Computing similarity between media programs

1. A computer-implemented method of associating media programs as part of a smart television platform, the method comprising:

obtaining, using at least one processing circuit, first metadata for a first media program comprising a plurality of first fields and second metadata for a second media program comprising a plurality of second fields;

extracting, using at least one processing circuit, a plurality of first tokens from each of the plurality of the first fields and a plurality of second tokens from each of the plurality of second fields;

calculating, for each of the first tokens, a first weight factor and calculating, for each of the second tokens, a second weight factor, wherein the first weight factor and the second weight factor are based on the difference between: the ratio of the logarithm of the number of token occurrences to the logarithm of the sum of one and the total number of possible occurrences, and one;

generating, for each of the first tokens, a first weighted token based on one of the first weight factors;

generating, for each of the second tokens, a second weighted token based on one of the second weight factors;

generating, for each of the first fields, a first vector comprising a plurality of the first weighted tokens;

generating, for each of the second fields, a second vector comprising a plurality of the second weighted tokens.

calculating a plurality of correlations indicative of similarities between the first fields and the second fields, wherein each of the plurality of correlations is a dot product of one of the first vectors and one of the second vectors that are selected based on a field type;

determining a similarity score between the first metadata and the second metadata by combining the plurality of correlations; and

associating the first and second media programs based on the determined similarity score.

2. The method of claim 1 , wherein the plurality of first tokens comprises text included in the plurality of the fields.

3. The method of claim 1 , wherein at least one of the first tokens comprising a combination of text keywords included in the field.

4. The method of claim 1 , further comprises calculating a term frequency for each of the plurality of first tokens equal to the sum of the logarithm of the number of the token's occurrences in a field and 1.

5. The method of claim 1 further comprising identifying a word cluster as a token.

6. The method of claim 1 wherein the plurality of first tokens are extracted from each of the plurality of first fields using a probabilistic topic model.

7. The method of claim 1 further comprising identifying a token having a frequency of occurrence exceeding a predetermined threshold and removing the identified token from the plurality of first tokens.

8. The method of claim 1 , further comprises multiplying a term frequency and an inverse document frequency to generate the first weight factor for each of the first tokens.

9. The method of claim 1 further comprising normalizing each of the first vectors to have a length of 1.

10. A computer-implemented method of associating media programs as part of a smart television platform, the method comprising:

(A) obtaining, using at least one processing circuit, first metadata for a first media program comprising a plurality of first fields and second metadata for a second media program comprising a plurality of second fields;

(B) extracting, using at least one processing circuit, a plurality of first tokens from each of the plurality of the first fields and a plurality of second tokens from each of the plurality of second fields;

(C) calculating a term frequency (TF) for each of the plurality of extracted tokens, and the TF representing a frequency of occurrence of the token within the field;

(D) calculating an inverse document frequency (IDF) for each token and the IDF representing a frequency of occurrence of the token across a plurality of the fields, wherein the IDF is calculated as the difference between: the ratio of the logarithm of the number of token occurrences to the logarithm of the sum of 1 and the total number of possible occurrences; and 1;

(E) combining the TF and the IDF for each token score to generate a weight factor for each token;

(F) determining a similarity score between a first vector comprising the plurality of weighted first tokens and a second vector comprising the plurality of weighted second tokens; and

(G) associating the first and the second media programs responsive to the similarity score.