Adversarial learning framework for persona-based dialogue modeling
Various embodiments may be generally directed to the use of an adversarial learning framework for persona-based dialogue modeling. In some embodiments, automated multi-turn dialogue response generation may be performed using a persona-based hierarchical recurrent encoder-decoder-based generative adversarial network (phredGAN). Such a phredGAN may feature a persona-based hierarchical recurrent encoder-decoder (PHRED) generator and a conditional discriminator. In some embodiments, the conditional discriminator may include an adversarial discriminator that is provided with attribute representations as inputs. In some other embodiments, the conditional discriminator may include an attribute discriminator, and attribute representations may be handled as targets of the attribute discriminator. The embodiments are not limited in this context.
1. A computer-implemented method, comprising:
receiving, by at least one server communicatively coupled with a user device, a dialogue utterance;
applying, by the at least one server, a generative adversarial network (GAN) to the dialogue utterance to generate response candidates to the utterance and determine a response from the response candidates to respond to the utterance, the GAN comprising a generator and a discriminator, and wherein the applying the GAN comprises:
generating, by the generator utilizing source attributes and target attributes, the response candidates responsive to the dialogue utterance, wherein the generator further comprising a context recurrent neural network (cRNN), an encoder RNN (eRNN), an attention RNN (aRNN), and a decoder RNN (dRNN), the cRNN to concatenate the source attribute with an output of the eRNN to generate an initial state for the dRNN, and the dRNN to concatenate the target attribute with an output of the aRNN and generate a decoder output based on the initial state, wherein the source attributes comprise one or more of a speaker identity, a speaker background, a speaker location, a speaker preference, a speaker sentiment, or combination thereof, and the target attributes comprise one or more of a respondent identity, a respondent background, a respondent location, a respondent preference, a respondent sentiment, or a combination thereof;
determining, by the discriminator, the response to respond to the dialogue utterance from the response candidates, wherein the discriminator comprises a word-level adversarial discriminator and is configured to utilize the target attributes as inputs to determine the response, wherein the word-level adversarial discriminator is a bidirectional RNN, and each unit of the bidirectional RNN comprising a gate recurrent unit (GRU); and
causing, by the at least one server, communication of the response to the user device responsive to the dialogue utterance.
2. The computer-implemented method of claim 1 , wherein output of the aRNN is coupled to the dRNN using an additive attention mechanism.
3. The computer-implemented method of claim 1 , wherein the eRNN is a bidirectional RNN and the cRNN is a unidirectional RNN, the eRNN and cRNN each comprising units, and each unit comprising a three-layer gate recurrent unit (GRU) with a hidden state size of 512.
4. The computer-implemented method of claim 1 , wherein the word-level adversarial discriminator (D adv ) is to determine the response, wherein an adversarial loss is cGAN adv = X i, C i+1, X i+1 [log D adv (X i , C i+1 , X i+1 )] X i ,Z i [1 log D adv (X i ,G(X i ,C i+1 ,Z 1 ))], wherein an optimization objective for the response is
G
*
,
D
adv
*
,
D
att
*
=
arg
min
G
(
(
max
D
adv
λ
G
adv
ℒ
cGAN
adv
(
G
,
D
adv
)
+
min
D
att
λ
G
att
ℒ
cGAN
att
(
G
,
D
att
)
+
λ
M
ℒ
MLE
(
G
)
)
and λ G att is set to 0, wherein G represents the generator, D adv represents an adversarial discriminator, X i represents a sequence of utterances, C i+1 represents the target attributes, X i+1 represents a ground truth, λ G adv represents an adversarial hyperparameter, λ G adv represents an attribute prediction hyperparameter, L cGAN att represents an attribute prediction loss, D att represents an attribute discriminator, λ M represents a word token hyperparameter, MLE (G) represents a maximum likelihood estimation (MLE) loss, and Z i represents a noise sample.
5. A system, comprising:
a networking interface;
memory to store instructions; and
processing circuitry, coupled with the memory and the networking interface, operable to execute the instructions, that when executed, cause the processing circuitry to:
detect a dialogue utterance received from a user device via the networking interface;
process the dialogue utterance utilizing a generative adversarial network (GAN) logic to generate response candidates to the dialogue utterance, wherein the GAN comprises generator logic and discriminator logic, and processing the dialogue utterance utilizing the GAN logic comprising:
generate, via the generator logic and utilizing utterance attributes, the response candidates responsive to the dialogue utterance, wherein the generator logic further comprises a context recurrent neural network (cRNN), an encoder RNN (eRNN), an attention RNN (aRNN), and a decoder RNN (dRNN), and the cRNN to concatenate a source attribute of the utterance attributes with an output of the eRNN to generate an initial state for the dRNN, and the dRNN to concatenate a target attribute of the utterance attributes with an output of the aRNN and generate a decoder output based on the initial state, and wherein the utterance attributes comprise one or more of a speaker identity, a speaker background, a speaker location, a speaker preference, a speaker sentiment, respondent identity, a respondent background, a respondent location, a respondent preference, a respondent sentiment, or a combination thereof;
determine, by the discriminator logic comprising a word-level adversarial discriminator (D adv ) logic utilizing at least one attribute of the utterance attributes as an input, a response to respond to the dialogue utterance from the response candidates wherein the word-level adversarial discriminator is a bidirectional RNN, and each unit of the bidirectional RNN comprising a gate recurrent unit (GRU); and
cause, via the networking interface, communication of the response to the user device responsive to the dialogue utterance.
6. The system of claim 5 , wherein output of the aRNN is coupled to the dRNN using an additive attention mechanism.
7. The system of claim 5 , wherein the eRNN is a bidirectional RNN and the cRNN is a unidirectional RNN, the eRNN and cRNN each comprising units, and each unit comprising a three-layer gate recurrent unit (GRU) with a hidden state size of 512.
8. The system of claim 5 , the discriminator logic to determine the response, wherein an adversarial loss is cGAN adv (G, D adv )= X i, C i+1, X i+1 [log D adv (X i , C i+1 , X i+1 )]+ X i ,Z i [1 log D adv (X i ,G(X i ,C i+1 ,Z 1 ))], and wherein an optimization objective for the response is
D
adv
*
,
D
att
*
=
arg
min
G
(
(
max
D
adv
λ
G
adv
ℒ
cGAN
adv
(
G
,
D
adv
)
+
min
D
att
λ
G
att
ℒ
cGAN
att
(
G
,
D
att
)
+
λ
M
ℒ
MLE
(
G
)
)
and λ G att is set to 0, wherein G represents the generator, D adv represents an adversarial discriminator, X i represents a sequence of utterances, C i+1 represents the target attributes, X i+1 represents a ground truth, λ G adv represents an adversarial hyperparameter, λ G att represents an attribute prediction hyperparameter, cGAN att represents an attribute prediction loss, D att represents an attribute discriminator, λ M represents a word token hyperparameter, MLE (G) represents a maximum likelihood estimation (MLE) loss, and Z i represents a noise sample.
9. A non-transitory computer-readable medium storing instructions that, when executed, cause one or more processors of one or more servers to:
receive, from a user device, a dialogue utterance;
generate, by generator logic of a generative adversarial network (GAN) utilizing source attributes and target attributes, response candidates responsive to the dialogue utterance, wherein the generator logic further comprising a context recurrent neural network (cRNN), an encoder RNN (eRNN), an attention RNN (aRNN), and a decoder RNN (dRNN), the cRNN to concatenate the source attribute with an output of the eRNN to generate an initial state for the dRNN, and the dRNN to concatenate the target attribute with an output of the aRNN and generate a decoder output based on the initial state, wherein the source attributes comprise one or more of a speaker identity, a speaker background, a speaker location, a speaker preference, a speaker sentiment, or combination thereof, and the target attributes comprise one or more of a respondent identity, a respondent background, a respondent location, a respondent preference, a respondent sentiment, or a combination thereof;
determine, by discriminator logic comprising a word-level adversarial discriminator (D adv ) of the GAN utilizing the target attributes, a response to respond to the dialogue utterance from the response candidates, wherein the word-level adversarial discriminator is a bidirectional RNN, and each unit of the bidirectional RNN comprising a gate recurrent unit (GRU); and
cause, via a networking interface, communication of the response to the user device responsive to the dialogue utterance.
10. The non-transitory medium of claim 9 , wherein the eRNN logic is a bidirectional RNN and the cRNN logic is a unidirectional RNN, the eRNN logic and cRNN logic each comprising units, and each unit comprising a three-layer gate recurrent unit (GRU) with a hidden state size of 512.
11. The non-transitory medium of claim 9 , the discriminator logic to determine the response, wherein an adversarial loss is cGAN adv (G, D adv )= X i, C i+1, X i+1 [log D adv (X i , C i+1 , X i+1 )]+ X i ,Z i [1−log D adv (X i ,G(X i ,C i+1 ,Z 1 ))], and wherein an optimization objective for the response is
G
*
,
D
adv
*
,
D
att
*
=
arg
min
G
(
(
max
D
adv
λ
G
adv
ℒ
cGAN
adv
(
G
,
D
adv
)
+
min
D
att
λG
att
ℒ
cGAN
att
(
G
,
D
att
)
+
λ
M
ℒ
MLE
(
G
)
)
where λ G att is set to 0, wherein G represents the generator, D adv represents an adversarial discriminator, X i represents a sequence of utterances, C i+1 represents the target attributes, λ i+1 represents a ground truth, λ G adv represents an adversarial hyperparameter, λ G att represents an attribute prediction hyperparameter, cGAN att resents an attribute prediction loss, D att represents an attribute discriminator, λ M represents a word token hyperparameter, MLE (G) represents a maximum likelihood estimation (MLE) loss, and Z i represents a noise sample.