Alphonso Inc.

Susmita Ghose

Sartaki Sinha Roy

Aashiq Muhamed

Aayush Agrawal

Ashish Ramgopal Baldua

Dawnis Chow

An automated method and computer program product are provided for performing multi-channel attribution for conversion events associated with a plurality of consumers. Each consumer has an associated consumer identifier. Each conversion event is associated with a brand or product that has corresponding media advertising electronically delivered on a plurality of different media-based delivery channels to the plurality of consumers via a plurality of media devices associated with respective consumers. Media advertising exposure is electronically detected for each conversion event. Attribution for each of the delivery channels is then electronically determined.

Multi-channel attribution based on timing and number of exposures relative to conversion events

Candidate brands are identified for an ad clip of a query video advertisement using two different techniques. Optical character recognition (OCR) is performed on a thumbnail of a plurality of different video ads in a curated database that includes brand-identifying metadata. In one technique, text vectorization is performed for each brand using all of the text that is extracted from each of the thumbnails for each brand, thereby providing a numerical brand embedding for each brand, a similar text vectorization is performed on the query video ad, and then a comparison is made between the embeddings to identify highest-ranking candidate brands based on brand-level identification. In another technique, a transformer-based model generates numerical embeddings for each ad clip in the curated database, and a comparison is made to these clip-based numerical embeddings to identify highest-ranking candidate brands based on clip-level identification. The results of the two techniques are combined together to provide one or more overall highest-ranking candidate brands.

Apparatus and method for identifying candidate brand names for an ad clip of a query video advertisement using OCR data

A computer program product is provided for predicting ad spend for a specific media program aired or streamed on a specific network at a specific date and time using a database of media program data that includes known ad spend for a subset of media programs, and viewership data for each of the media programs, including total viewership and viewership ratings. Each of the media programs is identified by its respective network, and date and time of airing or streaming. A random forest model is trained to predict ad spend using variables that are identified as being correlated with ad spend. The trained random forest model is then used to predict ad spend for a specific media program that is aired or streamed on a specific network at a specific date and time, and which has an unknown ad spend.

Computer program product that implements a machine learning process using a random forest model for predicting advertisement spending

A method is provided to identify whether video content, which includes a plurality of image frames, is likely to include an advertisement. The video content is split into a plurality of segments, each segment having a pre-specified duration. Subtitle text information is extracted from each segment and is passed through a natural language processing (NLP) language model to extract an embedding representing the subtitle text information for each of the segments, wherein the NLP language model is previously trained to differentiate between subtitle text information from video content items that were each previously identified as being an advertisement in comparison to subtitle text information from video content items that were each previously identified as not being an advertisement. The embedding representing the subtitle text information for each of the segments is passed through a classifier to obtain a probability regarding whether each segment is an advertisement or not.

Closed-caption processing using machine learning for media advertisement detection

A method is provided to classify whether video content is likely to be an advertisement or a non-advertisement. A curated database of video content items that includes a plurality of different video content items that were each previously identified as being an advertisement, and a plurality of different video content items that were each previously identified as not being an advertisement, are used to train a 2D CNN and a 3D CNN. The training of the 2D CNN includes learning characteristic visual and spatial features of advertisement images in the video content items compared to non-advertisement images in the video content items, the training resulting in weights being defined for the 2D CNN. The training of the 3D CNN includes learning a temporal structure and relationship over multiple image frames of the advertisements in the video content items compared to non-advertisement image frames in the video content items, the training resulting in weights being defined for the 3D CNN. The trained 2D CNN and 3D CNN are then used to determine the probability that newly identified video content should be classified as an advertisement.

3D convolutional neural networks for television advertisement detection

An artificial neural network architecture is provided for processing raw audio waveforms to create speaker representations that are used for text-independent speaker verification and recognition. The artificial neural network architecture includes a strided convolution layer, first and second sequentially connected residual blocks, a transformer layer, and a final fully connected (FC) layer. The strided convolution layer is configured to receive raw audio waveforms from a speaker. The first and the second residual blocks both include multiple convolutional and max pooling layers. The transformer layer is configured to aggregate frame level embeddings to an utterance level embedding. The output of the FC layer creates a speaker representation for the speaker whose raw audio waveforms were inputted into the strided convolution layer.

Growing community of inventors

Mountain View, CA, United States of America

Susmita Ghose