-
Notifications
You must be signed in to change notification settings - Fork 3
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Improvements to search #258
Comments
@ortnever @fabiancresson @PDumasBAR I've created this meta issue to have all the issues at the same place. Search is pretty hard to get right, after trying to tweak the original code, I've tried to use another technique, you can test it here : https://visualize-ad-search-rfimjjlwpq.herokuapp.com/en/browse, I have the feeling that the results are better, what do you think ? |
If you have weird results, do not hesitate to post the search query, so that I can list all the queries that you want to make. |
I just tried a few queries and it seems to require a lot of precision to work:
Another one:
|
Thanks for the feedback jeremy, it is true, I will look a bit more into it. |
This should be better now :) |
I tried to search the MFM-U Dataset (draft) I find the dataset with keyword "Air" |
For what I see, it works only on complete word. |
@ortnever This is due to how the search engine cuts word, I will look into it, thanks for the example. @PDumasBAR When I search for "Nois", I get the correct results (also with "noi" as shown below), are you sure you are using the deployment preview : https://visualize-ad-search-rfimjjlwpq.herokuapp.com/en/browse?includeDrafts=true&order=TITLE_ASC&search=noi |
OK I think the "search" is not finished and we must wait a little before testing. If I search "noi", it works, but the next search with "noise" doesn't work. Maybe my previous commentar was about that, first "noise" -> ok, then "noi" -> not ok, without reloading the page. |
Thanks for the feedback Pierre.
The search for now is very crude technically and I am exploring options to improve it. Search depends a lot on what you expect the tool to be able to search. This is why I try to involve you as early as possible, and regroup the used search terms at the top. For example I learn that BAFU searches datasets with acronyms. This might differ from generic users that would search with common words.
This is weird, it works for me and for @bprusinowski. I tested in Firefox and in Chrome, do you have more information ? |
Hi Patrick, I'm using EDGE, Version 95.0.1020.44 (Official build) (64-bit)
That's what I say. Not that the functionality is not good. Only it's a work in progress and testing make not so much sense until it reach a little more stable status. If it can helps, I think the search must react the same way as on opendata.swiss where the search look in all displayed fields. Of course title and description, but also in category, organization or tag (make less sense in visualize where the tags are not displayed). The opendata search is also "multi words" and coupled with an auto-complete list, given the user the opportunity to go directly to a matching dataset. I think this is the good direction. But also, only nice to have :) |
about "key word search - strange ordering of the results" #251 |
We (Vero and I) did some testing this morning and I want to summarize the behaviour we would like to see:
|
Thank you very much for taking the time to test the search and share your valuable feedback. The search functionality we have right now is implemented on a very high level: It loads all the datasets and the filtering/search is done on the server which is why the results shown may not be 100% relevant to the search query/keyword. We agree that there is potential to improve this feature. To apply the behaviour you described above, and make the search scalable and more performant, we would have to implement a full text search functionality from the database. @ptbrowne did some testing last year, and based on what we know at the moment, we assume that implementing a full text search functionality that aligns with the expected behaviour is technically feasible, but we would need to first invest some more time exploring this option to validate this assumption. We estimate that we could further explore the possibilities and implement a full text search in the course of one Sprint. If this is a priority, it is something we could consider for the final sprint (3.5). |
@sosiology do you mean that the search is not done with a SPARQL query ? |
Yep, the search has not been implemented with a SPARQL query at the moment. It should be done via the full text search functionality of stardog in the future. |
and one more question @ptbrowne about point 1 of @Mbch here above: is it not easy to "disable" to fact that the search is done with letters of the input word instead of words themselves ? I guess there is a kind of "fuzzy" search feature that could be disabled ? Or maybe say the smallest number of letters for the grouping ? |
Meta issue:
Search terms:
"Anzahl": #180
"Bade": #96
"OFPP"
"PRTR": #105
"MFM-U"
The text was updated successfully, but these errors were encountered: