Mongoose MoreLikeThis

Mongoose MLT is a mongoose plugin that adds a more-like-this query generation algorithm for searching relevant content.

It is heavily based on Lucene's MoreLikeThis, which uses tf-idf statistics to characterize document relevance based a small set of terms.

Usage

Define your schema, define your text indexes, and then load the plugin:

var MySchema = new mongoose.Schema({ ... });
MySchema.index({ '$**': 'text' });
MySchema.plugin(require('mongoose-mlt'), {
  limit: 100,
  tfThreshold: 2,
  termLimit: 25
});

The following are the possible options you can pass to the plugin:

• limit: Same as mongoose.models.Model.find().limit(n). It limits the number of returned results. Default is 100.
• tfThreshold: Determines the minimium term-frequency for any given term to be considered for search. Use a higher number to provide faster matching when using large document bodies. Use a lower number to provide more accurate matching when using small document bodies. Default is 2.
• termLimit: Limits the maximum number of terms to be used for search. It uses the n highest scoring terms. Use a higher number to provide more but less relevant results. Use a lower number to provide fewer but more relevant results. Default is 25.

There are several ways to use the mlt functionality. They all require a seed document or ObjectId:

Static method with a seed JSON object:

var jsonSeed = { title: 'Seed', body: 'This is the seed document' };
MyModel.mlt(jsonSeed, callback);

Static method with a seed Mongoose model:

var modelSeed = new MyModel({ title: 'Seed', body: 'This is the seed document' });
MyModel.mlt(modelSeed, callback);

Static method with a seed _id

MyModel.find({ title: 'Seed', body: 'This is the seed document' }, function(err, modelSeed) {
  MyModel.mlt(modelSeed._id, callback);
});

The mlt method itself takes the following parameters:

MyModel.mlt(seed, conditions, fields, options, callback);

where seed is the seed document or _id, and conditions, fields, options, and callback are the same as you would pass to Model.find().

The callback for mlt has err (which is null on success) and results (which is an array of relevant documents, sorted by similarity).