Generating accompanying text creative
A method comprises: collecting data including combinations of images and accompanying text and user feedback of the combinations; building training data sets based on the collected data; training a plurality of neural networks using the training data; generating a creative feature vector based on a specified image using a first network of the trained plurality of neural networks; generating a target audience vector based on a specified target audience using a second network of the trained plurality of networks; generating a sequence of words based on the vectors using a third network of the plurality of trained neural networks; and transmitting the generated sequence of words and the specified image to the target audience over a network.
1. A method, comprising:
collecting data including combinations of images and accompanying text and user feedback of the combinations;
building training data sets based on the collected data;
training a plurality of neural networks using the training data;
generating a creative feature vector based on a specified image using a first network of the trained plurality of neural networks;
generating a target audience vector based on a specified target audience using a second network of the trained plurality of networks;
generating a sequence of words based on the vectors using a third network of the plurality of trained neural networks; and
transmitting the generated sequence of words and the specified image to the target audience over a network.
2. The method of claim 1 , wherein the user feedback includes click throughs and conversions.
3. The method of claim 1 , further comprising collecting audience data for the combinations and further training the plurality of neural networks based on the collected audience data.
4. The method of claim 1 , wherein the third neural network is a recurrent neural network.
5. The method of claim 1 , wherein the third neural network comprises a series of long short-term memory units.
6. The method of claim 5 , wherein an input to a first long short-term memory unit in the series includes the vectors and an output of the first long short-term memory unit is the first word of the generated sequence of words.
7. The method of claim 6 , wherein an input to subsequent long short-term memory units in the series are the vectors and a previous word.
8. The method of claim 1 , wherein the generated creative feature vector is a lower dimensional representation of the specified image.
9. The method of claim 1 , wherein the generated target audience vector is an information-dense vector representation.
10. A non-transitory computer-readable medium having stored thereon instruction to cause a computer to execute a method, the method comprising:
collecting data including combinations of images and accompanying text and user feedback of the combinations;
building training data set based on the collected data;
training a plurality of neural networks using the training data;
generating a creative feature vector based on a specified image using a first network of the trained plurality of neural networks;
generating a target audience vector based on a specified target audience using a second network of the trained plurality of networks;
generating a sequence of words based on a concatenated vectors using a third network of the plurality of trained neural networks; and
transmitting the generated sequence of words and the specified image to the target audience over a network.
11. An apparatus, comprising:
a processor; and
a non-transitory memory having stored thereon instructions to cause the processor to execute a method, the method comprising
collecting data including combinations of images and accompanying text and user feedback of the combinations;
building training data set based on the collected data;
training a plurality of neural networks using the training data;
generating a creative feature vector based on a specified image using a first network of the trained plurality of neural networks;
generating a target audience vector based on a specified target audience using a second network of the trained plurality of networks;
generating a sequence of words based on a concatenated vectors using a third network of the plurality of trained neural networks; and
transmitting the generated sequence of words and the specified image to the target audience over a network.
12. The apparatus of claim 11 , wherein the user feedback includes click throughs and conversions.
13. The apparatus of claim 11 , wherein the method further comprises collecting audience data for the combinations and further training the plurality of neural networks based on the collected audience data.
14. The apparatus of claim 11 , wherein the third neural network is a recurrent neural network.
15. The apparatus of claim 11 , wherein the third neural network comprises a series of long short-term memory units.
16. The apparatus of claim 15 , wherein an input to a first long short-term memory unit in the series includes the vectors and an output of the first long short-term memory unit is a first word of the generated sequence of words.
17. The apparatus of claim 16 , wherein an input to subsequent long short-term memory units in the series is the vectors and a previous word.
18. The apparatus of claim 11 , wherein the generated creative feature vector is a lower dimensional representation of the specified image.
19. The apparatus of claim 11 , wherein the generated target audience vector is an information-dense vector representation.
20. The apparatus of claim 11 , wherein the method further comprising:
conducting A/B testing on the transmitted sequence of words and the specified image;
updating the training data set based on results of the A/B testing; and
retraining the plurality of neural networks accordingly.