-reflection and make immediate adjustments for their next lesson. Further down the page, feedback and suggestions guide teachers in proactively improving classroom balance in subsequent lessons.

While a teacher may prompt student participation during a lesson, it's essential to assess the nature of that engagement. Often, participation may be confined to binary questions, limiting the depth of student interaction. Achieving a balance between open-ended and binary questions is crucial in student-centered learning. Loquat offers another insightful report, the "Statements" report, which uncovers a distinct dimension of the classroom activity. This report delves into the duration of each statement made by both the teacher and students. If the average duration of a student’s statement is short (1-2 seconds), it indicates the absence of open-ended discussions. In such cases, students may not have articulated their thoughts or ideas thoroughly, opting for brief, one-word responses. Again, the teacher has access to a brief feedback section, where they can find more information and links to topics on how to pose open-ended questions.

Pilot testing

Loquat was piloted in Ghana, Guatemala, and Senegal.


Two women looking at report on LoqautThe initial pilot phase of Loquat was introduced in Ghana. In the first trial, teachers explored the app across various subjects, including Math, Science, Language (reading and reading comprehension), Our World Our People, and Debate in grades 1-5. While teachers appreciated the app's sound processing and report generation capabilities, there was less initial interest in exploring strategies to enhance talk time balance.

Encouraged by the valuable insights from the first trial, we expanded the pilot in Ghana to include 14 teachers in the second phase. This time, we extended the application to KG1, KG2, and grades 1-5. Following the initial analyses, teachers expressed eagerness to improve their teaching approach after seeing the visual reports. They actively incorporated more activities for children, employing gestures to reduce their talk time and direct students. The adoption of a child-centered approach emerged as a strategy to involve more students in the learning process.


In Guatemala, the pilot program involved teachers who participated in the application trial. The trial included subjects such as Math, Spanish, and Art Expression in grades 4-6. Teachers praised the app for its user-friendly interface and its ability to provide valuable insights that fostered increased interaction between students and teachers. Interestingly, none of the teachers utilized the self-reflection feature to discover ways to achieve a better balance.


In Senegal, the team was interested in learning about talk time balance in terms of spoken language. Therefore, we developed a special version of the app to determine if the language of instruction was French or the local language (i.e., "not French"). All reports were adjusted to show French or local language distribution throughout the lesson without disaggregating on adult or child. 

Frequently Asked Questions

I want to use Loquat in my country, do you support my language?

Rest assured, Loquat is a language-agnostic tool. It doesn't analyze the spoken language or the content of the speech. Instead, it utilizes speech characteristics to assign attributes, creating reporting summaries. In simple terms, the app can be seamlessly used in any country, supporting diverse languages and instructional contexts.

What about privacy?

Respecting user privacy is a top priority. When it comes to recording lessons, the audio stays exclusively on the teacher's device, ensuring confidentiality; it cannot be replayed or extracted. On the server end, all recordings are removed immediately after processing. The retention period is about 10 minutes before deletion. Additionally, the ML model only analyzes "speech" versus "non-speech". It does not record actual words. Loquat knows when someone is talking, but it does not know what that person is saying.

What comes next?

Looking ahead, we want to see Loquat evolve into a comprehensive teacher assistant tool. Our roadmap includes the addition of voice identification models that empower teachers to monitor specific student participation and may help in improving inclusivity. Going forward, we hope to incorporate affective models, enabling assessments of both classroom climate and content.



About the Expert

lhristov's picture
Senior Education Technical Advisor for RTI International's International Development Education Group. He has over 15 years of experience in software development, implementation, and support. He is responsible for the successful implementation, adoption, and client support for over 20 internal RTI data collection efforts and more than 30 external (SaaS subscribers) clients using all varieties of Tangerine® software in the education sector. Mr. Hristov is also involved in the design and development of RTI's "Loquat", a machine-learning mobile application that helps teachers balance their talk time to improve student learning outcomes.