Using the model as an information retrieval tool

Establishing a search engine involves many significant engineering tasks, so in this section, we’ll go through the essential steps to deploy an LSI model to be part of a search engine. These steps are as follows:

1. Load the saved objects.
2. Preprocess the new document.
3. Score the new document to get the latent topic scores.
4. Calculate the similarity scores with the new document.
5. Find documents with high similarity scores.

First, we will load the four saved objects. The objects include the dictionary list, the model, the BoW object, and the TF-IDF object.

Loading the dictionary list

Gensim has a utility function called datapath. It points to the physical location of the file. Here is the code for it:

from gensim.corpora import Dictionaryfrom gensim.test.utils import datapath
dict_file = datapath(path + “/gensim_dictionary_AGnews”)
gensim_dictionary = Dictionary.load(dict_file)

...