Home

Multi-Media Information Retrieval

This section provides more details and guidelines about submissions to SIGIR'99 in the area of Multi-Media IR.

As increasing amounts of non-text material are being stored on-line, providing efficient and effective ways of finding and accessing these materials is becoming a critical problem. We are soliciting original contributions that address the information retrieval questions raised when substantial quantities of images, graphics, speech, audio, and video are to be stored on-line and made available for search and retrieval.

The Multi-Media Information Retrieval Theme of SIGIR'99 specifically addresses the overlap between multi-media data and information retrieval, and so we discourage contributions that focus purely on algorithms for processing multi-media data without regard for the information retrieval problems being addressed. We also discourage contributions that focus only on the system and software aspects of storing and managing large amounts of these data types.

We welcome high-quality papers that describe evaluation experiments, case studies, and theoretical analyses. We especially encourage submissions from researchers not traditionally part of the SIGIR community who are nevertheless doing significant information retrieval work. Topics of interest include, but are not limited to:

CONTENT-BASED INDEXING STRATEGIES

• What kinds of indexing strategies can be developed for non-text material?
• Do non-text materials have natural sets of features that are good for IR applications?
• How do we process multi-media data to extract these features and build efficient search indexes?

QUERY FORMULATION AND QUERY LANGUAGES

• What kinds of questions do we ask of collections of non-text material?
• Are these different in kind from questions for text only collections?
• What does this tell us about query language syntax and semantics for multi-media?

CROSS-MEDIA AND MIXED-MEDIA RETRIEVAL

• How can we develop systems that support cross-media retrieval using a single form of query?
• What kinds of techniques do we need to search mixed-media data environments (e.g., word-processor documents with embedded images and charts)?

RESULTS ANALYSIS AND PRESENTATION

• How do we generate retrieval status values for multi-media data?
• Does the notion of relevance change in multi-media environments?
• What constitutes the basic unit of retrieval in a multi-media data environment?
• How do we combine results from different media sources into a single coherent answer to the retrieval question asked?

TEST COLLECTION DEVELOPMENT AND EVALUATION METHODOLOGIES

• What are the challenges in developing multi-media IR test collections?
• Are there any test collections that could be used today?
• How can we measure the performance of multi-media retrieval systems?
• Do conventional measures of precision and recall apply to media like video?
• Do we need alternative measures of effectiveness?