Software:You will need both NLTK and the gensim packages installed on your computer. It should bestraightforward to install gensim using pip or conda.a) Explain in general terms how word embeddings can be said to represent the meaningsof words, and relations such as similarity and analogy between words. Your answershould include brief definitions of the following terms, with appropriate examples:• Syntagmatic association or first-order co-occurrence.• Paradigmatic association or second-order co-occurrence.• The parallelogram model of relational similarity.[20 marks]b) It turns out that the way word embeddings model similarity and analogy can capture avariety of semantic relations between words. Follow the methods used in the Bird tutorialfor the queries below, using the NLTK excerpt from the Google News model:>>> from nltk.data import find>>> word2vec_sample = str(find('models/word2vec_sample/pruned.word2vec.txt'))>>> model = gensim.models.KeyedVectors.load_word2vec_format(word2vec_sample,binary=False)In each case, you should specify the top three words that match the query, and discusswhich of them (if any) come closest to your expected answer.i. Show how gensim solves the following queries:A. Man is to priest as woman is to ____B. They is to their as we is to ___C. Russia is to Moscow as Spain is to ___D. Long is to longest as old is to ___ii. It turns out that embeddings can capture morphosyntactic features such asnumber, tense, and case. Write gensim queries that will return:A. Past tenses of verbs, e.g. come -> came, have -> had, buy -> bought.B. Singular forms of verbs, e.g. come -> comes, have -> has, be -> is.C. Plural forms of nouns, e.g. card -> cards, child -> children.[15 marks]

a) Word embeddings are a type of word representation that allows words with similar meaning to have a similar representation. They are a distributed representation for text that is perhaps one of the key breakthroughs for the impressive performance of deep learning methods on challenging natural language processing problems.

• Syntagmatic association or first-order co-occurrence: This refers to words that often appear together in the same context or sentence. For example, in the sentence "I drink coffee in the morning", "drink" and "coffee" have a syntagmatic association because they appear together.

• Paradigmatic association or second-order co-occurrence: This refers to words that could be substituted for each other in a particular context. For example, in the sentence "I drink coffee in the morning", "coffee" could be replaced with "tea" without changing the grammatical correctness of the sentence. Therefore, "coffee" and "tea" have a paradigmatic association.

• The parallelogram model of relational similarity: This model is used to illustrate how word embeddings can capture both similarity and analogy between words. If we represent words as vectors, then the parallelogram model suggests that analogous words should be equidistant and parallel to each other. For example, the vector from 'man' to 'king' should be parallel and equal in length to the vector from 'woman' to 'queen'.

b) Using the gensim library and the Google News model, we can find the most similar words to a given word or find words that complete analogies. For example:

i. A. Man is to priest as woman is to ____: The gensim model might return 'nun', 'priestess', and 'minister' as the top three words. B. They is to their as we is to ___: The model might return 'our', 'us', and 'we're' as the top three words. C. Russia is to Moscow as Spain is to ___: The model might return 'Madrid', 'Barcelona', and 'Seville' as the top three words. D. Long is to longest as old is to ___: The model might return 'oldest', 'elder', and 'ancient' as the top three words.

ii. A. Past tenses of verbs: The gensim model can be used to find the past tense of a verb by finding the most similar word to the input verb and the word 'past'. For example, 'come' -> 'came', 'have' -> 'had', 'buy' -> 'bought'. B. Singular forms of verbs: Similarly, the model can find the singular form of a verb by finding the most similar word to the input verb and the word 'singular'. For example, 'come' -> 'comes', 'have' -> 'has', 'be' -> 'is'. C. Plural forms of nouns: The model can find the plural form of a noun by finding the most similar word to the input noun and the word 'plural'. For example, 'card' -> 'cards', 'child' -> 'children'.