How to choose similarity measurement between sentences and paragraphs

Question

Problems

1. How to find appropriate measurement method

There are several ways to measure sentence similarities, but I have no idea how to find appropriate method among them for my data (sentences).

Related Question on Stack overflow: is there a way to check similarity between two full sentences in python?

2. Sentence or paragraph based

If it is possible to acquire both one sentence and a paragraph which includes the sentence, which is more accurate to measure the similarity among sentences or paragraphs?

What I tried so far

1. I've tried to use one of the libraries to measure the similarity.

However, I'm struggling how to find more accurate method to measure similarities.

original = 'New York is a noisy city where hamburgers are famous.'
test = ['Berlin is a nostalgic city where sausages are famous.', 'Both New York and Belin are noisy cities, but hamburgers are famous in New York rather than in Berlin.']

import spacy
nlp = spacy.load("en_core_web_sm")


doc1 = nlp(original)
for doc2 in test:
    doc2 = nlp(doc2)
    print(doc1.similarity(doc2)) 

0.8682034221008
0.5078180005337849

Same as sentence based, it was figured out there are several methods to measure the similarity between paragraphs.

But there is no crew which is better (generally high-peformance) to compare sentence or paragraph base.

Related Question on Stack overflow: How to compute the similarity between two text documents?

Answer 1

Simphile NLP package creator here. Choosing the text similarity method that works best for your application can be difficult. Ideally you have a large sample set with many known positives (i.e. text that are sufficiently related to the reference). You would choose the method that orders the sample set such that the positives are concentrated at the head; this could be measured with the area under the precision-recall curve.

The Simphile package makes it easy to try several different methods:

Install:

pip install simphile

Choose your favorite method. This example shows three:

from simphile import jaccard_similarity, euclidian_similarity, compression_similarity

text_a = "I love dogs"
text_b = "I love cats"

print(f"Jaccard Similarity: {jaccard_similarity(text_a, text_b)}")
print(f"Euclidian Similarity: {euclidian_similarity(text_a, text_b)}")
print(f"Compression Similarity: {compression_similarity(text_a, text_b)}")

• Compression Similairty – leverages the pattern recognition of compression algorithms
• Euclidian Similarity – Treats text like points in multi-dimensional space and calculates their closeness
• Jaccard Similairy – Texts are more similar the more their words overlap

More docs and examples at the repo