Natural Language Processing Interview Questions and Answers
Artificial Intelligence is one of the most important topics of discussion these days. And, if you are someone aspiring to make a career in Artificial Intelligence, NLP is one thing that is very important to you.
This section covers commonly asked and expert level Natural Language Processing Interview questions and answers. The types of questions covered are general, conceptual and technical in nature. You can also find interesting examples and sample answers with each question.
Who are these Natural Language Processing Interview Questions useful for?
These interview questions will be very useful to all the candidates interviewing for the role of AI expert, AI Intern, NLP Expert, NLP Intern etc.
Both entry level freshers and experienced candidates will be benefited by these questions and answers.
1. What is NLP?
NLP stands for Natural Language Processing. It is a branch of AI that refers to the ability of a computer device to understand the human language, as it is spoken.
The auto-completion of words or queries in Google search, personal assistants Alexa, Siri, Google home are some of the examples of NLP being put to practical use in the real world.
NLP is an evolving field. It is a good idea to go prepared with some concepts that direct this field.
Video : Natural Language Processing Interview Questions and Answers
2. What do you know about Syntactic and Semantic Analysis in NLP?
NLP uses two important techniques called as Syntactic and Semantic Analysis.
Syntactic analysis studies the arrangement of words in a sentence to derive meaning from them based on the grammar rules of a language.
Some of the techniques used for Syntactic analysis are:
Semantics refer to the meaning that a text conveys. Semantic analysis makes use of computer algorithms to understand the meaning and interpretation of words and see how sentences are structured. Some of the techniques used for Semantic analysis are:
i.) Named Entity Recognition - Here you identify and categorize the words into preset categories. E.g. Names of people, places, animals etc.
ii.) Word sense disambiguation - Gives meaning to a word based on the context it is used in.
iii.) Natural language generation - Deriving semantic intentions from the database and converting them into human language.
3. Explain Stemming.
Stemming and Lemmatization are both data pre-processing steps.
However, they both work to establish that the two given words are the different forms of the same word, the approach the two follow is absolutely different.
Stemming - It follows a heuristic approach and works by cutting out the prefix or the suffix to find the stem.
For example - If we take the words Asking and Asked, stemming would cut out the tails "ing" and "ed" and get the stem "Ask" in both the cases.
But this not an optimal method because it can accidentally cut out the wrong letters and the stem won't really make any sense in that case.
So, we say that Stemming may fall a victim to "overstemming" or "understemming".
The three algorithms commonly used for stemming are:
i.) Porter Stemmer
ii.) Snowball Stemmer - also called as Porter2 Stemmer algorithm.
iii.) Lancaster Stemmer - This is the most aggressive algorithm out of the three and can sometimes render absolutely stems that don't make sense.
4. What is Lemmatization?
Lemmatization is a calculated approach to reach the base or the root of any word. Rather than just snipping the words at the head or the tail, this approach resolves the word to its dictionary meaning. This needs more efforts to prepare the system.
Here, the system maps the word to its origin. For e.g. the words "come", "came", "coming" are all mapped to the lemma "come".
Now, it is was a stemming algorithm, it would simply snip off the tail of the word and then won't know that all the three words share the same origin.
5. What are "Stop words"?
The words like "is, am, are, the, of, for, was, were, how, why" that we use to make a sentence are categorized as stop words because we don't want them to make the search engine or the application focus on them rather than the words that really matter.
6. What do you know about Zipf's law in NLP?
Zipf's law was named after the American linguist George Kingsley Zipf.
It states that in a corpus of given natural language utterances, the frequency of a word is inversely proportional to its rank in the frequency table.
This implies that the first most frequent word will appear approximately twice as often as the word that appears at 2nd position.
7. Can you name some commonly used Python NLP libraries?
Some of the commonly used Python NLP Libraries are:
i.) NLTK - This is a well known and complete NLP Library. It has a lot of third-party extensions and approaches for various tasks. Supports the largest number of languages.
ii.) Spacy
iii.) Scikit-learn
iv.) Gensim
v.) Pattern
vi.) Polyglot
