如何利用WordNet查找派生相关词及实现派生词到原词转换

Great questions! Let's break this down step by step using WordNet—whether you're hunting for all derivationally linked words or reversing derivations like founder → found, here's how to make it work.

WordNet stores derivational relationships between lemmas (word forms tied to specific meanings), so we can tap into that directly. Here's a straightforward approach using NLTK's WordNet interface:

• First, grab all synsets (sense groups) for your target word—since a word can have multiple meanings, we need to check every possible sense.
• For each lemma in each synset, use the derivationally_related_forms() method to pull linked words.
• Collect all these words, remove duplicates, and sort them for readability.

Here's a ready-to-use Python function:

from nltk.corpus import wordnet

def get_derivational_related_words(word):
    related_words = set()
    # Fetch all synsets for the input word
    synsets = wordnet.synsets(word)
    for syn in synsets:
        for lemma in syn.lemmas():
            # Get all derivationally linked lemmas
            deriv_lemmas = lemma.derivationally_related_forms()
            for deriv_lemma in deriv_lemmas:
                related_words.add(deriv_lemma.name())
    # Return sorted unique results
    return sorted(related_words)

# Test it out
print(get_derivational_related_words("found"))
# Output might include: ['founder', 'foundation', 'founding', 'refound']

Note: This returns all derivationally connected words across all parts of speech. If you want to filter by POS (e.g., only nouns), add a check for syn.pos() before processing.

For reversing derivations (like turning agent nouns to their base verbs), we use the same derivationally_related_forms() method, but add a filter to target the part of speech we expect the base word to be (usually a verb for examples like owner → own).

Here's how to do it:

• Fetch synsets and lemmas for the derived word.
• For each related lemma, check if its synset matches your target POS (e.g., verb for agent nouns).
• Collect and return the matching base words.

Example function:

from nltk.corpus import wordnet

def get_base_form(derived_word, target_pos=wordnet.VERB):
    base_words = set()
    synsets = wordnet.synsets(derived_word)
    for syn in synsets:
        for lemma in syn.lemmas():
            deriv_lemmas = lemma.derivationally_related_forms()
            for deriv_lemma in deriv_lemmas:
                # Only keep lemmas of the target part of speech
                if deriv_lemma.synset().pos() == target_pos:
                    base_words.add(deriv_lemma.name())
    # Return sorted results if any, else None
    return sorted(base_words) if base_words else None

# Test with your examples
print(get_base_form("founder"))  # Output: ['found']
print(get_base_form("owner"))    # Output: ['own']
print(get_base_form("writer"))   # Output: ['write']

If you're dealing with a different type of derivation (e.g., an adjective from a noun), just adjust the target_pos parameter—use wordnet.NOUN for nouns, wordnet.ADJ for adjectives, or wordnet.ADV for adverbs.

内容的提问来源于stack exchange，提问作者Ajitesh Mandal