Data processing apparatus and method

1 . A data processing apparatus comprising circuitry configured to:

receive a sequence of one or more input data elements, the one or more input data elements indicating position information of one or more objects in a sports event at respective successive time steps;

perform, using a first machine learning model, a sequence prediction process using the sequence of one or more input data elements to predict a first predicted data element at a first future time step in a first period of play of the sports event, the first predicted data element indicating predicted position information of the one or more objects at the first future time step;

determine whether a break event associated with a break in the sports event is predicted to occur, a break event indicating an end of the first period of play;

if a break event is not predicted to occur, continue performing the sequence prediction process using at least the first predicted data element to predict a second predicted data element at a second future time step in the first period of play, the second predicted data element indicating predicted position information of the one or more objects at the second future time step; and

if a break event is predicted to occur:

end the sequence prediction process;

use at least the first predicted data element to predict, using a second machine learning model that is different from the first machine learning model, a third predicted data element, the third predicted data element indicating predicted position information of the one or more objects at a start of a second period of play of the sports event; and

restart, using the first machine learning model, the sequence prediction process for the second period of play using the third predicted data element.

2 . The data processing apparatus according to claim 1 , wherein the position information comprises one or more of an absolute position of each of the one or more objects and a change in position of each of the one or more objects with respect to a previous time step.

3 . The data processing apparatus according to claim 1 , wherein the one or more objects comprise one or more participants of the sports event and a ball.

4 . The data processing apparatus according to claim 3 , wherein the position information comprises a distance of each of the one or more participants from the ball.

5 . The data processing apparatus according to claim 1 , wherein each data element comprises a sub-element for each object indicating the position information of the object and a sub-element indicating a type of event which occurs in the sports event.

6 . The data processing apparatus according to claim 5 , wherein the circuitry is configured to classify each data element to determine whether the indicated type of event is a break event.

7 . The data processing apparatus according to claim 5 , wherein the one or more objects comprise a participant of the sports event and the sub-element for the participant indicates at least one of an identity of the participant and an identity of a team of the participant.

8 . The data processing apparatus according to claim 5 , wherein the sub-elements of each of the predicted data elements are predicted using only data elements of one or more previous time steps.

9 . The data processing apparatus according to claim 5 , wherein at least one sub-element of each of the predicted data elements is predicted using at least one already-predicted sub-element at a current time step.

10 . The data processing apparatus according to claim 1 , wherein the sports event is a soccer match and the break event is a goal, corner kick, goal kick, throw in, free kick, penalty kick or offside call.

11 . The data processing apparatus according to claim 1 , wherein the sports event is a tennis match and the break event is a shot classified as a winner or an error.

12 . A data processing method comprising:

receiving a sequence of one or more input data elements, the one or more input data elements indicating position information of one or more objects in a sports event at respective successive time steps;

performing, using a first machine learning model, a sequence prediction process using the sequence of one or more input data elements to predict a first predicted data element at a first future time step in a first period of play of the sports event, the first predicted data element indicating predicted position information of the one or more objects at the first future time step;

determining whether a break event associated with a break in the sports event is predicted to occur, a break event indicating an end of the first period of play;

if a break event is not predicted to occur, continue performing the sequence prediction process using at least the first predicted data element to predict a second predicted data element at a second future time step in the first period of play, the second predicted data element indicating predicted position information of the one or more objects at the second future time step; and

if a break event is predicted to occur:

ending the sequence prediction process;

using at least the first predicted data element to predict, using a second machine learning model that is different from the first machine learning model, a third predicted data element, the third predicted data element indicating predicted position information of the one or more objects at a start of a second period of play of the sports event; and

restarting, using the first machine learning model, the sequence prediction process for the second period of play using the third predicted data element.

13 . A non-transitory computer-readable storage medium storing a program according to claim 12 .