Search by phrase and search with intelligence/planning

[lmnowels]

RoboHelp HTML 8 - Generating as WebHelp

When I try to enter "upgrade source," I only want the results that include that phrase, not any other combination of the two words. I need my users to get topics that relate best to the search word or phrase that was entered. If they ask for a certain word or phrase, I want the first result to be the one most likely to answer their question. I was planning to spend the time to set it up properly to make this happen.

But, as of right now, my search results appear to be almost random - just a mush and sometimes even excludes the topic that included the optimum content that the user was looking for.

I've spent time trying to set topic properties to include the optimum Index Keywords and setting the Keywords to Exclude from Search. I'm including my quotation marks in the search.

What am I doing wrong? Can you point me to a resource that can help me get the results I need? I appreciate your help.

[johndaigle · Accepted solution]

Hi, lmnowels.

WebHelp's Phrase search was first introduced with RH 8. As you've noticed, by placing two or more words of a phrase in quotes, that phrase will be favored in the search ranking and the topics returned will only be ones where that exact phrase appears at least once.

For example, in the sample project that ships with RH 8 called Web Graphics, I tested the phrase "graphic formats".  If I search without quotes, I get 16 topics for results. If I enclose the phrase in quotes I only get the 5 topics where the exact phrase appears. So, phrase searching is very helpful to a point. But it is not perfect. The downside is that search highlighting feature not only highlights the phrase instance, but individual words that make up the phrase. To my knowledge I don't think this can be filtered to show only the phrase instances. (This would be nice improvement for RoboHelp "Next")

Here's a review of how search results are ranked (in this order) which might help in your design.

1. Topics with the search query words in the topic title are ranked highest
2. Topics with the search query words that have been manually added by the author in the Search Keyword dialog are strongly favored. (Topic Properties > Keywords) Think of this as a synonym at the topic level (instead of the whole project)
   
3. Topics with the search query words in headings H1, H2, H3, H4 and so on are given greater weight.
   
4. Topics with the same ranking are sorted alphabetically.

One thing I'm confused a bit with your post is the comment about "setting the Keywords to Exclude from Search". These are actually two separate things. Whatever weight you give by creating a search Keyword, will be negated if you also check the box "Exclude from Search" as this refers to excluding the entire topic from coming up in the search results. Or, perhaps I didn't understand your post.

Another tip that folks often forget is that RH 8 offers the option of adding project Synonyms. This is done from File > Project Settings > General tab > Language > Advanced > Synonyms. (e.g., if someone searches for images, they will also get topics with the word graphics.

I created an Adobe Captivate movie which outlines the new search features here:

http://www.adobe.com/designcenter/robohelp/videos/

Look for the one on search.

Of course, we as authors are always demanding better search for our end-users and Adobe continues to improve the experience. Meanwhile, if you are looking for other options/enhancements, you might check out Peter Grainge's favorite "add on" called ZoomSearch.

http://www.grainge.org/pages/authoring/zoomsearch/zoomsearch.htm

Hope this helps.

John Daigle
Adobe Certified RoboHelp and Captivate Instructor
Evergreen, Colorado
303-679-08955
http://www.showmethedemo.com

[Jeff_Coatsworth]

Hi John,

Have you ever heard of anybody ever using that custom Google search on their WebHelp output? I wonder how hard it is to add to a project or just to the output pages.

[johndaigle]

I am vaguely aware that some folks have tried it successfully, but don't have a clear recollection of who it was or what issues there might have been. Perhaps it would be some process similar to ZoomSearch which Peter Grainge fancies.Or Rick Stone may have more info on this. Perhaps they will chime in.

(I did search the forums for "RoboHelp Google" and the results seemed to be more about Google Analytics rather than using Google's search engine as a substitute.)

John

[Jeff_Coatsworth]

Yes, that's what I remembered seeing too...

[lmnowels]

John, Thank you so much for your careful reply and your advice.

The most helpful information you provided was the ranking of the search results to help move the topics to the top. I will want to do this for the topics that are most likely to answer the questions, "What is it?" "How do you do it? (Steps)" "What cautions relate to it?" "How does it affect other processes?" ...(preferably, fairly consistently in that order)...

I had already carefully entered index terms and associated them with topics. Initially, I believe I tied them to more topics that was optimum. Sometimes less is better.

I had misunderstood the "Exclude from Search" checkbox. I thought that feature would exclude topics based on the Keywords in the field just above the box. For example, my topics all end with "Did that answer your questions? See also:  ... [hyperlinks to other topics] ..." (I had wanted to exclude the topics if the search term reflected words that simply labeled those hyperlinks to other topics.) But, I can see that the Exclude from Search box will always exclude that topic from all search. That's not very useful for my application. (I plan make recommendations for future enhancements.)

I believe one reason why I was feeling uncomfortable with the results is that the highlighting covered more than I was looking for. You described this well in your response. The search for "upgrade target," highlighted upgrade, target and upgrade target. The search for "old" highlighted old in folder, sold, and old. (I had already unchecked "Enable substring search.") So, the highlight cluttered the output, giving the impression that the search was not very selective. For now, I've turned off the highlight feature, but there are pros and cons. Some users will be unhappy either way. (I plan make recommendations on this feature also.)

I have not asked my company to add ZoomSearch to my PC system. I work for a large company. I do not have administrative rights and the IT group requires testing (by IT), approvals, etc. before any software can be loaded. I was hoping that I could accomplish my objectives through RoboHelp features.

I watched your Captivate training on Search and I will complete all of them.

I am making progress toward my goals to provide useful and efficient search to avoid frustrating the users. I appreciate your help very much.