We have a brand new look! Take a tour with us and explore the latest updates on Adobe Support Community.
Forgive me if this is brief in detail, but as is often the case, I have inherited a project I know very little about.
My organization is in the process of moving a help system for an enterprise level proprietary computer system from RoboHelp Classic to RoboHelp 2019.
The content and structure of the help topics is virtually identical in both systems. However searching is producing markedly different results.
In the old system (RoboHelp Classic), I can search for a help entry by title and it is the first item returned. When I perform the exact same search in RoboHelp 2019, the results are wildly different, and not at all helpful, and almost never contain the entry that I searched for.
Can anyone point me in the right direction? Is this an indexing issue? Or are there search settings in RH 2019 that I need to tweak? Any assistance would be appreciated as I am stumbling around in the dark at the moment.
The problem happens if the project is local or hosted. And it began when we migrated robohelp from the old version to the new. I was not on this project at that time, and the people most familiar with it have since left (hence me stumbling around blind here). But my understanding is that they took the project that was working in the old version, and published it using RoboHelp 2019, and thats when the search problems began. Which suggests to me there is something different about the way RH2019 handles search, or there is an incompatibility problem across versions that is manifesting as a search anomaly.
I understand your logic but on the other hand it the way search has changed is the problem, I would expect it to happen in all projects for all users. It is not happening for you in the supplied sample and others have not reported it.
Looking at the new output and viewing the page source, I see the heading, rather than the title, has a space in it. I don't think that should be an issue as heading and title are different things. Worth trying though.
<h1> Court-Ordered Obligations</h1>
Assuming that's not it, can you share the project? If you can, then see the Contact page on my site and send the project as instructed there. Do make sure you include a link to this thread and please do not email the project direct.
I finally got your sample help to work (for some reason it wasn't working at all "at work") and I see in the new output that random text seems to be bolded in the search results, rather than the words you searched for, or similar or parts of those words.
e.g. "you can prin t a request"
Which, if it was only happening on the server, would suggest some files related to the search weren't being copied up to the server correctly (I've seen whacky results in this case with RH11 output). But you say it happens when you freshly generate and view the project locally.
EDIT: oh wow, I just clicked on Parole/Probation Reports and the whole topic gets highlighted as a search hit. So weird.
@Peter, do you think it would be worth generating to a brand new folder, and perhaps also creating a new, fresh preset to, to see if that makes a difference?
A fresh generation to a new folder on the server could be worth trying. Steve has sent me his project but One Drive is messing up the download so what I have is unreliable. Steve is going to zip up the project before sending it again.
I wouldn't delete the existing server content just yet though in case his predecessor had put stuff there manually that the project depends on. Wait until the new upload is working on the server.
So after some continued troubleshooting it appears the issue may not be with RoboHelp directly, but with something beyond the generating/publishing step. The search appears to function properly after generating. The problems seem to arise only once the generated copy is published. I will continue troubleshooting on our end as time allows, and if something relevant to the community comes of it I will post back here, but otherwise this can probably be considered resolved. Thank you to everyone who assisted, but espescially to Peter.
Did anyone every identify a resolution to this? We're having a difficult time fully understanding - so we can then fully explain to our end users - how search functions (that is, weights/ranks results) in RH 2020.3. This used to be clear in 2017 and older/earlier (it was documented somewhere - for example: topic title first, then keywords, then h1, then h2, then h3, then h4, etc.) and now it's anyone's guess.
The results of the search are sorted using a score which is computed using BM25 algorithm along with other factors such as boost, proximity of keywords etc. In summary according to BM25 algorithm, the more a search term occurs in a single topic, the more that term will increase that topic’s score, but the more a search term occurs in the overall collection of topics, the less that term will increase a topic’s score.
For example, if a term occurs in several topics but in one topic it occurs very frequently then that topic will have higher score and will be displayed before other topics in the search result. But the frequency is not the only thing that is used to compute the topic’s score. Search term’s location also contributes the score of the topic. Following list is list of location in descending order of there boost value. Location with higher boost value adds more to the topic’s score.
In case search query has multiple terms then number of terms present in the document and there closeness also improves the score of the topic. For example of if you search for “Health and Safety” then any document which has both the keywords present will be ranked higher than the documents which have only one keyword present.
Closeness of the keywords plays a major role in the scoring. Any document which has all the keywords (within a window threshold of 50 characters) will always be ranked higher with all other documents which don’t have all the keywords, even if they are present in the body of the document and other documents have keyword in title or headings.
These are the main factors which decide the rank of the document in the search results. Hope this clears your doubts.
Please let us know if you any more queries.
This is good - thank you!