Koichi Kise, Ph.D., Professor
Dept. of Computer Science and Intelligent Systems, Graduate School of Engineering,
Osaka Prefecture University, Japan
Title: Document Analysis Meets Activity Recognition: A New Paradigm of Analyzing Documents in Combination With Users’ Reading Behavior
Abstract: As a new paradigm of research in the field of document analysis, I propose a notion of mutual analysis of documents and activities. Documents have been the single entity to be analyzed in the field. However their values cannot be fully estimated by just looking at their contents. People obtain information from documents by the activities of reading. The key to analyze documents for their values is in this point. The analysis of human reading activities gives us a way to obtain additional information on the contents of documents such as interesting and difficult. On the other hand, by augmenting the reading activities by the contents, we are able to know more about actors of these activities such as personal preferences.
In my talk I show some work on this topic including estimating the number of words read (wordometer), the type of documents and the level of understanding, as well as document annotations by reading behavior. I would also like to show some important open problems to be shared with the audience for discussion.
Short Bio: Koichi Kise received his B.E., M.E. and Ph.D. degrees in communication engineering from Osaka University, Osaka, Japan in 1986, 1988 and 1991, respectively.
From 2000 to 2001, he was a visiting professor at German Research Center for Artificial Intelligence (DFKI), Germany.
He is now a Professor of the Department of Computer Science and Intelligent Systems, Osaka Prefecture University, Japan.
He received awards including the best paper award of IEICE in 2008, the IAPR/ICDAR best paper awards in 2007 and 2013, the IAPR Nakano award in 2010, the ICFHR best paper award in 2010 and the ACPR best paper award in 2011. He is now working as the chair of the IAPR technical committee 11 (reading systems) and a member of the IAPR conferences and meetings committee. His major research activities are in analysis, recognition and retrieval of documents, images and activities.
Professor at the University of Fribourg, Switzerland
2013 Google Faculty Research Award, Google–USA
Title: Entity-Centric Data Management
Abstract: Until recently, structured (e.g., relational) and unstructured (e.g., textual) data were managed very differently: Structured data was queried declaratively using languages such as SQL, while unstructured data was searched using boolean queries over inverted indices. Today, we witness the rapid emergence of entity-centric techniques to bridge the gap between different types of content and manage both unstructured and structured data more effectively. I will start this talk by giving a few examples of entity-centric data management. I will then describe two recent systems that were built in my lab and revolve around entity-centric data management techniques: ZenCrowd, a socio-technical platform that automatically connects HTML documents to semi-structured entities, and TripleProv, a scalable, efficient, and provenance-enabled back-end to manage graphs of entities.
Brief Bio: Philippe Cudre-Mauroux is a Swiss-NSF Professor and the director of the eXascale Infolab at the University of Fribourg in Switzerland. Previously, he was a postdoctoral associate working in the Database Systems group at MIT. He received his Ph.D. from the Swiss Federal Institute of Technology EPFL, where he won both the Doctorate Award and the EPFL Press Mention in 2007. Before joining the University of Fribourg, he worked on distributed information and media management for HP, IBM Watson Research (NY), and Microsoft Research Asia. He was Program Chair of the International Semantic Web Conference in 2012 and General Chair of the International Symposium on Data-Driven Process Discovery and Analysis in 2012 and 2013. He recently won the Verisign Internet Infrastructures Award, a Swiss National Center in Research award, as well as a Google Facutly Research Award. His research interests are in next-generation, Big Data management infrastructures for non-relational data. Webpage: http://exascale.info/phil
Professor at the University at Buffalo, USA
2015 IAPR/ICDAR Outstanding Achievements Award Winner
Title: Accelerated Discovery: A Big Data Perspective
Abstract: Today, the advances in computing, storage, and machine learning algorithms make it possible for the entire scientific literature of any given field to be examined in its totality, so that papers across topics, years, authors, disciplines, and institutions can reveal linkages that have been thus far unapparent and which could lead to transformational discoveries. We present a vision for the ICDAR community to address this grand challenge by proposing a framework that would scale across disciplines and accelerate the entire process of scientific endeavor and discovery. Two major hurdles need to be overcome. First is the issue of the sheer volume of scientific literature in existence to date, which is conservatively pegged at nearly 100 million articles. The literature itself comes in a wide variety of formats, is of varying quality (necessitating assessment of the veracity of data and results), and continues to grow at a staggering rate (velocity). So, ironically, the body of scientific articles that typically present the analysis of data, themselves qualify as “Big Data”. The second hurdle is the quality of scientific search currently supported by search engines, which merely make access more efficient but falls short at facilitating easier and comprehensive understanding of the topic of the search. We present a roadmap to address the myriad of research challenges to bridge this gap that is crucial to the process of scientific discovery.
Brief Bio: Venu Govindaraju is a SUNY Distinguished Professor in the Department of Computer Science and Engineering, University at Buffalo, The State University of New York (UB). He is also currently the Vice President of Research and Economic Development at UB. Prof. Govindaraju received his undergraduate degree with honors (BTech) and Ph.D degrees in Computer Science from the Indian Institute of Technology (IIT) Kharagpur, India in 1986 and from the University at Buffalo, USA in 1992 respectively. He has been the Associate Director of the Center of Excellence for Document Analysis and Recognition (CEDAR) since 1995 and the founding director of the Center for Unified Biometrics and Sensors (CUBS) since its inception in 2003.
Prof. Govindaraju is the recipient of numerous peer honors such as the ICDAR Young Investigator award(2001), the MIT Global Indus Technovator Award (2004), the IEEE Technical Achievement Award (2010),and the Indian Institutes of Technology (IIT) Distinguished Alumnus Award(2014). He has been elected Fellow of the ACM (Association of Computing Machinery), the IEEE (Institute of Electrical and Electronics Engineers), the AAAS (American Association for the Advancement of Science), the IAPR (International Association of Pattern Recognition), and the SPIE (International Society of Optics and Photonics).
Prof. Govindaraju’s research has focused on the application of machine learning and pattern recognition techniques to application domains such as Document Analysis and Recognition and Biometrics. He has developed principled modeling approaches for pattern classification that have resulted in the development of robust, scalable systems. He has designed several algorithms for cursive handwriting recognition suitable for real time applications that demonstrated the benefits of innovative modeling of application constraints. He was amongst the first to explore human-like handwriting for designing CAPTCHAs to exploit the differential in handwriting reading proficiency between humans and machines. He defined the notion of lexicon density as a metric to measure the expected accuracy of handwritten word recognizers. He has also contributed to improvement in word recognition accuracy of unconstrained handwritten documents by applying OCR correction techniques in a bootstrapping framework where innovative topic categorization techniques are used to generate smaller topic-specific lexicons. In recognizing handwritten medical forms, he used partial recognition results to construct a linguistic model representing medical topic categories. His work in multilingual OCR ranges from the development of a recognition driven segmentation framework and the use of stochastic language models for Devanagari OCR (for the NSF Digital Libraries initiative), to the development of innovative pre-processing techniques and recognition strategies for Arabic OCR (for the DARPA MADCAT program). His language-motivated hierarchical modeling has been extended to computer vision applications such as scene understanding and classifying activities and gestures in unconstrained videos. He has also made contributions to the theoretical foundations of a general fusion architecture and taxonomy of trained combining functions (classifiers) and their input parameters which provides a principled guideline for choosing a particular fusion technique.
Prof. Govindaraju’s work has been particularly significant in the development of real-time engineered pattern recognition systems. The team led by Govindaraju and his colleagues at the University at Buffalo developed and delivered to the U.S. Postal Service a field-deployable real-time system for reading handwritten addresses on mailpieces yielding annual savings of several million dollars for the US Postal Service. A key element of the system was Prof. Govindaraju’s seminal work [IEEE TPAMI: 19(4):366-379 (1997)](that provided ground-breaking techniques to efficiently segment and process cursively written (often with illegible parts) words with the help of lexicons in real time. He was also the prime technical lead responsible for technology transfer to the Postal Services in US, Australia, and UK.
Prof. Govindaraju has authored 5 edited books, including the very first comprehensive book on the subject of OCR for Indic scripts, and about 400 refereed publications that include 78 journal papers, 22 book-chapters, and about 300 symposium/conference/workshop papers in areas of pattern recognition theory and its applications. He also serves on the editorial board of several premier journals including 3 IEEE transactions (IEEE-T-PAMI IEEE-T-SMC, IEEE-T-IFS) and IEEE Access. He is also the editor-in-chief of the IEEE Biometrics Council Compendium. Prof. Govindaraju has graduated 30 Ph.D scholars as major advisor. He has been awarded 4 patents including one on handwritten cursive word recognition.
Govindaraju has been involved in the organizing committees of IWFHR/ICFHR, ICDAR, IGS, DAS and ICPR conferences including serving as General Co-Chair of ICDAR 2013.
Program Manager, Information Innovation Office (I2O)
DARPA, USA
Title: The Role of Document Analysis in the Paperless Office: One perspective
Abstract: For decades we have been hearing of the demise of document understanding as a result of the “paperless office”, yet automation has actually created more paper than we had a dozen years ago. One question that we need to ask is “are we using paper the same way as we did before?” The fact is that the role of paper is shifting, and the type of document analysis tasks that are most relevant moving forward are shifting as well.
In recent years we have seen continued work on “conversion” to electronic representations, but we have also seen traditional document analytics applied to electronic representations themselves. For example, work has been published on indexing and retrieval of electronic sources from hard copy images as well as on the analysis of PDF and HTML layout structure. These problems are only a small part of a much larger need to extract information from material designed and organized for human, rather than machine, consumption.
This talk will highlight some of the current and evolving document analysis technologies that are required even in the offices of the future. These types of technologies won’t be rendered useless until we fundamentally change the definition of a “document”, which is not likely to occur nearly as quickly as some have predicted.
Brief Bio: David Doermann is a member of the research faculty at the University of Maryland College Park. He received a B.Sc. degree in computer science and mathematics from Bloomsburg University in 1987, and a M.Sc. degree in 1989 in the Department of Computer Science at the University of Maryland, College Park. He continued his studies in the Computer Vision Laboratory, where he earned a Ph.D. in 1993.Since then, he has served as co-director of the Laboratory for Language and Media Processing in the University of Maryland’s Institute for Advanced Computer Studies and as an adjunct member of the graduate faculty.
His team of researchers focuses on topics related to document image analysis and multimedia information processing. Their recent intelligent document image analysis projects include page decomposition, structural analysis and classification,page segmentation, logo recognition, document image compression, duplicate document image detection, image-based retrieval, character recognition, generation of synthetic OCR data, and signature verification. In video processing, projects have centered on the segmentation of compressed domain video sequences, structural representation and classification of video, detection of reformatted video sequences, and the performance evaluation of automated video analysis algorithms.
In 2002, Dr. Doermann received an Honorary Doctorate of Technology Sciences from the University of Oulu for his contributions to digital media processing and document analysis research. He is a founding co-editor of the International Journal on Document Analysis and Recognition, has the general chair or co-chair of over a half dozen international conferences and workshops, and is the general chair of the International Conference on Document Analysis and Recognition (ICDAR) to be held in Washington DC in 2013. He has over 30 journal publications and over 160 refereed conference papers.
Dr. Doermann is currently on assignment with DARPA, the innovative funding arm of US government where he runs programs on translation, speech processing, visual media reasoning and media forensics.