Semantic Analysis

import nltk
from nltk.corpus import wordnet as wn
from nltk.wsd import lesk
import spacy

nltk.download(['wordnet', 'punkt', 'stopwords', 'brown', 'semcor'], quiet=True)

# ─── 1. Word Sense Disambiguation (Lesk) ─────────────────────
contexts = [
    ("I went to the bank to deposit my paycheck", "bank"),
    ("The river bank was covered with wildflowers", "bank"),
    ("The plane was flying over the bank of clouds", "bank"),
]
print("Word Sense Disambiguation (Lesk):")
for context, word in contexts:
    tokens = context.split()
    sense = lesk(tokens, word, 'n')
    if sense:
        print(f"  '{context}'")
        print(f"  → Sense: {sense.name()}: {sense.definition()[:70]}
")

# ─── 2. Semantic Similarity with WordNet ─────────────────────
pairs = [
    ("dog", "wolf"),
    ("dog", "cat"),
    ("dog", "table"),
    ("car", "automobile"),
    ("happy", "joyful"),
]
print("WordNet Semantic Similarity:")
print(f"{'Pair':<25} {'Path Sim':>10} {'Wu-Palmer':>12}")
for w1, w2 in pairs:
    s1 = wn.synsets(w1, pos='n')
    s2 = wn.synsets(w2, pos='n')
    if s1 and s2:
        path = s1[0].path_similarity(s2[0]) or 0
        wup = s1[0].wup_similarity(s2[0]) or 0
        print(f"  {w1}-{w2:<20} {path:>10.3f} {wup:>12.3f}")

# ─── 3. Sentence Similarity with Sentence Transformers ───────
# pip install sentence-transformers
from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity
import numpy as np

model = SentenceTransformer('all-MiniLM-L6-v2')

sentences = [
    "The cat sat on the mat.",
    "A feline rested on the rug.",
    "The stock market crashed yesterday.",
    "I love pizza with extra cheese.",
]
embeddings = model.encode(sentences)
sim_matrix = cosine_similarity(embeddings)

print("\nSentence Similarity Matrix:")
for i, s1 in enumerate(sentences):
    for j, s2 in enumerate(sentences):
        if i < j:
            print(f"  {sim_matrix[i,j]:.3f} | '{s1[:40]}' ↔ '{s2[:40]}'")

# ─── 4. Semantic Role Labeling with AllenNLP ─────────────────
# pip install allennlp allennlp-models
from allennlp.predictors.predictor import Predictor

predictor = Predictor.from_path(
    "https://storage.googleapis.com/allennlp-public-models/structured-prediction-srl-bert.2020.12.15.tar.gz"
)

srl_sentence = "Alice carefully gave Bob an interesting book about linguistics."
result = predictor.predict(sentence=srl_sentence)
print(f"\nSRL for: '{srl_sentence}'")
for verb_info in result["verbs"]:
    print(f"  Verb: {verb_info['verb']}")
    print(f"  Tags: {list(zip(result['words'], verb_info['tags']))}")

# ─── 5. Simple SRL with spaCy patterns ───────────────────────
nlp = spacy.load("en_core_web_sm")
doc = nlp("Alice bought a laptop from Apple for one thousand dollars.")
print(f"\nSimple semantic roles from dependency parse:")
for token in doc:
    if token.dep_ == "ROOT":
        print(f"  Predicate (verb): {token.text}")
    elif token.dep_ == "nsubj":
        print(f"  ARG0 (agent/subject): {token.text}")
    elif token.dep_ in ("obj", "dobj"):
        print(f"  ARG1 (theme/object): {token.text}")
    elif token.dep_ == "prep" and token.text == "from":
        print(f"  ARG2 (source): {list(token.children)}")
    elif token.dep_ == "prep" and token.text == "for":
        print(f"  ARGM-MNY (money): {list(token.children)}")