SPEECH RECOGNITION

What is the primary goal of speech recognition?

• a) Convert speech into text

• b) Convert text into speech

• c) Translate languages

• d) Generate synthetic voices

Which deep learning architecture is commonly used for speech recognition tasks?

• a) CNN (Convolutional Neural Network)

• b) RNN (Recurrent Neural Network)

• c) GAN (Generative Adversarial Network)

• d) Transformer

3. What type of neural network is most effective for sequential speech data processing?

• a) Fully Connected Networks

• b) Convolutional Neural Networks (CNN)

• c) Recurrent Neural Networks (RNN)
  

• d) Autoencoders

4. Which of the following is a commonly used dataset for speech recognition?

• a) MNIST

• b) ImageNet

• c) LibriSpeech
  

• d) CIFAR-10

5.Which of the following challenges affect speech recognition models?

• a) Accents and dialects

• b) Background noise

• c) Homophones

• d) All of the above

6.What is the role of a Language Model (LM) in speech recognition?

• a) To convert speech into spectrograms

• b) To improve the accuracy of recognized words by predicting sequences

• c) To generate synthetic speech

• d) To reduce background noise

7.Why do homophones pose a challenge in speech recognition?

• a) They are difficult to pronounce

• b) They require additional computational power

• c) They sound the same but have different meanings, causing ambiguity

• d) They increase speech processing speed

8.Why are Convolutional Neural Networks (CNNs) used in speech recognition?

• a) To process text-based language models

• b) To extract spatial features from spectrograms or mel-spectrograms

• c) To convert text into speech

• d) To directly translate speech into another language

9. What is the main advantage of end-to-end speech recognition systems?

• a) They require separate feature extraction steps

• b) They directly convert raw audio to text without intermediate steps

• c) They only work with small datasets

• d) They are slower than traditional models

10.Why are language models used in speech recognition?

• a) To reduce training time

• b) To predict word probabilities and improve recognition accuracy

• c) To generate synthetic audio data

• d) To extract features from audio signals