Inference based on different portions of a training set using a single inference model
Systems, methods and non-transitory computer readable media for inference based on different portions of a training set using a single inference model are provided. Textual inputs may be received, each of which may include a source-identifying-keyword. An inference model may be a result of training a machine learning model using a plurality of training examples. Each training example may include a respective textual content and a respective media content. The training examples may be grouped based on source-identifying-keywords included in the textual contents. Different parameters of the inference model may be based on different groups, and thereby be associated with different source-identifying-keywords. When generating new media content using the inference model and a textual input, parameters associated with the source-identifying-keyword included in the textual input may be used.
1 . A non-transitory computer readable medium storing a software program comprising data and computer implementable instructions that when executed by at least one processor cause the at least one processor to perform operations for inference based on different portions of a training set using a single inference model, the operations comprising:
receiving a first textual input in a natural language, the first textual input includes a first source-identifying-keyword;
receiving a second textual input in the natural language, the second textual input includes a second source-identifying-keyword, the second source-identifying-keyword differs from the first source-identifying-keyword;
accessing an inference model, the inference model is a result of training a machine learning model using a plurality of training examples, each training example of the plurality of training examples includes a respective textual content in the natural language and a respective media content, the plurality of training examples includes a first group of at least two training examples associated with a first source and a second group of at least two training examples associated with a second source, the first group differs from the second group, the first source differs from the second source, each textual content of the textual contents associated with the first group includes the first source-identifying-keyword, each textual content of the textual contents associated with the second group includes the second source-identifying-keyword, a first parameter of the inference model is based on the first group, a second parameter of the inference model is based on the second group, the first parameter differs from the second parameter;
using the inference model and the first textual input to generate a first new media content, the first new media content is based on the first parameter of the inference model; and
using the inference model and the second textual input to generate a second new media content, the second new media content is based on the second parameter of the inference model.
2 . The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:
in response to the first textual input including the first source-identifying-keyword, updating a first data-record associated with the first source; and
in response to the first textual input including the second source-identifying-keyword, updating a second data-record associated with the second source.
3 . The non-transitory computer readable medium of claim 1 , wherein the first new media content includes a first new visual content, the second new media content includes a second new visual content, and wherein for each training example of the plurality of training examples, the respective media content includes a respective visual content.
4 . The non-transitory computer readable medium of claim 1 , wherein the first new media content includes a first new audio content, the second new media content includes a second new audio content, and wherein for each training example of the plurality of training examples, the respective media content includes a respective audio content.
5 . The non-transitory computer readable medium of claim 1 , wherein the first group and the second group have no training example in common.
6 . The non-transitory computer readable medium of claim 1 , wherein a particular training example is associated with both the first source and the second source, wherein the textual content included in the particular training example includes both the first source-identifying-keyword and the second source-identifying-keyword, and wherein both the first group and the second group includes the particular training example.
7 . The non-transitory computer readable medium of claim 1 , wherein the first parameter of the inference model includes a first artificial neuron, and the second parameter of the inference model includes a second artificial neuron.
8 . The non-transitory computer readable medium of claim 1 , wherein the first parameter of the inference model includes a first coefficient, and the second parameter of the inference model includes a second coefficient.
9 . The non-transitory computer readable medium of claim 1 , wherein the first parameter of the inference model includes a first threshold, and the second parameter of the inference model includes a second threshold.
10 . The non-transitory computer readable medium of claim 1 , wherein each one of the first textual input and the second textual input is indicative of a respective desire of an individual to generate a respective visual content depicting a person, the first new media content includes a depiction of the person with a characteristic of people depicted in media contents associated with the first group, and the second new media content includes a depiction of the person with a characteristic of people depicted in media contents associated with the second group.
11 . The non-transitory computer readable medium of claim 1 , wherein each one of the first textual input and the second textual input is indicative of a respective desire of an individual to generate a respective visual content depicting an event, the first new media content includes a depiction of the event with a characteristic of events depicted in media contents associated with the first group, and the second new media content includes a depiction of the event with a characteristic of events depicted in media contents associated with the second group.
12 . The non-transitory computer readable medium of claim 1 , wherein each one of the first textual input and the second textual input is indicative of a respective desire of an individual to generate a respective visual content depicting two objects, the first new media content includes a depiction of the two objects positioned at a spatial relation characteristic to media contents associated with the first group, and the second new media content includes a depiction of the two objects positioned at a spatial relation characteristic to media contents associated with the second group.
13 . The non-transitory computer readable medium of claim 1 , wherein each one of the first textual input and the second textual input is indicative of a respective desire of an individual to generate a respective visual content depicting two events, the first new media content includes a depiction of the two events in a temporal relation characteristic to media contents associated with the first group, and the second new media content includes a depiction of the two events in a temporal relation characteristic to media contents associated with the second group.
14 . The non-transitory computer readable medium of claim 1 , wherein each one of the first textual input and the second textual input is indicative of a respective desire of an individual to generate a respective audio content including a vocalization of a speaker, the first new media content includes the vocalization of the speaker with a characteristic of vocalizations in media contents associated with the first group, and the second new media content includes the vocalization of the speaker with a characteristic of vocalizations in media contents associated with the second group.
15 . The non-transitory computer readable medium of claim 1 , wherein each one of the first textual input and the second textual input is indicative of a respective desire of an individual to generate a respective textual content including a description, the first new media content includes the description in a language register characteristic to media contents associated with the first group, and the second new media content includes the description in a language register characteristic to media contents associated with the second group.
16 . The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:
receiving a third textual input, the third textual input does not include the first source-identifying-keyword or the second source-identifying-keyword;
using the inference model and the third textual input to generate a third new media content, the third new media content is based on both the first parameter and the second parameter of the inference model.
17 . The non-transitory computer readable medium of claim 1 , wherein the operations further comprise:
receiving a third textual input, the third textual input does not include the first source-identifying-keyword or the second source-identifying-keyword;
using the inference model and the third textual input to generate a third new media content, the third new media content is based on neither the first parameter nor the second parameter of the inference model.
18 . The non-transitory computer readable medium of claim 1 , wherein
the first group is associated with a first region of a mathematical space;
the second group is associated with a second region of the mathematical space;
the using the inference model and the first textual input to generate the first new media content includes using the inference model to identify a first mathematical object in the first region of the mathematical space corresponding to the first textual input, and using the identified first mathematical object to generate the first new media content; and
the using the inference model and the second textual input to generate the second new media content includes using the inference model to identify a second mathematical object in the second region of the mathematical space corresponding to the second textual input, and using the identified second mathematical object to generate the second new media content.
19 . A method for inference based on different portions of a training set using a single inference model, the method comprising:
receiving a first textual input in a natural language, the first textual input includes a first source-identifying-keyword;
receiving a second textual input in the natural language, the second textual input includes a second source-identifying-keyword, the second source-identifying-keyword differs from the first source-identifying-keyword;
accessing an inference model, the inference model is a result of training a machine learning model using a plurality of training examples, each training example of the plurality of training examples includes a respective textual content in the natural language and a respective media content, the plurality of training examples includes a first group of at least two training examples associated with a first source and a second group of at least two training examples associated with a second source, the first group differs from the second group, the first source differs from the second source, each textual content of the textual contents associated with the first group includes the first source-identifying-keyword, each textual content of the textual contents associated with the second group includes the second source-identifying-keyword, a first parameter of the inference model is based on the first group, a second parameter of the inference model is based on the second group, the first parameter differs from the second parameter;
using the inference model and the first textual input to generate a first new media content, the first new media content is based on the first parameter of the inference model; and
using the inference model and the second textual input to generate a second new media content, the second new media content is based on the second parameter of the inference model.
20 . A system for inference based on different portions of a training set using a single inference model, the system comprising:
at least one processor configured to perform the operations of:
receiving a first textual input in a natural language, the first textual input includes a first source-identifying-keyword;
receiving a second textual input in the natural language, the second textual input includes a second source-identifying-keyword, the second source-identifying-keyword differs from the first source-identifying-keyword;
accessing an inference model, the inference model is a result of training a machine learning model using a plurality of training examples, each training example of the plurality of training examples includes a respective textual content in the natural language and a respective media content, the plurality of training examples includes a first group of at least two training examples associated with a first source and a second group of at least two training examples associated with a second source, the first group differs from the second group, the first source differs from the second source, each textual content of the textual contents associated with the first group includes the first source-identifying-keyword, each textual content of the textual contents associated with the second group includes the second source-identifying-keyword, a first parameter of the inference model is based on the first group, a second parameter of the inference model is based on the second group, the first parameter differs from the second parameter;
using the inference model and the first textual input to generate a first new media content, the first new media content is based on the first parameter of the inference model; and
using the inference model and the second textual input to generate a second new media content, the second new media content is based on the second parameter of the inference model.