George Kour

IBM Research · AI Research

The ability to compare the semantic similarity between text corpora is important in a variety of natural language processing applications. However, standard methods for evaluating these metrics have yet to be established. We propose a set of automatic and interpretable measures for assessing the characteristics of corpus-level semantic similarity m...

Models for text generation have become focal for many research tasks and especially for the generation of sentence corpora. However, understanding the properties of an automatically generated text corpus remains challenging. We propose a set of tools that examine the properties of generated text corpora. Applying these tools on various generated co...

Conversational systems or chatbots are an example of AI-Infused Applications (AIIA). Chatbots are especially important as they are often the first interaction of clients with a business and are the entry point of a business into the AI (Artificial Intelligence) world. The quality of the chatbot is, therefore, key. However, as is the case in general...

Testing Machine Learning (ML) models and AI-Infused Applications (AIIAs), or systems that contain ML models, is highly challenging. In addition to the challenges of testing classical software, it is acceptable and expected that statistical ML models sometimes output incorrect results. A major challenge is to determine when the level of incorrectnes...

Based on recent advances in natural language modeling and those in text generation capabilities, we propose a novel data augmentation method for text classification tasks. We use a powerful pre-trained neural network model to artificially synthesize new labeled data for supervised learning. We mainly focus on cases with scarce labeled data. Our met...

Based on recent advances in natural language modeling and those in text generation capabilities, we propose a novel data augmentation method for text classification tasks. We use a powerful pre-trained neural network model to artificially synthesize new labeled data for supervised classification tasks. We mainly focus on cases with scarce labeled d...

In this paper, we aim at identifying, on a trial-by-trial basis, the underlying decision-making process of an animal in a complex and changing environment. We propose a method for identifying the set of stochastic policies employed by the agent and estimating the transition dynamics between policies based on its behavior in a multidimensional discr...

Deep neural networks work well at approximating complicated functions when provided with data and trained by gradient descent methods. At the same time, there is a vast amount of existing functions that programmatically solve different tasks in a precise manner eliminating the need for training. In many cases, it is possible to decompose a task to...

Existing applications include a huge amount of knowledge that is out of reach for deep neural networks. This paper presents a novel approach for integrating calls to existing applications into deep learning architectures. Using this approach, we estimate each application's functionality with an estimator, which is implemented as a deep neural netwo...

The problem of accurately predicting handling time for software defects is of great practical importance. However, it is difficult to suggest a practical generic algorithm for such estimates, due in part to the limited information available when opening a defect and the lack of a uniform standard for defect structure. We suggest an algorithm to add...

We present an approach for Business Intelligence (BI), where market share changes are tracked, evaluated, and prioritized dynamically and interactively. Out of all the hundreds or thousands of possible combinations of sub-markets and players, the system brings to the user those combinations where the most significant changes have happened, grouped...

Real-time performance is necessary in applications involving on-line handwriting recognition. However, conventional approaches usually wait until the entire curve is traced out before starting the analysis, inevitably causing delays in the recognition process. In regards to the Arabic script, the postponed analysis may be attributed to the cursive...

Despite the long-standing belief that digital computers will challenge the future of handwriting, pen and paper remain commonly used means for communication and recording of information in daily life. In addition to the growing use of keyboard-less devices such as smart-phones and tablets, which are too small to have a convenient keyboard, handwrit...

Delaying the analysis launch until the completion of the handwritten word scribing, restricts on-line recognition systems to meet the highly responsiveness demands expected from such applications, and prevents implementing advanced features of input typing such as automatic word completion and real-time automatic spelling. This paper proposes an ef...