Chumki Basu

Technical paper recommendation: a study in combining multiple information sources

The growing need to manage and exploit the proliferation of online data sources is opening up new opportunities for bringing people closer to the resources they need. For instance, consider a recommendation service through which researchers can receive daily pointers to journal papers in their fields of interest. We survey some of the known approaches to the problem of technical paper recommendation and ask how they can be extended to deal with multiple information sources. More specifically, we focus on a variant of this problem - recommending conference paper submissions to reviewing committee members - which offers us a testbed to try different approaches. Using WHIRL - an information integration system - we are able to implement different recommendation algorithms derived from information retrieval principles. We also use a novel autonomous procedure for gathering reviewer interest information from the Web. We evaluate our approach and compare it to other methods using preference data provided by members of the AAAI-98 conference reviewing committee along with data about the actual submissions.

Learning to personalize

102 August 2000/Vol. 43, No. 8 COMMUNICATIONS OF THE ACM question in the design of such self-customizing software is what kind of patterns can be recognized by the learning algorithms. At one end, the system may do little more than recognize superficial patterns in a single user’s interactions. At the other, the system may exploit deeper knowledge about the user, what tasks the user is performing, as well as information about what other users have previously done. The challenge becomes one of identifying what information is available for the given “learning to personalize” task and what methods are best suited to the available information. When I used the email program on my PC to forward the file of this article to the editor of this magazine I executed a series of actions that are mostly the same ones I would take to forward any file to another user. I typically click on an item on a menu that pops up a window for the composition of an email message. A fairly routine sequence of actions then follows—I compose the message, select a menu item that creates a pop-up window into which I enter the name of the desired file to be forwarded, finally completing the LEARNING to Personalize Haym Hirsh, Chumki Basu, and Brian D. Davison

Telcordia LSI Engine: implementation and scalability issues

Latent Semantic Indexing (LSI), a vector space-based approach to information retrieval, has been proven to be an effective tool in correlating and retrieving relevant documents. While much work has been published on LSI, most of it addresses the algorithmic or theoretical basis of the model. Little, if any, presents implementation issues in practice. We describe a production-level implementation of LSI. The system integrates components including document collection and preprocessing, singular value decomposition (SVD), multilingual processing, and a tree-based access method for similarity querying. We discuss implementation issues encountered during the development of the system. In particular, we address scalability issues in the query engine and various components of the system, and present lessons learned.

Putting legacy data on the Web: a repository definition language

Abstract The objective of InfoHarness is to provide integrated and rapid access to huge amounts of heterogeneous legacy information through WWW browsers. This is achieved with the help of metadata that contains information about the type, representation, and location of physical data. The proposed InfoHarness Repository Definition Language (IRDL) aims to simplify the metadata generation process. It provides high flexibility in associating typed logical information units with portions of physical data and in defining relationships between these units. The proposed stable abstract class hierarchy provides support for statements of the language that introduce new data types, as well as new indexing technologies.

ViTex: Video To Tex and Its Application in Aerial Video Surveillance

Given the huge amount of aerial surveillance video, captured daily, an automated video understanding system is needed to extract information and to generate metadata that is easy to search, browse and summarize, and which can be readily understood by an end user. In this paper, we propose a Video-To-Text engine called ViTex that automatically generates text descriptions of the content of a video. The ViTex engine first segments an input video sequence according to pre-defined semantic classes using a Mixture-of- Expert blob segmentation algorithm. The resulting segmentation is coerced into a semantic concept graph and based on domain knowledge and a semantic concept hierarchy. Then, the initial semantic concept graph is summarized and pruned. Finally, according to the summarized semantic concept graph and its changes over time, text descriptions are automatically generated using one of the three description schemes: key-frame, key-object and key-change descriptions. We have applied the ViTex engine to aerial surveillance video and compared its performance with ground-truth text descriptions generated by humans.

The Geospatial Interoperability Problem: Lessons Learned from Building the Geolens Prototypye

In 1994 NASA issued a Cooperative Agreement Notice to support new research on digital library technology that would enable broader public use of its Earth science data over the Internet. As a response to this CAN, the Universal Spatial Data Access Consortium (USDAC) was formed and it proposed to prototype the GeoLens system that would not only give broader public access to NASA’s Earth observation data, but also made these data interoperate with other geospatial data served by the Federal government. Part of the challenge of the GeoLens Project has been to decompose the larger geospatial interoperability problem into constituent issues. This chapter will address these issues and describe solutions implemented in our GeoLens prototype. The purpose of this exercise is to support an end-to-end scenario, beginning with geospatial data discovery and ending with conflation of geospatial data extracts from extremely heterogeneous sources. The larger goal is to investigate the opportunity for new information processing standards and innovative digital library technology to play a key role in the realization of this scenario.

Mining Spatial Associations with Limited Sensory Information

Abstract Human navigation in an unknown environment requires an understanding of the spatial relationships of the terrain. For example, a soldier who is on a reconnaissance mission in a new city needs to “know” the spatial layout of the surroundings with high confidence. Oftentimes, this understanding must be acquired within a very short amount of time and with limited sensory inputs. The soldier would benefit from a digital avatar that draws inferences about the spatial layout of the city based on an initial set of observations and guides the soldier either in further exploring the environment or in making decisions based on these inferences. In this paper, we present and evaluate an inductive approach to learning spatial associations using sensory data that is available from the simulation environment of a computer game, Unreal Tournament. We study two kinds of spatial relationships between nodes on a level of a game map: nodes that are placed near each other to satisfy some spatial requirement and nodes that are placed near each other to satisfy the design preferences of a level architect. We show that we can infer both kinds of relationships using an association rule mining algorithm. Furthermore, we show how to use an ontology to distinguish between these relationships in order to discover different types of spatial arrangements on a specific map. We discuss how the inferred associations can be used to control an avatar that makes recommendations for navigating unexplored areas on a map. We conclude with some thoughts on the applicability of our methods to scenarios in the real world, beyond the simulation environment of a game, and on how the learned associations can be represented and queried by a simple question-answer type system.