Semantic similarity computation with different state-of-the-art metrics

Description • Installation • Usage • License

Description

TaxoSS is a semantic similarity library for Python which implements the state-of-the-art semantic similarity metrics like Resnik, JCN, and HSS.

Requirements

• Python 3.6 or later
• NLTK
• NumPy
• Pandas

Installation

TaxoSS can be installed through pip (the Python package manager) in the following way:

pip install taxoss

Usage

Semantic similarity functions

You can compute the semantic similarity in the following way:

from TaxoSS.functions import semantic_similarity
semantic_similarity('brother', 'sister', 'hss')

3.353513521371089

The function semantic_similarity(word1, word2, kind, ic) has these options for the argument kind:

• hss -> HSS (default)
• wup -> WUP
• lcs -> LC
• path_sim -> Shortest Path
• resnik -> Resnik
• jcn -> Jiang-Conrath
• lin -> Lin
• seco -> Seco

For the argument ic see the following section.

Information Content

Using a Wikipedia copus for calculating the Information Content (default of the argument ic):

from TaxoSS.functions import semantic_similarity
semantic_similarity('cat', 'dog', 'resnik')

6.169410755220327

Calculating Information Conent from a given corpus:

from TaxoSS.calculate_IC import calculate_IC
from TaxoSS.functions import semantic_similarity

calculate_IC(path_to_corpus, path_to_save_IC_file)
semantic_similarity('cat', 'dog', 'resnik', path_to_save_IC_file)

with path_to_save_IC_file a path into the virtual environment TaxoSS package, e.g. venv/lib/python3.6/site-packages/TaxoSS/data/prova_IC.csv.

Benchmark