Flalingo FLAI: Real Performance Analysis in English Learning with AI

FLAI is an AI-based language analysis system that extracts 47 pedagogical metrics from every lesson, measuring talk ratio, vocabulary diversity, and level alignment.

"How is a good language lesson measured?" Most people answer this question with phrases like, "based on how happy the student is" or "based on the teacher's experience." However, modern educational science shows us that real learning success can be assessed much more accurately with concrete, measurable metrics.

Critical questions in language learning include: Should the teacher or the student talk more? Which words should be taught? What should the student's speaking speed be? How much pausing is normal during the lesson? The answers to these questions must now be based on data, not guesswork.

FLAI (Foreign Language AI Analyzer) extracts 47 different pedagogical metrics from every lesson. These metrics not only answer the question of "what happened," but also provide objective answers to "how well did it happen" and "what can be done." In this blog, we will examine in detail the core metrics measured by FLAI, their pedagogical importance, and their contribution to the learning process.

1. Talk Ratio Analysis: Whose Voice is Heard in the Lesson? 

Talk Ratio

One of the most critical metrics in language learning is the student-teacher speaking balance. Research shows that for a student to progress in the language, they must actively engage in "production." While passive listening is important, actively using the language is essential for true learning.

FLAI measures the following in every lesson:

• Student's total speaking time (seconds and percentage)
• Teacher's total speaking time (seconds and percentage)
• The balance ratio between the two parties

Pedagogical Ideal: In language learning literature, it is recommended that the student actively speaks for at least 60% of the lesson time. The teacher is the guide, the corrector, and the model, but the student should be the main speaker of the lesson.

FLAI's Approach: The system calculates the talk ratio of every lesson and automatically detects the following statuses:

• ✅ Ideal Balance (Student 55–70%, Teacher 30–45%): The lesson is pedagogically healthy.
• ⚠️ Teacher Dominant (Teacher 70%+): The student is not given sufficient speaking opportunity.
• 🚨 Critical Status (Teacher 95%+): FLAI automatically tags this lesson with an "error" status and sends a notification to the platform administrator.

Real Example: When a lesson is analyzed:

This result shows that the student spoke only 38%. FLAI provides the following suggestion to the teacher: "Your student did not have enough opportunity to speak in this lesson. Encourage the student to speak more by asking more open-ended questions in the next lesson."

Why Is It Important? Many teachers leave happy, thinking, "I taught my student a lot," when in fact the student may have had no opportunity to practice. FLAI objectively reveals this situation and raises awareness.

2. Speaking Speed and Fluency: Tempo and Rhythm ⏱️

Talking Speed

Speaking speed measures how many words per minute (WPM) are spoken. This metric is an important indicator of the student's command of the language.

Action for the Teacher: FLAI monitors the student's speaking speed over time. If the student consistently stays below 60 WPM, it suggests to the teacher: "Focus on developing your student's vocabulary." If the speed is increasing over time (e.g., from 70 WPM to 110 WPM in 3 months), this is a clear indicator of progress and can be used to motivate the student.

Pause Times

Pauses during speech provide information about the student's thinking processes, word search efforts, or confidence level.

FLAI Metrics:

• Total Pause Time: Total duration of silence
• Average Pause Duration: Average length of a pause
• Pause in PHW (Pauses per Hundred Words): How many times pausing occurs per 100 words

Pedagogical Meaning:

• Short Pauses (1-2 seconds): Natural, thinking time
• Medium Pauses (3-5 seconds): Word search, slight hesitation
• Long Pauses (5+ seconds): Serious vocabulary gap or confidence issue

If the student pauses more than 10 times per 100 words, this signals to the teacher that:

• The student is being pushed out of their comfort zone (this can be good, struggle = learning)
• If they are struggling too much, the lesson level should be lowered slightly

Muted Time

In online lessons, students sometimes mute their microphones. FLAI also monitors these durations.

Analysis:

• Long muted time → The student may be passive or uncomfortable in the lesson.
• Frequent muting/unmuting → Technical issue or student reluctance to participate.

3. Vocabulary Analysis: Metrics at the Heart of Learning 

Words are the foundation of language learning. FLAI analyzes vocabulary from multiple perspectives.

Total Words

The total number of words the student used throughout the lesson.

This is the simplest indicator of how active the student was in the lesson. Ideally, a student should use at least 400–600 words in a 60-minute lesson.

Unique Words

This metric measures the number of different words used by removing repetitions from the "Total Words" count. For example, in the sentence "I think... I think it is very good," although there are 7 total words, the unique word count is 5 because "I" and "think" are repeated. This metric shows the student's vocabulary diversity and is generally assessed with the ratio TTR = Unique / Total.

TTR may naturally decrease as the text length increases; therefore, results should be interpreted by considering the student's level and the context of the speech.

Pedagogical Value: An increase in TTR over time shows that the student's vocabulary is expanding. FLAI tracks this metric weekly and monthly to provide progress reports to the teacher and student.

New Words

One of FLAI's strongest features is the automatic detection of new words using GPT-4. The system analyzes the dialogue and detects new words that the teacher consciously taught.

Dictionary Integration: For every new word, FLAI automatically:

• Pulls the definition from Cambridge or Oxford dictionaries
• Adds example sentences
• Provides pronunciation information (IPA)

FLAI offers not only the definition but also synonym suggestions for the new words learned by the student. For example, for the word "adventure," the system suggests "journey," "expedition," and "quest" alternatives; for "curious," it suggests "interested," "inquisitive," and "eager."

This approach enriches the student's vocabulary not only numerically but also in terms of semantic diversity. Thus, the student learns to express the same concept in different contexts, and their speaking fluency is strengthened.

In this way, the student has the opportunity to learn not only the new word but also its alternatives.

Pedagogical Benefit:

• The teacher can say, "I taught 8 new words in this lesson."
• The student can review the words they learned afterward.
• Which words were taught is tracked when creating the lesson curriculum.

Most Common Words

A list of the words the student used the most.

Analysis:

• If filler words like "I think," "like," "maybe" are used too frequently → The student is hesitant or lacks vocabulary.
• If specific, meaningful words are frequent (e.g., "fascinating," "demonstrate," "analyze") → The student is at an advanced level.

Action Suggestion: If the student repeats the same words too often, FLAI suggests to the teacher: "Teach alternatives for these words."

Word Difficulty Analysis

FLAI categorizes every word according to CEFR (Common European Framework of Reference) standards: A1, A2, B1, B2, C1, C2.

Pedagogical Critical Meaning: This metric measures the suitability of the lesson to the student's level.

Scenario 1: Level Mismatch In a lesson conducted for an A2 level student:

This distribution is highly problematic! The teacher is using words far above the student's level. Result:

• The student does not understand.
• Motivation drops.
• Learning does not occur.

FLAI detects this situation and sends a warning to the teacher: "55% of the words used in this lesson are above the student's level. Use simpler vocabulary."

Scenario 2: Content That is Too Easy For a B2 student:

This is also a problem! The student gets bored, and does not progress. FLAI suggestion: "Your student is ready for more challenging content. Add B2-C1 level vocabulary."

Rare Word Percentage

FLAI uses English word frequency lists such as Oxford 5000 and COCA to filter out "rare" words in the text; this reveals the true difficulty level of the text. Since the rare word percentage in your data is 12.5%, this falls within the 10–20% range, indicating a "normal, balanced" profile (for comparison: below 5% = very basic speech; 30%+ = very academic or unnecessarily complex).

This ratio offers a meaningful challenge for B1–B2 level students while maintaining fluency. Rare words are valuable for advanced (B2–C1) students in building vocabulary and nuance; however, they can create cognitive load at the beginner (A1–A2) level.

Therefore, in A1–B1 groups, it is ideal to prefer more common synonyms for the same concept, add short dictionary notes/hints for rare words, and support them with in-context examples at the first encounter; in B2–C1, this ratio can be maintained and selectively increased with targeted academic/professional terms.

4. Language Detection and Analysis: Immersion Quality 

Language Detection

FLAI automatically detects all languages used in the lesson and analyzes student and teacher speeches separately. In this way, it measures how frequently each language is used and objectively reports the student's level of exposure to the target language (English).

In this example, the student spoke a total of 456 words, and 76 of them were Turkish, meaning approximately 16.6% use of the native language. This ratio is typical for occasional native language use in English lessons. Especially for beginner and intermediate students (between 10–20%), this ratio is natural; students turn to their native language when explaining complex thoughts or seeking comprehension confirmation.

However, the goal is to reduce this ratio over time. Ideally, the Turkish usage rate should drop below 10% from the B1 level, and below 5% at B2 and above. FLAI regularly monitors this data to numerically show how the student's ability to think and express themselves in the target language is developing. This helps the student increase their self-awareness; the teacher can optimize their lesson strategies based on this data.

Why Is It Important? "Full immersion" in language learning is critical. The less a student uses their native language, the more they begin to think in the target language.

If native language use is high, FLAI suggests to the teacher: "Your student frequently switches to Turkish. You can reduce native language use by explaining unknown words in English."

5. Safety and Content Moderation 

For Children's Lessons (Moderation API)

FLAI uses an advanced content moderation system to ensure safety in children's lessons. The system analyzes all conversations during the lesson and automatically detects potential risk elements. These elements include inappropriate content, violent language, hate speech, and sexual content, which could negatively impact the student's development.

At the code level, this control mechanism is explicitly defined in the relevant section of the crontab.py file. If the lesson is marked as a children's lesson, the system runs the <MODERATION_FUNCTION> function. This function checks whether inappropriate content is detected in the dialogue. If the conversation is marked as "flagged," the system immediately triggers a series of automatic actions.

First, an automatic notification is sent to the platform administrator via Chat. This allows administrators to be instantly aware of potentially risky situations. Then, the lesson recording is taken for manual review, and the context of the content is evaluated by human moderators. If necessary, the relevant teacher is contacted, an explanation is requested, or corrective measures are taken.

This system demonstrates that FLAI prioritizes child safety and places great importance not only on teaching quality but also on the responsibility of creating an ethical and safe learning environment. Thus, parents can be assured that their children are learning English in a completely safe and monitored environment.

6. Real Scenario Examples: From Data to Action 

Scenario 1: Imbalanced Talk Lesson

The FLAI report detailed the communication balance and the student's active speaking performance in the lesson conducted by teacher Melis with her B1 level student, Ahmet. The data obtained from the first 60-minute lesson showed that the student's speaking opportunities during the lesson were limited.

When the total talk ratio was examined, the student spoke only 15%, while the teacher spoke 85%. This imbalance indicates that student-centered speaking practice was not sufficiently provided.

Additionally, the student used a total of 89 words, which is well below the target range of 400–600 words. The speaking speed was also determined to be 65 words/minute (WPM); this value is below the target range of 90–120 WPM.

In line with the FLAI system's suggestion, teacher Melis was advised to give her student more speaking time, ask open-ended questions, ask the student to recount a story or personal experience, and keep her own sentences shorter. These suggestions aimed to encourage the student to speak more naturally and confidently.

In the next lesson, teacher Melis applied these suggestions, and the results changed remarkably. According to the new FLAI report, the student's talk ratio increased from 15% to 58%, and the teacher's share dropped to 42%.

Furthermore, the total number of words used by the student increased from 89 to 412, reaching the target range. This improvement indicates significant progress in both the student's word productivity and participation level in the lesson.

Overall, this result clearly demonstrates the impact of FLAI's data-driven teacher feedback. The strategic changes implemented by teacher Melis strengthened the student's active learning process; thus, the lesson shifted from a teacher-centered structure to a student-centered, interactive one.

Scenario 2: Level Mismatch

The FLAI report analyzed the language suitability of the lesson conducted between A2 level student Zeynep and her teacher Can. The report detailed the difficulty level of the words used throughout the lesson and detected a significant imbalance.

According to the data, 65% of the words used in the lesson content are at the B2 and above level, which is significantly higher than the student's current A2 level. In addition, the rare word percentage was calculated as 35%; this rate is considered "high" according to FLAI standards.

Consequently, it is an expected situation for Zeynep to experience frequent loss of comprehension and low interaction during the lesson.

In light of this data, the FLAI system advised teacher Can to choose words appropriate to the student's comprehension level. The report suggested preferring simpler and more everyday expressions instead of complex and abstract words.

For example, alternatives such as "nice" or "good" instead of "sophisticated," "think about" instead of "contemplate," and "certain to happen" instead of "inevitable" were offered. This approach facilitates the student's grasp of the meaning and the learning of words within a natural context.

In subsequent lessons, teacher Can began implementing these suggestions. Teaching with simpler and level-appropriate vocabulary increased Zeynep's comprehension-based participation.

The student can now follow the lesson more comfortably, struggles less in parts she doesn't understand, and remains active throughout the lesson. This change positively affected not only the learning experience but also the student's satisfaction and desire for retention.

This example demonstrates the power of AI-supported analysis by the FLAI system, which directly contributes to teaching quality. Teachers managing level alignment correctly ensures that students maintain their motivation and achieve sustainable learning success.

Scenario 3: Progress Tracking - Mehmet's 12-Week Journey

FLAI tracked Mehmet's 12-week journey lesson by lesson, clearly revealing his progress. The number of unique words the student used increased by 180% from 150 to 420 between Week 1 and Week 12.

This increase shows that not only the number of memorized words but also Mehmet's flexibility in selecting and using words in different contexts has strengthened. That is, his expressive diversity and richness of narrative have also noticeably improved as his vocabulary expanded.

There is also a significant leap in fluency: Speaking speed increased by approximately 64%, from 70 WPM to 115 WPM. This indicates that the confidence in sentence construction and automaticity have strengthened, and pauses have decreased.

Since the 90–120 WPM band is generally considered healthy for comfortable communication, Mehmet's final value approaching the upper limit of this range presents a picture consistent with the fluency goal.

Interaction balance also improved: The student's talk ratio increased from 45% to 65% — a +20 point (approx. 44% relative increase). This indicates that the floor time in the lesson shifted towards the student and that productive speaking opportunities were effectively utilized.

In short, in 12 weeks, Mehmet is using more and more diverse words, speaking more fluently, and taking up more speaking time in the lesson; this triple progress forms a strong foundation for lasting language acquisition.

Conclusion: From Data to Learning 

The metrics measured by FLAI are not just numbers. Each one represents a critical point in the student's learning journey. The talk ratio shows how active the student is; the vocabulary shows how much they have progressed; and the speaking speed shows how comfortable they are.

In traditional systems, this information is lost, remains speculative, or is never measured. With FLAI, every lesson becomes a data goldmine, and this data:

• Gives teachers the opportunity to improve their teaching methods.
• Provides students with clear goals and motivation.
• Enables platform administrators to perform quality control and facilitate growth.

The combination of Data + human teacher creates the perfect learning experience. FLAI is the technology that makes this equation possible.