Korpus is a tf-idf algorithm implementation , simply helps you create a corpus of documents which you can query it to find similiar documents for a given input. So, what is tf-idf?
Wiki definition (http://en.wikipedia.org/wiki/Tf-idf):
Tf–idf, term frequency–inverse document frequency, is a numerical statistic which reflects how important a word is to a document in a collection or corpus.
It is often used as a weighting factor in information retrieval and text mining. The tf-idf value increases proportionally to the number of times a word appears in the document, but is offset by the frequency of the word in the corpus, which helps to control for the fact that some words are generally more common than others.
Basically, korpus is your best friend, while you are trying to approach what the actual input meant to be using your pre-indexed document base. This is the approach what Lucene (the most popular java full text search engine) uses in backyard (http://lucene.apache.org/core/old_versioned_docs/versions/2_9_0/api/all/org/apache/lucene/search/Similarity.html)
Let's take a look the example below. Currently a document defined as a
(id, content) tuple and we are going to create a corpus with a bunch of
idioms using these document tuples. Once the corpus created, under the hood,
our idioms are automatically indexed and weighted which is meant to be ready
for querying:
>>> from korpus import Corpus
>>> common_idioms = [
(1, 'Piece of cake'),
(2, 'Costs an arm and a leg'),
(3, 'Break a leg'),
(4, 'Hit the books'),
(5, 'Let the cat out of the bag'),
(6, 'Hit the nail on the head'),
(7, 'When pigs fly'),
(8, 'You can’t judge a book by its cover'),
(9, 'Bite off more than you can chew '),
(10, 'Scratch someone’s back'),
]
>>> corpus = Corpus(common_idioms)
>>> resutls = corpus.query('Hit the nail', min_score=0.2)
[(6, 0.6134307406647964, 4), (4, 0.2928327297980855, 4)]
We tried to find similiar idioms by our input Hit the nail with a minimum
similarity score of 0.2. The returned list of objects contains the
information about matched items in corresponding corpus. In this case there two
matched items of ids 6 and 4 with similarity scores
0.6134307406647964 and 0.2928327297980855 beside the total match count
of 4
This means there are 4 matched results. Two of them are above the min_score
threshold those are:
* Hit the nail on the head (0.613) * Hit the books (0.292)
Coming soon...
- Mumin Ozturk (@mumino)
- Kadir Pekel (@kadirpekel)