SYSTEMS AND METHODS FOR TRAINING LANGUAGE MODELS TO GENERATE INSTRUCTIONS FOR NETWORK OPTIMIZATION
Described herein are systems and methods for generating instructions for network optimization using language models. A system can maintain data corresponding to a plurality of historical wager opportunities. Using this data, the system can generate a training dataset comprising a plurality of training examples. Each training example can include an input prompt requesting evaluation of a condition related to wager opportunities, where the conditional intent specifies the condition, and a corresponding output message identifying one or more commands to evaluate the condition. The commands can be derived based on the conditional intent and the request. The system can update a language model using the generated training dataset.
1 . A system, comprising:
one or more processors coupled to non-transitory memory, the one or more processors configured to:
maintain a plurality of historical wager opportunities;
generate, using the data corresponding to the plurality of historical wagers, a training dataset comprising a plurality of training examples, at least one training example comprising:
(i) a respective input prompt including a request indicating a conditional intent relating to wager opportunities, the conditional intent specifying a condition to be evaluated for the request, and
(ii) a respective output message identifying one or more commands to evaluate the condition, the one or more commands derived based on the conditional intent and the request; and
update a language model using the training dataset.
2 . The system of claim 1 , wherein the one or more processors are further configured to:
classify the condition in a first training example of the plurality of training examples; and
select the one or more commands for the first training example based on the classification of the condition.
3 . The system of claim 2 , wherein the condition for the first training example identifies an occurrence of an event, and the one or more processors are further configured to:
select the one or more commands for the first training example that cause determination of whether the occurrence of the event is satisfied.
4 . The system of claim 1 , wherein the respective input prompt further comprises data corresponding to the live event retrieved from one or more databases.
5 . The system of claim 1 , wherein the respective input prompt of a first training example of the plurality of training examples comprises a plurality of conditions, and wherein the one or more processors are further configured to:
generate a plurality of commands for the respective output message of the training example.
6 . The system of claim 1 , wherein the one or more commands are identified using at least one command token in the respective output message of each of the plurality of training examples.
7 . The system of claim 1 , wherein the request of a first training example of the plurality of training examples specifies a future condition to monitor, and wherein the one or more processors are further configured to:
generate the respective output message of the first training example to include at least one command to monitor the future condition.
8 . The system of claim 7 , wherein the at least one command includes a command to update a list of conditions to monitor at periodic intervals.
9 . The system of claim 7 , wherein the one or more processors are further configured to:
generate the respective output message of the first training example to include at least one second command to perform an action determined from the respective input prompt of the first training example upon determining that the future condition is satisfied.
10 . The system of claim 1 , wherein the action comprises an operation to update a player profile with an indication of a wager opportunity.
11 . A method, comprising:
maintaining, by one or more processors coupled to non-transitory memory, a plurality of historical wager opportunities;
generating, by the one or more processors, using the data corresponding to the plurality of historical wagers, a training dataset comprising a plurality of training examples, at least one training example comprising:
(i) a respective input prompt including a request indicating a conditional intent relating to wager opportunities, the conditional intent specifying a condition to be evaluated for the request, and
(ii) a respective output message identifying one or more commands to evaluate the condition, the one or more commands derived based on the conditional intent and the request; and
updating, by the one or more processors, a language model using the training dataset.
12 . The method of claim 11 , further comprising:
classifying, by the one or more processors, the condition in a first training example of the plurality of training examples; and
selecting, by the one or more processors, the one or more commands for the first training example based on the classification of the condition.
13 . The method of claim 12 , wherein the condition for the first training example identifies an occurrence of an event, further comprising:
selecting, by the one or more processors, the one or more commands for the first training example that cause determination of whether the occurrence of the event is satisfied.
14 . The method of claim 11 , wherein the respective input prompt further comprises data corresponding to the live event retrieved from one or more databases.
15 . The method of claim 11 , wherein the respective input prompt of a first training example of the plurality of training examples comprises a plurality of conditions, further comprising:
generating, by the one or more processors, a plurality of commands for the respective output message of the training example.
16 . The method of claim 11 , wherein the one or more commands are identified using at least one command token in the respective output message of each of the plurality of training examples.
17 . The method of claim 11 , wherein the request of a first training example of the plurality of training examples specifies a future condition to monitor, further comprising:
generating, by the one or more processors, the respective output message of the first training example to include at least one command to monitor the future condition.
18 . The method of claim 17 , wherein the at least one command includes a command to update a list of conditions to monitor at periodic intervals.
19 . The method of claim 17 , further comprising:
generating, by the one or more processors, the respective output message of the first training example to include at least one second command to perform an action determined from the respective input prompt of the first training example upon determining that the future condition is satisfied.
20 . The method of claim 11 , wherein the action comprises an operation to update a player profile with an indication of a wager opportunity.