Monthly Archives: September 2005

You are browsing the site archives by month.

Question #3, Part 5 – Phrase Matching and Blog Searching

What can search engines could do better?

One person responded:

    * …I’ll do a search and get results with just one of the key words in it…e.g. if I did a search on Tamil Tigers, I would probably get results with just the word Tigers, which I’m not at all interested in.

    Comments: partial phrase matching is indeed an issue with search engines. Quoting searches can help with this, but can also overly limit the result set. Proximity can be a problem, as in pages that mention Tamil and Tigers, but they don’t mention them connected as a phrase (probably unlikely for this example, but you get the point). Phrase matching is really the key to relevancy. For most one-word queries it’s either easy enough to assume what the user is searching for (eBay), or ambiguous enough (jaguar) that a breadth of results is appropriate. But with a multi-term query you get into the world of concept matching, as in when a web site defines an entity as Navy Blue Shirt, but the searcher searches for Dark Blue Shirt.

    But maybe the bigger challenge is the challenge of “small mentions”. These days, with a two-word query like Tamil Tigers I think most results will match on both terms. However, what about pages that make only a passing reference to the term?

Another respondent wrote:

    * Update blog searches quicker.

    Comments: most blog search engines are near real-time right now, though I guess “near” is not real-time. I notice a lag sometimes between when I publish a posting on the Search Lounge and when it gets pushed to my RSS feeds.

Question #3, Part 4 – Natural Language

What can search engines could do better?

* Perhaps be better at responding to articulated queries, i.e. understand the semantics of a plain English query.

Comments: Natural language, one of search’s holy grails. Rather than building out natural language systems, most major search engines rely on users to learn a new syntax: the language of keywords.

Generations from now people will speak a different language because of this. Instead of whole sentences people will use phrases. Instead of saying to a friend, “What is the largest city in Brazil?”, people will say “Brazil largest city.” Or instead of saying at a cocktail party, “What are the ingredients in a Sazerac?” people will say, “Sazerac ingredients list?”

Question #3, Part 3 – Official Sites and Language Settings

What can search engines could do better?

**Official Sites**

The respondent wrote:

    *Always list the official site first. (So if I do a search for Nabisco I want nabisco.com listed first, not several articles about Nabisco).

    Comments: A few years ago search engines weren’t very good at always getting the official page in the first position, but now they generally are. But what if users click more often on the non-official page because the official page isn’t the best source?

**Language Settings**

The respondent wrote:

    * Ask me for the languages I am able to comprehend since links that are mostly written in Czech, Chinese or Icelandic do not really help me.

    Comments: most engines do let users set this preference, but it’s definitely not foolproof. Results in other languages still show up often enough to be noticeable.

Question #3, Part 2 – Tips and How-Tos

What can search engines could do better?

A few people wrote they want search engines to explain more about how they work. Although I understand this sentiment, do people really want that?. What I’m thinking they want is to get better results; and if they’re not getting good results they think maybe they’re doing something wrong, but they don’t know what. So rather than search engines going into the minute details of ranking algorithms and so forth, when users don’t get good search results the engines should step in and offer ways to improve their search.

But I think this issue addresses a serious topic. Namely, search engines are really complex systems, but people want to be able to enter a couple of words in the search box and get exactly what they want. How do search engines balance on the tight line between sharing information on how search works, while not overwhelming people with too much info?

Here are some of the responses on this topic:

    *Offer more tips on searching.
    *…providing me with a better understanding of how they [search engines] use the inputs to conduct a search would help. This info is no doubt already available, but I’m usually in convenience mode when doing searches and unless this info is up front, I usually won’t take the time to go searching for it.
    * What I wish there was more (maybe there already is and i don’t know how to find it) is a “how to…” search engine. In my case, for arts and crafts, but it could be for anything else.
    *Provide more intuitive search refinements than boolean language, such as dropdown menus and check boxes with refinements explained in common English.

Question #3, Part 1 – Providing Answers on Search Results Page

What can search engines could do better?

I got 12 real responses to this question, which is more than I expected since it’s a tough question.
I’m going to break out the answers into separate, small posts, starting with…

    *Add direct reference data. e.g. “when is labor day?” pulls up result: “September 5, 2005″ .

    I love it when the SERP answers my question and I don’t have to click through to the page! Like the way Brainboost does it. Engines are experimenting more and more with this concept, like Instant Search.

Post Navigation