|DB meeting:||Friday, May 11th, 2:00 pm, DC1331|
|Topic:||Modeling Video Data
With widely varying use of video data, the simple VCR functions can no longer satisfy users' requirement. Techniques for content-based video data modeling, indexing and retrieval have been developed to address these demands. However, efficient modeling of video remains a research topic. Compared to text or image data, video data is more complicated, because of its temporality and high semantic charateristics. In this presentation, shot based, annotation based and video object based modeling techniques are reviewed and compared.
|DB seminar:||Monday, May 14, 11:00AM, DC1304|
|Speaker:||Peter Buneman, University of Pennsylvania|
|DB meeting:||Friday, May 18th, 2:00 pm, DC1331|
|Topic:||Data Management issues in emerging P2P network programming platforms,
such as JXTA (see
http://www.jxta.org). JXTA is defined as a set protocols, which use XML-encoded messages, designed to be independent of transport protocols. It can be implemented on top of TCP/IP, HTTP, Bluetooth, etc.
|DB seminar:||Monday, May 28th, 11:00AM, DC1304|
|Speaker:||Klaus Meyer-Wegener, Technical University of Dresden|
|Title:||Combining Data Independence and Realtime in Media Servers|
|DB meeting:||Friday, June 1st, 2:00 pm, DC1331|
|Topic:||I will talk about the idea Grant and I had some time ago about storing/querying XML in an ordinary rdbms; what works and what doesn't (and speculations on why).|
|DB meeting:||Friday, June 8th, 2:00 pm, DC1331|
|Topic:||Optimizing temporal queries: efficient handling of duplicates
The SQL/TP language allows users to compose queries against an abstract database, and then translates these queries to SQL for execution on a commercial RDBMS. The translation is defined to provide SQL duplicate semantics. By considering the context of translation, we are able to use a more efficient translation when the number of duplicates does not affect the correctness of the result.
|DB seminar:||Monday, June 11, 11:00AM, DC1304|
|Speaker:||Jarek Gryz, York University|
|Title:||Query Optimization via Data Mining|
|DB meeting:||Friday, June 15th, 2:00 pm, DC1331|
|Topic:||The secrets of getting precise answers when querying incomplete
I will show how to do bag relational algebra with aggregation over a particular representation of incomplete information called c-tables (Grahne '84). The c-table representation is an extension of the Codd relational model with richer semantics for null values. The reason c-tables were chosen for the exploration is that they are the least expressive relational representation of incomplete information over which relational algebra is closed and can be "well defined". The need for duplicate preserving relational algebra over this representation of incomplete information is justified by introducing aggregation over c-tables with duplicates.
|DB meeting:||Friday, June 22nd, 2:00 pm, DC1331|
|Topic:||Word Spread as an Indicator of Text Category
Traditional information retrieval has generally viewed text as an unordered collection of words. This so called "bag-of-words" view is sufficient for many retrieval tasks including query matching and document clustering. However, by definition the bag-of-words view lacks insight into the flow of information in the text. In this talk I will present an alternative view of text that retains some of the topic flow information by preserving relative word positions, and describe our ongoing research on how this information can be used to automatically categorize text.
|DB meeting:||Friday, June 29th, 2:00 pm, DC1331|
|Topic:||I will survey some of the literature on dynamic histograms - that is, histograms that can adapt automatically and incrementally to changes in the underlying data distribution.|
|DB meeting:||Friday, July 6th, 2:00 pm, DC1331|