Case context: A company trains a speech recognition model on 8,000 hours of clean recordings made in a quiet office booth. The model performs very well on its held-out office test set. Later, the same model is evaluated in a kiosk installed inside a busy train station, and the error rate rises sharply.

Question: Which environmental factors are the most likely cause of the drop in accuracy, and why do they matter?

Sample answer: The most likely causes are background crowd noise, loudspeaker announcements, and reverberation from the station walls and floor. The model was trained mostly on clean audio, so it learned patterns that do not match the noisy station setting. This data mismatch makes the incoming speech harder to separate from the surrounding sound, which leads to more recognition errors.

Key points:

• Clean office recordings differ from the noisy kiosk environment
• Crowd noise and announcements interfere with speech
• Reverberation changes the audio characteristics the model hears
• The performance drop is explained by training-test mismatch

Rubric: The answer should identify the station noise sources, explain that the model was trained on clean audio, and connect the mismatch between training and deployment conditions to the reduced accuracy.