15
Search Update April 1-3, 2009 Joshua Ganderson <[email protected]> Laura Baalman <[email protected]>

Search Update April 1-3, 2009 Joshua Ganderson Laura Baalman

Embed Size (px)

Citation preview

Search UpdateApril 1-3, 2009

Joshua Ganderson <[email protected]>Laura Baalman <[email protected]>

Google Funded Image Search: Updated User Interface

Addition of Viking 1 and 2

Addition of HiRISE Images

View Results in Google Earth Mars Mode

Help for View Results in Google Earth Mars Mode

Google Funded Image Search: In The Works

• Planetary Content Group at Ames

• Adding Data - CRISM, CTX, and Themis

• Exploring New Options – Recent meetings with Google staff and consultation with application designers for ways to better provide access and functionality

Google Funded Image Search: Under Discussion

Met with sponsor at Google and were encouraged to focus efforts on user contributions and collaboration

• User Contributions - User assisted ranking, tagging, error reporting, and layering of more free form user data

• Outside Data – Continue to leverage other data sources for improved search and data quality

• More Data – Continue to add missions

• More Collaboration – Determine who else is doing related work and try to centralize repository of metadata

Why Search As Next Prototype?

1.Identified Area of Need – Data retrieval a primary function of the PDS, improving search is a obvious value add

2.User Research + Requests – All user research and a number of suggestions focus on improvements or refinements to the search process. Specifically, SBN requested an implementation of product level search.

3.Fragmented Tools and Knowledge – There are a number of tools and repositories for information that are not synchronized and duplicate effort

4.An Experiment in Leverageability – An entirely unified search is unlikely and the expertise on each type of search rests across the nodes. However, centralizing metainfo and providing an api for a variety of searches might meet a high % of need.

Generalized Product Level Search: Purpose

• Enable Product Level Search – Very limited product level search exists. Specific to type of search and a limited portion of PDS data. Need to support search based on product contents and characteristics.

• Simplify Search – A simple textual search may provide good enough results and lower the bar for retrieving data

• Make Search Flexible – Simple search not good enough in many cases so more complex and structured queries must be supported

• Centralize Index – To support consistency and correlation, searches are best conducted against the same metainfo

• Join Non-PDS Data – As with Google funded image search, outside sources of information can enrich meta information. Allows for better search and can augment data able to provide to the user.

Generalized Product Level Search: Implementation

1. Crawl - Crawl all hosted PDS data and node websites

2. Parse – Parse any labels and associate metainfo with product or volume. Also extract pertinent info from product if applicable.

3. Simple Then Complex - Start with simple text but support faceted and ranged search on specific fields

4. Join - Pull in outside data and augment metainfo

5. Transform - Allow for volume, mission, or instrument specific transformations

6. User Contributions – Feedback, comments, error reporting, visualizations, ranking, etc

7. API - Allow access as a web service to allow one-off and very specialized search tools

Generalized Product Level Search: Exceptions

• Not One Size Fits All - Understand that types of search will not fit model

• Bang For Buck - Intended to fit largest portion of search with least cost in development

• Libraries To Support Exceptions - As stabilized, portions might be turned into libraries to facilitate specialized search

• Central Metainfo DB – Even if flexible search doesn't meet needs, the metainfo collected should still be a valuable resource for any custom tools nodes need to develop

Generalized Product Level Search: 2010

Regardless of what happens with the 2010 spec and supporting tools developed by the EN, this will not be a wasted effort.

•Easily Updated – The 2010 data should have more accessible metainfo. Integrating the processing of this type of data should be simple.

•Serves an Immediate Need – We're responsible for providing the best access we can to the data we have. PDS3 data not going away.

•Educates Search – The meta info required to search doesn't change. Any new search tools should be easier to develop using this example.

•Search Design Continues to Be Valid – Regardless of what the underlying functionality is, much of the search interface will still be valid. This can be recycled into or layered over new search tools.

•External Pipelines Still Viable – If we implement feeds from outside repositories of information, this will still be worthwhile in any new search tools.

Generalized Product Level Search: Status

• In Progress – Have proof of concept doing crawl, label parsing, metainfo/product association

• EN Meeting - Ames and EN face to face next week

• User Collaboration – Next steps once ingestion process is complete is to involve users in metainfo choices and ordering discussions

• Prototype Availability – Estimating 6 weeks

Questions?