Adithya Rao

Prajwal Chidananda, Ayan Sinha, Adithya Rao et al.

2D Key-point estimation is an important precursor to 3D pose estimation problems for human body and hands. In this work, we discuss the data, architecture, and training procedure necessary to deploy extremely efficient 2.5D hand pose estimation on embedded devices with highly constrained memory and compute envelope, such as AR/VR wearables. Our 2.5D hand pose estimation consists of 2D key-point estimation of joint positions on an egocentric image, captured by a depth sensor, and lifted to 2.5D using the corresponding depth values. Our contributions are two fold: (a) We discuss data labeling and augmentation strategies, the modules in the network architecture that collectively lead to $3\%$ the flop count and $2\%$ the number of parameters when compared to the state of the art MobileNetV2 architecture. (b) We propose an auxiliary multi-task training strategy needed to compensate for the small capacity of the network while achieving comparable performance to MobileNetV2. Our 32-bit trained model has a memory footprint of less than 300 Kilobytes, operates at more than 50 Hz with less than 35 MFLOPs.

Nemanja Spasojevic, Prantik Bhattacharyya, Adithya Rao

Mining topical experts on social media is a problem that has gained significant attention due to its wide-ranging applications. Here we present the first study that combines data from four major social networks -- Twitter, Facebook, Google+ and LinkedIn, along with the Wikipedia graph and internet webpage text and metadata, to rank topical experts across the global population of users. We perform an in-depth analysis of 37 features derived from various data sources such as message text, user lists, webpages, social graphs and wikipedia. This large-scale study includes more than 12 billion messages over a 90-day sliding window and 58 billion social graph edges. Comparison reveals that features derived from Twitter Lists, Wikipedia, internet webpages and Twitter Followers are especially good indicators of expertise. We train an expertise ranking model using these features on a large ground truth dataset containing almost 90,000 labels. This model is applied within a production system that ranks over 650 million experts in more than 9,000 topical domains on a daily basis. We provide results and examples on the effectiveness of our expert ranking system, along with empirical validation. Finally, we make the topical expertise data available through open REST APIs for wider use.

Adithya Rao, Nemanja Spasojevic

In this work, we apply word embeddings and neural networks with Long Short-Term Memory (LSTM) to text classification problems, where the classification criteria are decided by the context of the application. We examine two applications in particular. The first is that of Actionability, where we build models to classify social media messages from customers of service providers as Actionable or Non-Actionable. We build models for over 30 different languages for actionability, and most of the models achieve accuracy around 85%, with some reaching over 90% accuracy. We also show that using LSTM neural networks with word embeddings vastly outperform traditional techniques. Second, we explore classification of messages with respect to political leaning, where social media messages are classified as Democratic or Republican. The model is able to classify messages with a high accuracy of 87.57%. As part of our experiments, we vary different hyperparameters of the neural networks, and report the effect of such variation on the accuracy. These actionability models have been deployed to production and help company agents provide customer support by prioritizing which messages to respond to. The model for political leaning has been opened and made available for wider use.

Nemanja Spasojevic, Adithya Rao

Text actionability detection is the problem of classifying user authored natural language text, according to whether it can be acted upon by a responding agent. In this paper, we propose a supervised learning framework for domain-aware, large-scale actionability classification of social media messages. We derive lexicons, perform an in-depth analysis for over 25 text based features, and explore strategies to handle domains that have limited training data. We apply these methods to over 46 million messages spanning 75 companies and 35 languages, from both Facebook and Twitter. The models achieve an aggregate population-weighted F measure of 0.78 and accuracy of 0.74, with values of over 0.9 in some cases.

Adithya Rao, Nemanja Spasojevic, Zhisheng Li et al.

In this work, we present the Klout Score, an influence scoring system that assigns scores to 750 million users across 9 different social networks on a daily basis. We propose a hierarchical framework for generating an influence score for each user, by incorporating information for the user from multiple networks and communities. Over 3600 features that capture signals of influential interactions are aggregated across multiple dimensions for each user. The features are scalably generated by processing over 45 billion interactions from social networks every day, as well as by incorporating factors that indicate real world influence. Supervised models trained from labeled data determine the weights for features, and the final Klout Score is obtained by hierarchically combining communities and networks. We validate the correctness of the score by showing that users with higher scores are able to spread information more effectively in a network. Finally, we use several comparisons to other ranking systems to show that highly influential and recognizable users across different domains have high Klout scores.