Skip to main content

IBM R&D Labs in Israel

IBM Research Colloquium on Cognitive Computing 2014

Tab navigation

09:00 Registration

09:45 Opening Remarks

10:00 Personalization in Cortana,
Hadas Bitran, Microsoft

Abstract: The new Microsoft personal assistant, Cortana, uses a conversational natural user interface to interact with users. Cortana knows you best and aims to help you in ways that work for you.
The Cortana experiences are built by understanding the user, and matching the user understanding to the world knowledge in a personalized way. This session will demonstrate Cortana's natural user interaction and will discuss how Cortana personalizes the experiences she serves to users, using live examples.

Bio: Hadas Bitran is Principal Group Program Manager at the Microsoft Israel R&D Center. She manages a product team in charge of personalization of experiences in Bing. Hadas holds a B.Sc. in computer science from Tel Aviv University and an MBA from Kellogg School of Management, Northwestern University.

10:30 Conversational Speech Understanding, Applications and Challenges,
Moshe Wasserblat, Intel

Abstract: This lecture is focused on conversational speech understanding. This technology deals with understanding, analyzing, and extracting valuable business and personal insight from human-to-human verbal conversations (e.g., meetings). Unlike existing human-to-machine solutions (e.g. SIRI), the challenges induced by Conversational Understanding are currently unaddressed by the industry. It is a generic technology that can enable multiple capabilities critical to rising usages (e.g. Meeting Assistants, Argumentation, Customers Experience, Life-logging etc…).

Bio: Mr. Moshe Wasserblat is currently a senior NLP architect in the Intel Cognitive Computing group. In his former role he was with NICE Systems for more than 17 years, where he founded and led the Speech Analytics Research Team. His main interests are in the field of Speech Processing and Conversational Natural Language Processing. He was the co-founder coordinator of the EXCITEMENT FP7 ICT program on Semantic Inference and served as organizer and manager of several Israeli Chief Scientist initiatives. He has filed more than 40 patents in the field of language technology and also has several publications in international conferences and journals.

11:00 The Role of IBM Watson in Advancing the Era of Cognitive Computing,
Rob High, IBM Watson

Abstract: Humans have been employing machines for eons to augment and amplify their physical strength. Hydraulics, pneumatics, internal combustion, and electric motors have all enabled us to do things that we couldn't do before. Similarly, with the advent of electronic computing in the 1940s, we have been able to amplify our mathematical strength to process vast amounts of quantifiable data. This has enabled us to solve enormous business problems -- from payroll and banking account transaction processing, through high speed telecommunications and digital media. However, there is a vast space of the human condition that computers have traditionally not been well suited -- the problem of deeply understanding human expression and human reasoning. With Watson, we have begun a new era of computing that promises to amplify our consume information more deeply, and to inspire greater and more meaningful decisions. In this talk, I will outline some of the things IBM has been doing with Watson to employ the principle of cognitive computing to decision processes -- from Healthcare through to Financial Investments, all the way through to product selection and support.

Bio: Rob High currently serves as CTO for IBM's Watson Solutions, driving cognitive system development for the application of deep natural language processing in inferring answers from massive quantities of unstructured information. He was previously the chief architect for IBM's SOA Foundation -- a composition of IBM software products spanning WebSphere, Rational, Tivoli, Lotus and Information Management intended to enable the entire lifecycle of SOA-based applications. Before that, Rob was the chief architect for IBM's WebSphere Foundation -- covering the WebSphere Application Server, Network Deployment, Extended Deployment, Community Edition, Express Edition, and their relationship to WebSphere Portal, WebSphere Process Server, WebSphere ESB, WebSphere Commerce and other related offerings.

11:30 Break

11:45 Attention as a Key Factor of Effective Learning,
Lilach Shalev-Mevorach, Tel-Aviv University

Abstract: Attention is a cognitive resource that plays a major role in most everyday activities. Contemporary theories in cognitive neuroscience characterize attention as a multifaceted system comprised of several distinct functions. In recent years there is a growing interest in understanding the relation between the functions of attention and normal development of social, behavioral and academic skills. Various studies have demonstrated the relationship between visual attention and specific academic skills. For instance, the ability to focus attention on a spatially restricted area is a major predictor of the development of writing in young children. The same mechanism, namely, selective (spatial) attention is also important in early stages of reading. However, attention plays a crucial role in learning not only in young children. One of the obstacles of adult e-learners is the difficulty in sustaining attention to online lectures over long periods of time. As a result, quite frequently, students are facing mind wandering. It was found that interpolated active memory tests reduce mind wandering, increase note taking and improve learning of online lectures. Thus, a potentially powerful way that may improve the efficiency of learning is to provide conditions that will facilitate allocation of attention to the learning material. This interesting line of research will be further discussed.

Bio: Prof. Lilach Shalev-Mevorach is Head of the Attention Lab, School of Neuroscience at Tel-Aviv University. Her research topics include: neuropsychology of attention and attention deficits, assessment of attention, cognitive training, the role of attention in learning, visual perception and attention in high functioning autism. The research methods involved include: behavioral, cognitive and brain imaging (EEG and fMRI).

12:15 Using Crowd-Based Data Selection to Improve the Predictive Power of Search Trend Data,
Tomer Geva, Tel-Aviv University

Abstract: Large-scale data generated by crowds provide a myriad of opportunities for monitoring and modeling people's intentions, preferences, and opinions. A crucial step in analyzing such "Big Data" is identifying the relevant data items that should be provided as input to the modeling process. Interestingly, this important step has received limited attention in previous research. This paper proposes a novel crowd-based approach to this data selection problem: leveraging crowds to amplify the predictive capacity of search trend data (Google Trends). We developed an online word association task that taps into people's "thought-collection" process when thinking about a focal term. We empirically tested this method in two domains that have been used as test-beds for prediction. The method yields predictions that are equivalent or superior to those obtained in previous studies (using alternative data selection methods) and to predictions obtained using various benchmark data selection methods.

Bio: Tomer Geva is an Assistant Professor at the Recanati Business School, Tel Aviv University. Tomer's research focuses on understanding the utility and informativeness of large scale data, for the purpose of deriving business decisions. Tomer's work involves usage of data-science methods for predictive modeling as well as econometric analysis. Tomer teaches graduate and undergraduate Data Science and Business Data Analytics courses. Tomer holds a PhD from Tel-Aviv University, an MBA (cum laude) from Tel-Aviv university, and a BSc (cum laude) in Industrial Engineering from the Technion – Israel Institute of Technology.

12:45 Emotion Analysis in the Organization,
Michal Shmueli-Scheuer, IBM Research - Haifa

Abstract: Modern workplaces involve a lot of emotions. Emotions play an integral role in employee motivation, personal commitment and social interactions at work. With the development of enterprise social networking platforms that allow people to communicate and express themselves in the enterprise, there is an excellent, genuine and valid picture of emotions expressed in real-life communication of organization members.

Our goal in this work is to show that text-based analysis of emotional expressions in such platforms provides important insights about the emotional climate within an organization. Our analysis shows that this platform provides a picture of the emotional culture or etiquette: which emotions are expressed, when and by whom. We show differences in the emotions expressed by various segments of the employees. And we show emotion expressions at different times of the day and the week, as well as in relation to external events.

Bio: Dr. Michal Shmueli-Scheuer is a researcher in the Information Retrieval group at IBM Research – Haifa. Michal received her PhD in Information and Computer Science at the University of California, Irvine in 2009. Her area of expertise is in the fields of large scale analytics, database, and information systems, focusing on user behavior analytics, affective computing, and information management on the web. She has authored numerous papers on data management and information retrieval in leading conferences.

13:00 Demos/Posters 3 minutes madness

13:15 Lunch and Demos

14:15 Keynote: Probabilistic Topic Models and User Behavior,
David Blei, Columbia University

Abstract: Probabilistic topic models provide a suite of tools for analyzing large document collections. Topic modeling algorithms discover the latent themes that underlie the documents and identify how each document exhibits those themes. Topic modeling can be used to help explore, summarize, and form predictions about documents. Topic modeling ideas have been adapted to many domains, including images, music, networks, genomics, and neuroscience.

Traditional topic modeling algorithms analyze a document collection and estimate its latent thematic structure. However, many collections contain an additional type of data: how people use the documents. For example, readers click on articles in a newspaper website, scientists place articles in their personal libraries, and lawmakers vote on a collection of bills. Behavior data is essential both for making predictions about users (such as for a recommendation system) and for understanding how a collection and its users are organized.

In this talk, I will review the basics of topic modeling and describe our recent research on collaborative topic models, models that simultaneously analyze a collection of texts and its corresponding user behavior. We studied collaborative topic models on 80,000 scientists' libraries from Mendeley and 100,000 users' click data from the arXiv. Collaborative topic models enable interpretable recommendation systems, capturing scientists' preferences and pointing them to articles of interest. Further, these models can organize the articles according to the discovered patterns of readership. For example, we can identify articles that are important within a field and articles that transcend disciplinary boundaries.

More broadly, topic modeling is a case study in the large field of applied probabilistic modeling. Finally, I will survey some recent advances in this field. I will show how modern probabilistic modeling gives data scientists a rich language for expressing statistical assumptions and scalable algorithms for uncovering hidden patterns in massive data.

15:15 Break

15:30 Better Word Representations,
Yoav Goldberg, Bar Ilan University

Abstract: To a computer, "burger", "hamburger" and "pizza" are all unique symbols with no connection between them. There is a need to represent words such that words with similar meanings have similar representations. In this talk, I provide a brief survey of the traditional approach to word representation, where each word is represented as a sparse high dimensional vector of the contexts it appeared in. I will then describe the more recent approach in which each word is "embedded" in a relatively low dimensional continous space using techniques derived from the neural-networks literature.

These neural-embedded word vectors, popularized by the word2vec software package, result in word representations that exhibit some intriguing semantic properties. While the traditional and the neural approaches seem very different from each other, I argue that they are in fact quite similar. I will demonstrate that the intriguing properties of the embedded representations can be recovered also in the traditional representations, and that many components can be transferred between the two approaches, yielding better representations under both frameworks.

Bio: Yoav Goldberg is a senior lecturer in Bar Ilan's Computer Science Department. His research interests revolve around natural language processing and machine learning, in particular syntactic and semantic analysis and representation of textual data. Prior to joining Bar Ilan, Yoav worked as a researcher in Google's natural language understanding research team. He obtained his PhD from Ben Gurion University.

16:00 IBM Debating Technologies - How Persuasive Can a Computer Be?,
Dan Gutfreund, IBM Research - Haifa

Abstract: In complex decision making scenarios there are no right or wrong answers but rather different opinions, points of view, and arguments. Presenting perspective, arguing and persuading have always been considered abilities exclusive to humans - until now. In this talk I will present "the Debater", an IBM Research project whose goal is to develop technologies that will assist humans to debate and reason by digesting huge corpora of unstructured data and automatically generating arguments for or against a given point of view.

Bio: Dr. Dan Gutfreund received his PhD in computer science from the Hebrew University in Jerusalem, Israel. Following that he spent four years at Harvard University and MIT as a post-doctorate fellow and a lecturer. In 2009 Dan joined the Analytics department at IBM Research - Haifa. Today Dan manages the Machine Learning for Text Analytics group and leads the efforts around IBM Debating Technologies. Dan's research interests span various aspects of computer science, including: machine learning, computational complexity, error correcting codes and foundations of cryptography.

16:15 Concluding Remarks

16:30 Demos Reception