Abstract: Systems and methods for video query contextualization can include a router model that determines how to process and respond to the query associated with the video. The systems and methods can include obtaining an input query and video data, processing the input query and the video data with the router model to generate a video clip and routing data, and the routing data can then be utilized to determine which processing system to utilize to process the video clip and the input query. The video clip can then be processed with the determined processing system to generate a query response that may be provided to the user.

Abstract: Methods and systems for media item characterization based on multimodal embeddings are provided herein. A media item including a sequence of video frames is identified. A set of video embeddings representing visual features of the sequence of video frames is obtained. A set of audio embeddings representing audio features of the sequence of video frames is obtained. A set of audiovisual embeddings is generated based on the set of video embeddings and the set of audio embeddings. Each of the set of audiovisual embeddings represents a visual feature and an audio feature of a respective video frame of the sequence of video frames. One or more media characteristics associated with the media item are determined based on the set of audiovisual embeddings.

Abstract: Systems and methods for directing behavior of a generative artificial intelligence (AI) system are provided. In particular, a computing device may obtain an input prompt associated with a requested task for one or more generative artificial intelligence (AI) systems, obtain one or more attributes based on the input prompt, modify the input prompt based on the one or more embedded attributes, and provide the modified input prompt to the one or more generative AI systems.

Abstract: Methods and systems for predicting titles for contents segments of media items at a platform using machine-learning are provided herein. A media item is provided to users of a platform, the media item having a plurality of content segments comprising a first content segment and a second content segment preceding the first content segment in the media item. The first content segment and a title of the second content segment are provided as input to a machine-learning model trained to predict a title for the first content segment that is consistent with the title of the second content segment. One or more outputs of the machine-learning model are obtained which indicate the title for the first content segment. An indication of each content segment and a respective title of each content segment are provided for presentation to at least one user of the one or more users.

Abstract: A computing system is described that includes user interface components configured to receive typed user input; and one or more processors. The one or more processors are configured to: receive, by a computing system and at a first time, a first portion of text typed by a user in an electronic message being edited; predict, based on the first portion of text, a first candidate portion of text to follow the first portion of text; output, for display, the predicted first candidate portion of text for optional selection to append to the first portion of text; determine, at a second time that is after the first time, that the electronic message is directed to a sensitive topic; and responsive to determining that the electronic message is directed to a sensitive topic, refrain from outputting subsequent candidate portions of text for optional selection to append to text in the electronic message.

Abstract: A computing system is described that includes user interface components configured to receive typed user input; and one or more processors. The one or more processors are configured to: receive, by a computing system and at a first time, a first portion of text typed by a user in an electronic message being edited; predict, based on the first portion of text, a first candidate portion of text to follow the first portion of text; output, for display, the predicted first candidate portion of text for optional selection to append to the first portion of text; determine, at a second time that is after the first time, that the electronic message is directed to a sensitive topic; and responsive to determining that the electronic message is directed to a sensitive topic, refrain from outputting subsequent candidate portions of text for optional selection to append to text in the electronic message.

Abstract: A multi-horizon predictor system that predicts a future parameter value for multiple horizons based on time-series data of the parameter, external data, and machine-learning. For a given time horizon, a time series data splitter splits the time into training data corresponding to a training time period, and a validation time period corresponding to a validation time period between the training time period and the given horizon. A model tuner tunes the prediction model of the given horizon fitting an initial prediction model to the parameter using the training data thereby using machine learning. The model tuner also tunes the initial prediction model by adjusting an effect of the external data on the prediction to generate a final prediction model for the given horizon using the validation data. A multi-horizon predictor causes the time series data splitter and the model tuner to operate for each of multiple horizons.

Abstract: A computing system is described that includes user interface components configured to receive typed user input; and one or more processors. The one or more processors are configured to: receive, by a computing system and at a first time, a first portion of text typed by a user in an electronic message being edited; predict, based on the first portion of text, a first candidate portion of text to follow the first portion of text; output, for display, the predicted first candidate portion of text for optional selection to append to the first portion of text; determine, at a second time that is after the first time, that the electronic message is directed to a sensitive topic; and responsive to determining that the electronic message is directed to a sensitive topic, refrain from outputting subsequent candidate portions of text for optional selection to append to text in the electronic message.

Abstract: A multi-horizon predictor system that predicts a future parameter value for multiple horizons based on time-series data of the parameter, external data, and machine-learning. For a given time horizon, a time series data splitter splits the time into training data corresponding to a training time period, and a validation time period corresponding to a validation time period between the training time period and the given horizon. A model tuner tunes the prediction model of the given horizon fitting an initial prediction model to the parameter using the training data thereby using machine learning. The model tuner also tunes the initial prediction model by adjusting an effect of the external data on the prediction to generate a final prediction model for the given horizon using the validation data. A multi-horizon predictor causes the time series data splitter and the model tuner to operate for each of multiple horizons.

Abstract: A system that uses power spectrum analysis and auto-correlation function analysis to perform seasonality estimation of time series data. A power spectrum analyzer calculates and analyzes a power spectrum of a received time series data. An auto-correlation function analyzer calculates at least one auto-correlation function of the received time series, and generates a resulting set of one or more candidate seasonalities. A seasonality estimator estimates one or more seasonalities of the received time series using at least a portion of the analyzed result from the power spectrum analyzer and using the set of one or more candidates generated by the auto-correlation function analyzer. Accordingly, the estimation of candidate seasonality uses both auto-correlation and power spectrum analysis, thereby at least in some circumstances improving the seasonality estimation compared to auto-correlation function analysis alone or power spectrum analysis alone.