NCA-GENL Question 63
Select 3You are tasked with building a traditional machine learning pipeline to classify text data into predefined categories. You decide to use Python packages such as spaCy, NumPy, and Keras to implement the solution. Which of the following steps are correct for this task?
- A
Use spaCy to preprocess text data by tokenizing and lemmatizing the text.
- B
Apply NumPy to create dense embeddings of the text data directly.
- C
Use Keras to build and train a neural network classifier on the preprocessed text data.
- D
Leverage spaCy to directly train a neural network model for classification without additional libraries.
- E
Convert the preprocessed text data into numerical features using techniques such as one-hot encoding or TF-IDF.
Show answer and explanation
Correct answers: A, C, E
Explanation
To classify text data using a traditional machine learning pipeline, preprocessing the text using tools like spaCy, converting it into numerical features (e.g., TF-IDF), and training a classifier using a library like Keras are essential steps. Each library in the pipeline has its specific role: spaCy for preprocessing, NumPy for numerical operations, and Keras for model training. Incorrect options involve either misusing libraries or skipping critical steps, such as feature conversion.
- A. Correct.
Correct: spaCy is commonly used for preprocessing text data, such as tokenization, lemmatization, and removing stop words, which are essential steps in text classification tasks.
- B. Incorrect.
Incorrect: NumPy is a powerful library for numerical computations, but it is not used for creating text embeddings directly. Instead, embeddings are typically generated using specialized libraries or models like word2vec or spaCy.
- C. Correct.
Correct: Keras is a deep learning library that can be used to build and train neural network classifiers on preprocessed numerical features derived from text data.
- D. Incorrect.
Incorrect: While spaCy is excellent for text preprocessing, it is not designed to directly train neural network models for classification. Other libraries like Keras or PyTorch are more suited for this purpose.
- E. Correct.
Correct: Converting text data into numerical features is a crucial step in traditional machine learning pipelines. Techniques like one-hot encoding or TF-IDF transform text into a format suitable for classifiers.