Abstract
The composition of multimedia presentations is a time-and resource-consuming task if not afforded in a well-defined manner. This is particularly true when people having different roles and following different high-level directives, collaborate in the authoring and assembling of a final product. For this reason we adopt the Select, Assemble, Transform and Present (SATP) approach to coordinate the presentation authoring and a tag cloud-based search engine in order to help users in efficiently retrieving useful assets. In the first part of this paper we present MediaPresenter, the framework we developed to support companies in the creation of multimedia communication means. In the second part we describe how we adopt keyword search techniques coupled with Tag Clouds in order to summarize the results over the stored data.
Introduction: The Multimedia Authoring Bottleneck
Creating high-quality multimedia presentations in a corporate environment is often a logistical headache. Companies typically face a "lose-lose" scenario: they either outsource the work to expensive agencies (which creates bottlenecks and high costs) or they do it internally using tools like PowerPoint, which lack centralized asset management.
In the internal model, assets like logos, background images, and videos are often scattered across individual hard drives. Even when a central database exists, finding the right file is difficult for non-technical users who don't know the specific filenames or database schemas. There is a clear need for a collaborative, "agile" platform that allows different team members (designers, managers, presenters) to share, search, and reuse digital assets efficiently.
The SATP Workflow
The authors propose a framework called MediaPresenter, built on a 3-tier web architecture using the Software-as-a-Service (SaaS) paradigm. The system enforces a structured four-stage pipeline known as SATP:
- Select: Finding assets via keyword search and tag clouds.
- Assemble: Merging assets into slides or "series."
- Transform: Rendering the presentation into various formats (PDF, PPTX, SWF, PNG).
- Present: Displaying the final product via a web browser.
MediaBank: Intelligent Search and Tagging
The core innovation of the platform is MediaBank, a search engine designed to bridge the gap between complex database structures and human intuition. It employs two primary technical strategies:
1. Relational Keyword Search
Instead of simple table lookups, the search engine treats the database as a graph of related entities (primary assets linked to secondary metadata like tags or templates). To rank search results, the authors adapted the classic Term Frequency-Inverse Document Frequency (tf*idf) algorithm. They introduced a weight ($w_B$) that decreases as the "distance" between an attribute and the primary asset in the database schema increases. This ensures that a keyword match in a "Title" (close to the asset) is ranked higher than a match in a distant "Template Description."
2. Tag Clouds as Search Refiners
Rather than just being a "word cloud" for aesthetics, the tag cloud in MediaBank acts as a dynamic filter. Using a greedy "Maximum Coverage" algorithm, the system selects a set of tags that represents the widest possible variety of search results. Clicking a tag narrows the search results, guiding users through "hidden relationships" in the content they might not have initially considered.
System Implementation and Architecture
The system is implemented as a cross-platform, browser-based authoring tool using Adobe Flex and a Java-based backend (BlazeDS). By using XML schemas and JAXB (Java Architecture for XML Binding), the system can translate a single presentation into multiple proprietary formats (like PDF or PowerPoint) without being locked into a specific vendor's ecosystem. This decoupled exporting allows for high flexibility in how the final content is consumed.
Conclusion and Future Work
MediaPresenter successfully manages different user permissions and collaborative roles, allowing for efficient multimedia production. However, the current system relies heavily on manual tagging, which remains a bottleneck. Future work aims to implement automatic tag detection for documents and products, potentially using AI and semantic mapping techniques to automatically annotate titles and descriptions, further reducing the manual effort required to keep the MediaBank organized.
Summary updated