System and method of using a knowledge representation for features in a machine learning classifier
Systems and methods are provided for classifying at least one unlabeled content item with a machine-learning classifier. A knowledge representation synthesized based on an object of interest is used as a source of features for which training data is evaluated. The machine learning classifier is trained based on features based on the attributes and the training data.
1 . A computer-implemented method of transforming a set of raw data into a set of featurized data using an object of interest comprising a set of unstructured data for training machine learning models, the method comprising executing via a computer system comprising at least one processor and at least one non-transitory memory:
receiving, by the at least one processor, the set of raw data for use with the machine learning models, the set of raw data comprising content items;
receiving, by the at least one processor, the object of interest containing terms and/or properties;
synthesizing, by the at least one processor executing executable instructions stored in the at least one non-transitory memory, a knowledge representation based on terms and/or properties of the object of interest, the knowledge representation being encoded as a non-transitory computer-readable data structure, the synthesizing comprising:
generating an initial knowledge representation using the object of interest as input, the initial knowledge representation containing a plurality of concepts and at least one relationship between the plurality of concepts;
executing a semantic synthesis process on the initial knowledge representation to obtain an augmented initial knowledge representation, the semantic synthesis process including (i) augmenting the initial knowledge representation using at least one information source external to the object of interest, (ii) deriving at least a first concept or a first relationship between two or more concepts that is not present in the initial knowledge representation based on a semantic relationship between the plurality of concepts in the initial knowledge representation and the at least a first concept or a first relationship, and (iii) adding the at least a first concept or a first relationship to the initial knowledge representation;
deriving one or more attributes from the augmented initial knowledge representation, wherein the one or more attributes are properties of the augmented initial knowledge representation;
deriving a first set of one or more content item features from the derived one or more attributes for use in obtaining featurized content items with derived features;
determining evaluation values of the derived first set of one or more content item features for each of the content items of the set of raw data by evaluating the content items of the set of raw data with respect to the derived first set of one or more content item features;
associating the evaluation values of the derived first set of one or more content item features with each of the content items of the set of raw data to obtain the set of featurized data; and
training the machine learning models using the set of featurized data as training data.
2 . The method of claim 1 , further comprising executing the machine learning models using the set of featurized data as input data.
3 . The method of claim 2 , wherein the input data is training labeled data.
4 . The method of claim 2 , wherein the input data is evaluation data.
5 . The method of claim 1 , wherein the machine learning models use supervised learning.
6 . The method of claim 1 , wherein the object of interest is one of a topic, a tweet, a webpage, a website, a document, a document title, a message, an advertisement, and a search query.
7 . The method of claim 1 , wherein the semantic synthesis process comprises generative rules applied to an elemental data structure to generate a new complex knowledge representation.
8 . The method of claim 1 , wherein the one or more content item features include one of:
a total number of concepts intersecting between the augmented initial knowledge representation and at least one content item;
a number of broader concepts intersecting between the augmented initial knowledge representation and at least one content item;
a number of narrower concepts intersecting between the augmented initial knowledge representation and at least one content item; and
a sum of weights of concepts intersecting between the augmented initial knowledge representation and at least one content item.
9 . A system for transforming a set of raw data into a set of featurized data using an object of interest comprising a set of unstructured data for training machine learning models, the system comprising:
at least one non-transitory memory, and at least one processor configured to perform a method comprising:
receiving the set of raw data for use with the machine learning models, the set of raw data comprising content items;
receiving the object of interest containing terms and/or properties;
synthesizing, by the at least one processor executing executable instructions stored in the at least one non-transitory memory, a knowledge representation based on terms and/or properties of the object of interest, the knowledge representation being encoded as a non-transitory computer-readable data structure, the synthesizing comprising:
generating an initial knowledge representation using the object of interest as input, the initial knowledge representation containing a plurality of concepts and at least one relationship between the plurality of concepts;
executing a semantic synthesis process on the initial knowledge representation to obtain an augmented initial knowledge representation, the semantic synthesis process including (i) augmenting the initial knowledge representation using at least one information source external to the object of interest, (ii) deriving at least a first concept or a first relationship between two or more concepts that is not present in the initial knowledge representation based on a semantic relationship between the plurality of concepts in the initial knowledge representation and the at least a first concept or a first relationship, and (iii) adding the at least a first concept or a first relationship to the initial knowledge representation;
deriving one or more attributes from the augmented initial knowledge representation, wherein the one or more attributes are properties of the augmented initial knowledge representation;
deriving a first set of one or more content item features from the derived one or more attributes for use in obtaining featurized content items with derived features;
determining evaluation values of the derived first set of one or more content item features for each of the content items of the set of raw data by evaluating the content items of the set of raw data with respect to the derived first set of one or more content item features;
associating the evaluation values of the derived first set of one or more content item features with each of the content items of the set of raw data to obtain the set of featurized data; and
training the machine learning models using the set of featurized data as training data.
10 . The system of claim 9 , further comprising executing the machine learning models using the set of featurized data as input data.
11 . The system of claim 10 , wherein the input data is training labeled data.
12 . The system of claim 10 , wherein the input data is evaluation data.
13 . The system of claim 9 , wherein the machine learning models use supervised learning.
14 . The system of claim 9 , wherein the object of interest is one of a topic, a tweet, a webpage, a website, a document, a document title, a message, an advertisement, and a search query.
15 . The system of claim 9 , wherein the semantic synthesis process comprises generative rules applied to an elemental data structure to generate a new complex knowledge representation.
16 . The system of claim 9 , wherein the one or more content item features include one of:
a total number of concepts intersecting between the augmented initial knowledge representation and at least one content item;
a number of broader concepts intersecting between the augmented initial knowledge representation and at least one content item;
a number of narrower concepts intersecting between the augmented initial knowledge representation and at least one content item; and
a sum of weights of concepts intersecting between the augmented initial knowledge representation and at least one content item.
17 . At least one non-transitory computer readable storage medium storing processor-executable instructions that, when executed by at least one processor, cause the at least one processor to perform a method of transforming a set of raw data into a set of featurized data using an object of interest comprising a set of unstructured data for training machine learning models, the method comprising:
receiving the set of raw data for use with the machine learning models, the set of raw data comprising content items;
receiving the object of interest containing terms and/or properties;
synthesizing a knowledge representation based on terms and/or properties of the object of interest, the knowledge representation being encoded as a non-transitory computer-readable data structure, the synthesizing comprising:
generating an initial knowledge representation using the object of interest as input, the initial knowledge representation containing a plurality of concepts and at least one relationship between the plurality of concepts;
executing a semantic synthesis process on the initial knowledge representation to obtain an augmented initial knowledge representation, the semantic synthesis process including (i) augmenting the initial knowledge representation using at least one information source external to the object of interest, (ii) deriving at least a first concept or a first relationship between two or more concepts that is not present in the initial knowledge representation based on a semantic relationship between the plurality of concepts in the initial knowledge representation and the at least a first concept or a first relationship, and (iii) adding the at least a first concept or a first relationship to the initial knowledge representation;
deriving one or more attributes from the augmented initial knowledge representation, wherein the one or more attributes are properties of the augmented initial knowledge representation;
deriving a first set of one or more content item features from the derived one or more attributes for use in obtaining featurized content items with derived features;
determining evaluation values of the derived first set of one or more content item features for each of the content items of the set of raw data by evaluating the content items of the set of raw data with respect to the derived first set of one or more content item features;
associating the evaluation values of the derived first set of one or more content item features with each of the content items of the set of raw data to obtain the set of featurized data; and
training the machine learning models using the set of featurized data as training data.
18 . The computer readable storage medium of claim 17 , further comprising executing the machine learning models using the set of featurized data as input data.