Whole words only

Tuesday, March 16, 2010

Enhanced history command

The search engine history command allows you to put a most-popular searches list on your website. Based on a customer request, the search terms in the most-popular list are now active links; that is, each term is a link that, when clicked, runs that search against your index. To try out the history feature, follow the instructions in the Search Guide under Activity Reports.

Friday, March 5, 2010

Try out our next generation search engine

We are beginning final testing of the next generation Blossom search engine. The new engine has a better understanding of English so that it can generate improved page extracts and so that searching is more precise. It is also able to separate common navigation text from page contents. Let us know if you would like to try out the new engine on your search index by sending email to Blossom Support.

Thursday, February 25, 2010

Targeted incremental updates for fast turnaround

There are now three ways to update a search index:
  1. Automatically. Periodically our spider visits your website and looks for new pages. The process is fully automatic. Normally you should rely on this automatic update.
  2. Manually. From the Search Configuration page at Blossom.com, you can trigger a complete rebuild of your index. This makes sense after a major reworking of a site. To trigger a rebuild, follow the "Update an index" link and click on "Full Rebuild".
  3. Incrementally. When you add new pages to your site and you'd like them to appear quickly in your index, you can perform an incremental update where you tell us where the new pages are. To trigger an incremental update, follow the "Update an index" link and click on "Incremental Update". You'll be asked for a file containing the URLs of the new pages.

Friday, December 11, 2009

Customizable entities on the way

A significant upgrade of the Blossom search system is in development. One new feature is a facility for defining custom entities. The search engine has several entity types built in, such as phone number, email addresses, and time of day. An entity consists of two parts: a trigger and a pattern. When a trigger is detected in a query, it can be replaced by text that matches the pattern. For example, a search for "office phone number" would match not only the words "office phone number" but also the word "office" near a phone number, like "office: 888-123-4567". The custom entity facility will allow site designers to define their own triggers and patterns. More details will be provided when the facility is available for testing. In the meantime, to help us design the facility, we are interested in hearing about how you might use custom entities. If you think you have a use for them, please send a note support@blossoft.com.

Wednesday, October 14, 2009

List of image-only PDF files

Some sites have PDF files that don't contain text; often they are generated from scanned documents. The lack of text makes the files invisible to the search engine. To reduce the number of times image-only PDF files are downloaded, a list is kept for each index. You can now download that list from the Search Configuration page. Follow the link "Retrieve the list of image-only PDF files being ignored" in the "Search Index Settings" section.

The list is cleared whenever a complete respidering of the index is triggered, such as when you request a non-scheduled index update (by following the "Update an index" link on the Search Configuration page).

Thursday, October 8, 2009

New option for database-driven sites

Some websites have pages generated dynamically from a repository. Often on these sites there are multiple URLs that can generate the same page. Blogs are an example; an article may have multiple labels, each label providing a path to the same text. For example, www.domain.com/july/article1 and www.domain.com/announcements/article1 might both refer to the same article.

A new option has been added to the Blossom spider telling it to ignore all but the last component of a URL when determining whether two URLs are the same. Thus in the example above, "article1" would only be retrieved once. This will reduce the number of duplicate documents downloaded from a site, saving both bandwidth and potential page count.

Contact Blossom Support if you think your site might benefit from using this option.

Sunday, April 19, 2009

Spaces in URLs

Over the past few weeks we've been experimenting with different rules for handling space characters in URLs. Technically, spaces are not allowed in URLs; they should be encoded as %20. Nevertheless, many websites use spaces in file names.

We looked at treatment of spaces in common browsers. In every case browsers translate interior spaces to %20. The treatment of leading and trailing spaces varies. The most forgiving behavior is to remove them. Thus we now remove leading/trailing spaces, but keep interior spaces. Here are a couple of examples to illustrate:

<a href=" leadingSpace.html"> is treated as <a href="leadingSpace.html">

While

<a href="interior Space.html"> becomes <a href="interior%20Space.html">