Chapter 4. Searching the Multimedia Warehouse

This chapter covers how to perform search as and queries against digital objects in a multimedia warehouse. It includes the different ways a multimedia database can be searched and the different types of search that can be performed. It will cover how to interpret and understand what is returned, as well as how to intelligently display the results that come back.

Multilingual data

Not all countries in the world or organizations use one language. Some are bilingual, others support multiple languages. Data coming in from other sites might be in a different language. Museums around the world regularly share items, and the digital information relating to them might be in a different language.

The challenge for all these organizations is how to handle this data in the best way. The issues can be broken down to the following:

• Storing

• Displaying

• Translating

• Searching

All spoken languages can be represented in text; some phonetically, others in a representative structure. Different languages use different symbols to represent the pronunciation. Some languages use one symbol to represent a sound (phonetic), others use a mixture of one or more to represent the different sounds. Languages such as, English have introduced a large number of exceptions to this, so much that dictionaries need to include a pronunciation guide. English also has multiple symbols representing...

Security

The most efficient way of controlling which users should be allowed to access images is best done via roles. A role is a very simple concept but a lot of people who have never used them struggle to understand what they do.

A role enables fine-grain access to a set of images. It enables security to be configured based on the business requirements. Roles also enable security to be changed very quickly and easily.

With security one has to always remember that it's a balancing act. The more restrictive and tighter the security, the less chance there is of unauthorized access. The tighter and more restrictive the security is, the harder it can actually be for a user to do their job, to perform queries, or do things quickly. The less security, the greater the chance of a digital object being stolen or damaged. One goal of a warehouse is to allow users to do ad hoc and complex queries without the burden or restrictions of an OLTP environment. Each organization is different and has their...

Searching

When it comes to searching, most people have a Google or Bing mentality. This means they have been inadvertently trained in search behavior because of their extensive usage of these search tools. They are used to typing in a word or two and have the results come back in a ranked order. This type of search is a course grained search. It is effective when you are blindly reaching out trying to find a match. Users then look at the results and narrow down their search manually by checking the results to see if it was what they were after.

This type of searching is very effective when the information is spread across a large number of systems. It is not effective when more refined or fine grain search methods are requested. These search systems are also based around the concept of HTML page being the only type of digital object being searched on.

A number of these search engines struggle or just cannot handle fine grained queries. If a user wanted to find all images that were portrait...

Summary

Searching within a multimedia warehouse requires an understanding of fuzzy concepts and being able to effectively visualize the results. An effective search engine for querying multimedia needs to factor in the need for course grain and fine grain searching. As most digital objects are described using metadata, the search engine needs the capabilities to query against it. In addition, the search engine needs features to perform queries specific to the attributes of the type of digital object. Most search engines which have a relational data centric focus, lack the capabilities to do enable these advanced querying features. The search engine also needs to be able to adjust its search behavior based on the data types of the metadata. This includes being able to do fuzzy searching against circa data and intelligent querying against accession and object numbers.

Exercises

1. Design a metadata structure that supports searching on video.

2. Design a search algorithm for searching on animal sounds in audio files.

3. Design a search engine for researchers that enables collaboration, in that different researches can contribute to the search results and share that information between other researches. New results added are highlighted and their reasons for addition are included. The search engine might involve finding all historical information on a particular painting, meaning the search engine has to work across multiple sources.