Comments

# Richard Edwards said on June 6, 2006 4:18 PM:

I like this very much!  The "narrow by" facets are nicely done.  It is very helpful that the facets themselves are shown  I still want the search results in RSS, but I'll bet that is coming soon.  Actually, I really want the results in RDF with all the metadata.  But meanwhile in the real-world I'll enjoy the new prototype.
Thanks! -red

# Rob Caron said on June 6, 2006 6:12 PM:

A much-improved MSDN & TechNet Search that provides some advanced search filtering to make life...

# stephbu said on June 6, 2006 6:14 PM:

Thanks Richard - all the regular "goodies" such as RSS and OpenSearch would come along for the ride in the finished product.

The fit and finish around the markup is pretty loose, we've not spent time optimizing markup - as much as anything else we were just playing with some of the data we get back from MSN, exploring ideas that seem more subjective i.e.

- "Include" vs. "Exclude" scoping.
- Forms of suggestions to help refinement such as keywords, categories, sources etc...

All feedback most welcome

# Richard said on June 7, 2006 9:04 AM:

What I would like is the ability to narrow based on technology or platform... by exclusion.

I am usually not interested in .NET Compact, so would like to see no hits from this.

When there are multiple versions of a technology I would like to easily be able to see which hits apply to which (e.g. seach for "BACKUP DATABASE" gives both SQL2k and SQL2k5 hits, but it is not obvious which is which).

# Troy Harvey said on June 7, 2006 10:52 AM:

Great job on this! I also really like the "narrow by" elements.

Throw a <label> on that exact phrase item:

<label for="exactPhrase">Exact Phrase</label>

# jandrick said on June 7, 2006 12:35 PM:

Thanks for the great feedback so far.  I hope more people will continue to give us feedback.  I want to address a couple of items.   We are workig on the keywords features mentioned above, we had this working in a prototype, but we wanted to make some changes before we deployed that feature.  We are working on product and technology filters as well.  Since this is protype we don't have RSS turned on yet, but it will be apart of a final release if not the next iteration. We are also working on XHTML compliance and accessability.

# stephbu said on June 7, 2006 1:06 PM:

Thanks Troy & Richard for trying out the prototype.  Great accessibility suggestion taken for the next cut.  

Let me also define what I think inclusion and exclusion mean.

For me I think of "exclusion" as - "the results from all sources except those who are members of <scope>" e.g. Exclude Blogs would mean results from all sources except Blogs.

Inclusion for me means "all the results from only those who are member of <scope>" e.g. Blog Inclusion means all results only from Blogs.

The "narrow by" interface currently uses inclusion, mainly because it makes the "winnowing" process feel quicker.  Moreover designing a meaningful interface for exclusion that shows what's in and out of the query is pretty difficult.  

On the flipside, exclusion can through fine-grain removal of small portions of the resultset yield really tough to find results that you'd otherwise have to page to find, e.g take out Blogs Source, MFC Topic, System.XML Namespace etc...  

I personally prefer include to exclude, quicker to narrow vs. more clicks and a tougher-to-read UI.

# kdeadwy said on June 9, 2006 5:36 PM:

I am a user of a Sharepoint Portal, and am confused by the search results I receive. Is there some place I can look to see how the various operators work? I am NOT someone who implements Sharepoint solutions, just a user.

I'm attaching a description of the confusing results I'm getting to help you understand what I'm seeing:

Looked in a document listed on the home page called 'White paper transition of the grid.pdf' from the Future Energy Systems landing page, and found the phrase 'renawable energy sources' (looks like a typo in the document).

In the portal search, entered renawable energy sources (no quotes around the phrase), and no documents were returned. I had expected to see AT LEAST that one document, and then any other documents that have ANY of those 3 words.

Then, I tried putting in 'renawable energy sources' using the single quote. Again, NO documents were found.

Then I tried it with "renawable energy sources" using the double quote.  Yet again, NO documents were found.

The only way I got that document to appear using the simple search was to use renawable+energy+sources, and then it found ALL documents with those 3 words in them, any one of the 3 words...

THEN, I went to the advanced search and entered renawable energy sources in the field for 'with the exact phrase', but no quotes around the phrase. NO documents found.

If I tried the exact phrase with single quotes ('renawable energy sources'), still NO documents found.

I had to put EITHER double quotes around the phrase ("renawable energy sources") OR pluses between the words (renawable+energy+sources) to have it return JUST the one document.

Why did the simple search return documents with ANY of the words, but the advanced search just the one document? In the simple search, I expected ONLY the document that had all 3 words in it, not any of the words.

# Niklas Nihl said on June 10, 2006 5:40 AM:

# TrackBack said on June 11, 2006 1:03 AM:

# TrackBack said on June 11, 2006 1:05 AM:

# TrackBack said on June 11, 2006 1:06 AM:

# SEARCH ENGINES WEB ♥ said on June 13, 2006 6:49 PM:

Console AND WriteLine
Console & WriteLine

are bringing up TWO slightly different SERPs now

# Leo said on June 15, 2006 1:16 PM:

Do you remember the old Altavista? they have done this ten years ago...I wish some helpful like using a wildcard *, %, like and so on...We spend a lot of time searching documents, more than working with them...

# Leo said on June 15, 2006 1:23 PM:

I am a database developer, in other words a search guru ho-ho-ho, it is possible to reproduce a search function like sql SOUNDEX? (obviously with the language parameter required). Also, remembering the old-school-altavista, there was another interesting function: NEAR...do you remember it?

# stephbu said on June 19, 2006 8:20 PM:

Leo thanks for trying out the experience, and commenting - we agree that advanced search operators aren't new to the Search world, but many operators are certainly new for MSDN Search.  (e.g. the "prefers" and "NOT" syntax)

Putting out an explanation of how those operators is also something that most engines never do.  We hope this is helpful to our audience.

We don't support NEAR or SOUNDEX-like functions at this point, but agree these would be great operators.  We will examine this for future iterations.

# steven said on June 27, 2006 11:00 AM:

Just some thought on controlling query scopes.

Short/general queries have actually been overwhelmingly dominating, which may be less true when working with collections of less variability like MSDN, but quite obvious for Web searches (e.g., many major search engines do not even support advanced search UI to form complex queries like [(a OR b) AND (c OR d)] because nobody is really using it).

The reason for this seem to be more complicated than merely users' tendency (or, stupidity :):
  - users are able to initiate a search with simple/short queries, but usually only over the course of the search will they be able to have a better idea about what is available for retrieval and to express their informational need using more specific/scoped queires;
  - however, as the user try to refine the query by using more conditions, other issues arise: for example, 1. focused queries often suppress other documents that may conclude the search quicker - those being directly relevant or providing hints on query refining - but do not satisfy the exact conditions, i.e., documents only appearing in the answer set for a shorter/broader query or other focused queries 2. unless a scoped query is precise and accurate enough, which can be very difficult to achieve in a couple of attempts to refine the initial short one, the quality of the answer set can hardly be improved enough to justify the effort spent in forming complex queries using the existing UIs;
  - besides, long queries do act quite "consistently" as obstacles when compared with short ones: for example, ranking of the answer set usually deteriorates severely as the number of terms in the query
increases (mainly because the existing statistical and text/title/anchor analysis techniques can only be effective in determining descriptive terms of the whole document at a very coarse level while complex queries are usually more detail-oriented).

Although short/simple queries and long/complex ones both have complementing advantages and disadvantages, they become conflicting with each other in the existing search engines because individual queries are isolated requests despite the fact that a lot of them can be related as the user refines the queries for the same search.

So simply augmenting the existing UI may bring very little improvement. For example, the "narrow by" functionality can be a very nice addition, but how can it work on a more varying collection than MSDN without clogging the result presentation, and is the "include" approach really preferable to the "exclude" one (aren't they also both desirable and complementing one another)?

Some other related questions may be:
  1. Should search engines' job be to announce documents' "availability" in their result lists (i.e., somewhere among the 15,520,000,000 of them) or to make the documents indeed reachable and recognizable by the users?
  2. Since our search engines all have become so automated and simple to use (and thus logically "prefer" short queries so much), are we entering an era of unanimity when
     - all people start to look at the same few "authoritative" pages when they search on the Web
       using the not-that-many commonly used keywords, circulatively asserting the pages'
       "importance" day after day;
     - while all the other "un-authoritative" ones start to fade out unnoticed before they can ever
       obtain the "important/relevant" status (or whatever the search engines have in mind)?
  3. How much extra time and frustration does the enforced sequential sifting of the result list cost, even when the questions are eventually answered?
  4. Does everybody need to be a ranking expert (and to be confined so much) before composing his/her masterpiece and putting it on the Web to "compete" in this age of "search-culture"?

All the problems might be solved or eased by putting related queries into a unified search session.

Sorry if this is not an appropriate place to discuss general search engines! Just that I didn't find similar topics on MSN search blog to join. :)

# The Hon. said on July 10, 2006 4:11 PM:

certified and asking?

New Comments to this post are disabled