Text Similarity using BM25 & WordNet

Prerequisite for running code

1. Python 2.x - https://askubuntu.com/questions/101591/how-do-i-install-the-latest-python-2-7-x-or-3-x-on-ubuntu
2. Numpy - pip install numpy
3. Scipy - pip install scipy
4. NLTK - pip install nltk
5. Pattern - pip install pattern
6. Sklearn - pip install sklearn

Command for running code

python execute.py

OUTPUT

Algorithms

Syntactic Similarity -

I used BM25(Best Matching) algorithm for syntactic similarity, it generates the similarity score between two sentences.

BM25 Algorithm -

bm25_score(CD,QD) = 𝚺(i=1 to n) idf(qi)*(f(qi,CD)*k1+1)/(f(qi,CD)+k1*(1-b+(b*|CD|/avgdl)))
idf(t) = 1 + log(C/1+df(t))

Where,

    CD = corpus document, e.g.- list of all the answers
    QD = query document, e.g.- list of model answer
    idf(qi) = inverse document frequency (IDF) of the term qi in CD
    C = count of the total number of documents in CD
    df(t) = frequency of the number of documents in which the term t is present
    f(qi, CD) = frequency of the term qi in CD
    |CD| = total length of the CD
    avgdl = average document length of CD
    k1, b = Constants

Semantic Similarity -

I used NLTK's WordNet corpus for generating the semantic similarity score between two sentences. I used synsets function to get all the lexnames of a word then calulated the path similarity between words then took the maximum value among all the lexnames for a single word. After that I calculated the average of all scores for a single sentence and that is the value of semantic similarity score.

Final Score is the average of bm25_score and semantic_similarity_score.