Trip Database Blog

A fairly productive meeting with Sequence has moved the snippets idea forward. The underlying search software we use has an auto-snippet function – which is a good start. So Sequence will create a testing system to demonstrate how it would work in reality. Assuming that’s good we still need to decide how to display it. Do we take up screen ‘real estate’ and display it like Google? Alternatively, we only display it via a roll-over function?

The other snippet function was the ability to grab the documents conclusions; where they exist. This seems ‘do-able’.

Assuming things go smoothly we should have something to rollout by March.

I ended up e-mailing the authors of the ‘Eye tracking in MSN Search…’ to discuss a few ideas for TRIP. I suggested a hybrid method whereby a snippet (short extract of text taken from the main document) is available but only visible via a rollover. So you retain the number of results per page and users can easily see the snippet – best of both worlds?

An interesting reply (very pleased they did reply!):

In previous work, we found that the cognitive cost associated with deciding that you need to look at the expanded abstract usually outweighs the benefit associated with presenting more results on a page; the rollover disclosure stymies users’ ability to quickly scan the results to find what they’re looking for.

I’ve taken that to mean that the extra time taken to decide to look at the extra snippet information would be better spent scanning more results.

I still think the rollover is better for TRIP, for two main reasons:

• We have less screen ‘real estate’ due to the categorisation system, so space is even more precious.
• TRIP is a specialised not general search engine. Users are likely to know more from the title and source in TRIP than in MSN search

So, I’m off to Sequence (our web people) to discuss the options. I’m actually favouring a ‘two snippet’ approach. Snippet one will be the first 5-7 lines of text and snippet two will be the conclusion of the article. As many articles don’t have conclusions (e.g. guidelines) many records will only have ‘snippet one’.

Sequence are very good at taking my ideas and working them into something tangible. No doubt more tomorrow on this idea!

A recently published paper has been a bit of an eye-opener: “Eye tracking in MSN Search: Investigating snippet length, target position and task types“

It examines how people use search and highlights two main types:

• Navigational – where a user knows where they want to get to (e.g. a hotel, an organisation) and use the web to find the homepage
• Informational – where a user has a question and needs to use the web to find an answer.

Form the above two descriptions I would see TRIP as being almost exclusively in the informational sector.

The researchers methodology used eye tracking to see where people looked on a search page and the effect of different variables on the search speed and accuracy. One feature they examined was the length of a snippet. The snippet being (typically) the 1-2 lines of text that typically appear in search engine results below the title. NOTE: TRIP does not have a snippet feature.

Without going on too much (read the article for more info) they found that, in informational tasks, the longer the snippet the quicker and more accurate the search. The main downside being that the longer the snippets the more users had to scroll to see the results.

Makes me think that we need to introduce a snippet function. Is this ‘Evidence-based search’?

We’re preparing ourself for the next round of improvements on TRIP. We’ve got some ideas including:

• Use RSS feeds
• Introduce more web 2.0 ideas (allow people to comment on articles, rate them etc)
• Try and grab the conclusion, clinical bottom lines from articles so they can be viewed from within TRIP – as opposed to clicking on the link.
• Enhance the search algorithm to further improve it.
• An Amazon style feature for each entry with something like “people who looked at this article also looked at….”
• Improve the search relevancy

However, these are just some thoughts we’re discussing. Here’s your opportunity to help decide where to take TRIP. Just let us know what you’d like to see, new features, improvements etc. and we’ll see what we can do. Contact us via this blog or the TRIP Database contact form.

We look forward to hearing from you.

An interesting post in BMJ Career Focus (click here). Why am I highlighting this – because networking will form part of Gwagle…..

The Clinical Cases and Images blog has highlighted the rather nice schmedley site.

Interesting pattern shown over December. Below is a chart of our weekly search stats. At the start of the graph we’d been free for 3 weeks and were receiving around 25,000 searches per week. This rose and peaked at the end of December (at 80,000 searches per week), then started to drop, the low point being 27,000 searches around Christmas day. Since then the searches have picked up and are as good as at our peak number of searches.

People distracted in December – surely not!

Reported here: A Harris Interactive poll conducted for Yahoo Local found that 67 percent of respondents said they would be likely to post an online rating or review, and 79 percent of respondents said they’d be likely to be influenced by such a rating or review.

This highlights the rise – is it an expectation – of web 2.0. Gwagle, hopes to use this willingness to share experience when it launches.

A collection of educational medical videos for students and doctors click here.